VOYEUR: Merge of upstream

30 files changed, 1545 insertions, 1334 deletions

diff --git a/graphics/VectorRenderer.h b/graphics/VectorRenderer.h
index 0467cac946..8e1c5e91e1 100644
--- a/graphics/VectorRenderer.h
+++ b/graphics/VectorRenderer.h
@@ -73,6 +73,8 @@ struct DrawStep {
 	uint8 shadow, stroke, factor, radius, bevel; /**< Misc options... */
 
 	uint8 fillMode; /**< active fill mode */
+	uint8 shadowFillMode; /**< fill mode of the shadow used */
+
 	uint32 extraData; /**< Generic parameter for extra options (orientation/bevel) */
 
 	uint32 scale; /**< scale of all the coordinates in FIXED POINT with 16 bits mantissa */
@@ -103,7 +105,7 @@ VectorRenderer *createRenderer(int mode);
  */
 class VectorRenderer {
 public:
-	VectorRenderer() : _activeSurface(NULL), _fillMode(kFillDisabled), _shadowOffset(0),
+	VectorRenderer() : _activeSurface(NULL), _fillMode(kFillDisabled), _shadowOffset(0), _shadowFillMode(kShadowExponential), 
 		_disableShadows(false), _strokeWidth(1), _gradientFactor(1) {
 
 	}
@@ -126,6 +128,11 @@ public:
 		kTriangleRight
 	};
 
+	enum ShadowFillMode {
+		kShadowLinear = 0,
+		kShadowExponential = 1
+	};
+
 	/**
 	 * Draws a line by considering the special cases for optimization.
 	 *
@@ -278,7 +285,7 @@ public:
 	 * Clears the active surface.
 	 */
 	virtual void clearSurface() {
-		byte *src = (byte *)_activeSurface->pixels;
+		byte *src = (byte *)_activeSurface->getPixels();
 		memset(src, 0, _activeSurface->pitch * _activeSurface->h);
 	}
 
@@ -292,6 +299,10 @@ public:
 		_fillMode = mode;
 	}
 
+	virtual void setShadowFillMode(ShadowFillMode mode) {
+		_shadowFillMode = mode;
+	}
+
 	/**
 	 * Sets the stroke width. All shapes drawn with a stroke will
 	 * have that width. Pass 0 to disable shape stroking.
@@ -466,7 +477,7 @@ public:
 	 */
 	virtual void drawString(const Graphics::Font *font, const Common::String &text,
 	                        const Common::Rect &area, Graphics::TextAlign alignH,
-	                        GUI::ThemeEngine::TextAlignVertical alignV, int deltax, bool useEllipsis) = 0;
+	                        GUI::ThemeEngine::TextAlignVertical alignV, int deltax, bool useEllipsis, const Common::Rect &textDrawableArea) = 0;
 
 	/**
 	 * Allows to temporarily enable/disable all shadows drawing.
@@ -485,6 +496,7 @@ protected:
 	Surface *_activeSurface; /**< Pointer to the surface currently being drawn */
 
 	FillMode _fillMode; /**< Defines in which way (if any) are filled the drawn shapes */
+	ShadowFillMode _shadowFillMode;
 
 	int _shadowOffset; /**< offset for drawn shadows */
 	int _bevel; /**< amount of fake bevel */
diff --git a/graphics/VectorRendererSpec.cpp b/graphics/VectorRendererSpec.cpp
index 6a3ee306a5..491a9d7f42 100644
--- a/graphics/VectorRendererSpec.cpp
+++ b/graphics/VectorRendererSpec.cpp
@@ -119,6 +119,38 @@ inline frac_t fp_sqroot(uint32 x) {
 	*(ptr4 + (y) + (px)) = color2; \
 } while (0)
 
+#define BE_DRAWCIRCLE_BCOLOR_TR_CW(ptr,x,y,px,py,a) do { \
+	this->blendPixelPtr(ptr + (y) - (px), color, a); \
+} while (0)
+
+#define BE_DRAWCIRCLE_BCOLOR_TR_CCW(ptr,x,y,px,py,a) do { \
+	this->blendPixelPtr(ptr + (x) - (py), color, a); \
+} while (0)
+
+#define BE_DRAWCIRCLE_BCOLOR_TL_CW(ptr,x,y,px,py,a) do { \
+	this->blendPixelPtr(ptr - (x) - (py), color, a); \
+} while (0)
+
+#define BE_DRAWCIRCLE_BCOLOR_TL_CCW(ptr,x,y,px,py,a) do { \
+	this->blendPixelPtr(ptr - (y) - (px), color, a); \
+} while (0)
+
+#define BE_DRAWCIRCLE_BCOLOR_BL_CW(ptr,x,y,px,py,a) do { \
+	this->blendPixelPtr(ptr - (y) + (px), color, a); \
+} while (0)
+
+#define BE_DRAWCIRCLE_BCOLOR_BL_CCW(ptr,x,y,px,py,a) do { \
+	this->blendPixelPtr(ptr - (x) + (py), color, a); \
+} while (0)
+
+#define BE_DRAWCIRCLE_BCOLOR_BR_CW(ptr,x,y,px,py,a) do { \
+	this->blendPixelPtr(ptr + (x) + (py), color, a); \
+} while (0)
+
+#define BE_DRAWCIRCLE_BCOLOR_BR_CCW(ptr,x,y,px,py,a) do { \
+	this->blendPixelPtr(ptr + (y) + (px), color, a); \
+} while (0)
+
 #define BE_DRAWCIRCLE_XCOLOR_TOP(ptr1,ptr2,x,y,px,py) do { \
 	*(ptr1 + (y) - (px)) = color1; \
 	*(ptr1 + (x) - (py)) = color2; \
@@ -222,6 +254,37 @@ inline frac_t fp_sqroot(uint32 x) {
 	this->blendPixelPtr(ptr4 + (y) + (px), color2, a); \
 } while (0)
 
+#define WU_DRAWCIRCLE_BCOLOR_TR_CW(ptr,x,y,px,py,a) do { \
+	this->blendPixelPtr(ptr + (y) - (px), color, a); \
+} while (0)
+
+#define WU_DRAWCIRCLE_BCOLOR_TR_CCW(ptr,x,y,px,py,a) do { \
+	this->blendPixelPtr(ptr + (x) - (py), color, a); \
+} while (0)
+
+#define WU_DRAWCIRCLE_BCOLOR_TL_CW(ptr,x,y,px,py,a) do { \
+	this->blendPixelPtr(ptr - (x) - (py), color, a); \
+} while (0)
+
+#define WU_DRAWCIRCLE_BCOLOR_TL_CCW(ptr,x,y,px,py,a) do { \
+	this->blendPixelPtr(ptr - (y) - (px), color, a); \
+} while (0)
+
+#define WU_DRAWCIRCLE_BCOLOR_BL_CW(ptr,x,y,px,py,a) do { \
+	this->blendPixelPtr(ptr - (y) + (px), color, a); \
+} while (0)
+
+#define WU_DRAWCIRCLE_BCOLOR_BL_CCW(ptr,x,y,px,py,a) do { \
+	this->blendPixelPtr(ptr - (x) + (py), color, a); \
+} while (0)
+
+#define WU_DRAWCIRCLE_BCOLOR_BR_CW(ptr,x,y,px,py,a) do { \
+	this->blendPixelPtr(ptr + (x) + (py), color, a); \
+} while (0)
+
+#define WU_DRAWCIRCLE_BCOLOR_BR_CCW(ptr,x,y,px,py,a) do { \
+	this->blendPixelPtr(ptr + (y) + (px), color, a); \
+} while (0)
 
 // optimized Wu's algorithm
 #define WU_ALGORITHM() do { \
@@ -277,16 +340,24 @@ void colorFill(PixelType *first, PixelType *last, PixelType color) {
 
 VectorRenderer *createRenderer(int mode) {
 #ifdef DISABLE_FANCY_THEMES
-	assert(mode == GUI::ThemeEngine::kGfxStandard16bit);
+	assert(mode == GUI::ThemeEngine::kGfxStandard);
 #endif
 
 	PixelFormat format = g_system->getOverlayFormat();
 	switch (mode) {
-	case GUI::ThemeEngine::kGfxStandard16bit:
-		return new VectorRendererSpec<OverlayColor>(format);
+	case GUI::ThemeEngine::kGfxStandard:
+		if (g_system->getOverlayFormat().bytesPerPixel == 4)
+			return new VectorRendererSpec<uint32>(format);
+		else if (g_system->getOverlayFormat().bytesPerPixel == 2)
+			return new VectorRendererSpec<uint16>(format);
+		break;
 #ifndef DISABLE_FANCY_THEMES
-	case GUI::ThemeEngine::kGfxAntialias16bit:
-		return new VectorRendererAA<OverlayColor>(format);
+	case GUI::ThemeEngine::kGfxAntialias:
+		if (g_system->getOverlayFormat().bytesPerPixel == 4)
+			return new VectorRendererAA<uint32>(format);
+		else if (g_system->getOverlayFormat().bytesPerPixel == 2)
+			return new VectorRendererAA<uint16>(format);
+		break;
 #endif
 	default:
 		break;
@@ -317,9 +388,15 @@ setGradientColors(uint8 r1, uint8 g1, uint8 b1, uint8 r2, uint8 g2, uint8 b2) {
 	_gradientEnd = _format.RGBToColor(r2, g2, b2);
 	_gradientStart = _format.RGBToColor(r1, g1, b1);
 
-	_gradientBytes[0] = (_gradientEnd & _redMask) - (_gradientStart & _redMask);
-	_gradientBytes[1] = (_gradientEnd & _greenMask) - (_gradientStart & _greenMask);
-	_gradientBytes[2] = (_gradientEnd & _blueMask) - (_gradientStart & _blueMask);
+	if (sizeof(PixelType) == 4) {
+		_gradientBytes[0] = ((_gradientEnd & _redMask) >> _format.rShift) - ((_gradientStart & _redMask) >> _format.rShift);
+		_gradientBytes[1] = ((_gradientEnd & _greenMask) >> _format.gShift) - ((_gradientStart & _greenMask) >> _format.gShift);
+		_gradientBytes[2] = ((_gradientEnd & _blueMask) >> _format.bShift) - ((_gradientStart & _blueMask) >> _format.bShift);
+	} else {
+		_gradientBytes[0] = (_gradientEnd & _redMask) - (_gradientStart & _redMask);
+		_gradientBytes[1] = (_gradientEnd & _greenMask) - (_gradientStart & _greenMask);
+		_gradientBytes[2] = (_gradientEnd & _blueMask) - (_gradientStart & _blueMask);
+	}
 }
 
 template<typename PixelType>
@@ -328,9 +405,15 @@ calcGradient(uint32 pos, uint32 max) {
 	PixelType output = 0;
 	pos = (MIN(pos * Base::_gradientFactor, max) << 12) / max;
 
-	output |= ((_gradientStart & _redMask) + ((_gradientBytes[0] * pos) >> 12)) & _redMask;
-	output |= ((_gradientStart & _greenMask) + ((_gradientBytes[1] * pos) >> 12)) & _greenMask;
-	output |= ((_gradientStart & _blueMask) + ((_gradientBytes[2] * pos) >> 12)) & _blueMask;
+	if (sizeof(PixelType) == 4) {
+		output |= ((_gradientStart & _redMask) + (((_gradientBytes[0] * pos) >> 12) << _format.rShift)) & _redMask;
+		output |= ((_gradientStart & _greenMask) + (((_gradientBytes[1] * pos) >> 12) << _format.gShift)) & _greenMask;
+		output |= ((_gradientStart & _blueMask) + (((_gradientBytes[2] * pos) >> 12) << _format.bShift)) & _blueMask;
+	} else {
+		output |= ((_gradientStart & _redMask) + ((_gradientBytes[0] * pos) >> 12)) & _redMask;
+		output |= ((_gradientStart & _greenMask) + ((_gradientBytes[1] * pos) >> 12)) & _greenMask;
+		output |= ((_gradientStart & _blueMask) + ((_gradientBytes[2] * pos) >> 12)) & _blueMask;
+	}
 	output |= _alphaMask;
 
 	return output;
@@ -397,7 +480,7 @@ gradientFill(PixelType *ptr, int width, int x, int y) {
 template<typename PixelType>
 void VectorRendererSpec<PixelType>::
 fillSurface() {
-	byte *ptr = (byte *)_activeSurface->getBasePtr(0, 0);
+	byte *ptr = (byte *)_activeSurface->getPixels();
 
 	int h = _activeSurface->h;
 	int pitch = _activeSurface->pitch;
@@ -453,7 +536,7 @@ template<typename PixelType>
 void VectorRendererSpec<PixelType>::
 blitSubSurface(const Graphics::Surface *source, const Common::Rect &r) {
 	byte *dst_ptr = (byte *)_activeSurface->getBasePtr(r.left, r.top);
-	const byte *src_ptr = (const byte *)source->getBasePtr(0, 0);
+	const byte *src_ptr = (const byte *)source->getPixels();
 
 	const int dst_pitch = _activeSurface->pitch;
 	const int src_pitch = source->pitch;
@@ -481,7 +564,7 @@ blitAlphaBitmap(const Graphics::Surface *source, const Common::Rect &r) {
 		y = y + (r.height() >> 1) - (source->h >> 1);
 
 	PixelType *dst_ptr = (PixelType *)_activeSurface->getBasePtr(x, y);
-	const PixelType *src_ptr = (const PixelType *)source->getBasePtr(0, 0);
+	const PixelType *src_ptr = (const PixelType *)source->getPixels();
 
 	int dst_pitch = _activeSurface->pitch / _activeSurface->format.bytesPerPixel;
 	int src_pitch = source->pitch / source->format.bytesPerPixel;
@@ -508,7 +591,7 @@ template<typename PixelType>
 void VectorRendererSpec<PixelType>::
 applyScreenShading(GUI::ThemeEngine::ShadingStyle shadingStyle) {
 	int pixels = _activeSurface->w * _activeSurface->h;
-	PixelType *ptr = (PixelType *)_activeSurface->getBasePtr(0, 0);
+	PixelType *ptr = (PixelType *)_activeSurface->getPixels();
 	uint8 r, g, b;
 	uint lum;
 
@@ -537,20 +620,41 @@ applyScreenShading(GUI::ThemeEngine::ShadingStyle shadingStyle) {
 template<typename PixelType>
 inline void VectorRendererSpec<PixelType>::
 blendPixelPtr(PixelType *ptr, PixelType color, uint8 alpha) {
-	int idst = *ptr;
-	int isrc = color;
-
-	*ptr = (PixelType)(
-		(_redMask & ((idst & _redMask) +
-		((int)(((int)(isrc & _redMask) -
-		(int)(idst & _redMask)) * alpha) >> 8))) |
-		(_greenMask & ((idst & _greenMask) +
-		((int)(((int)(isrc & _greenMask) -
-		(int)(idst & _greenMask)) * alpha) >> 8))) |
-		(_blueMask & ((idst & _blueMask) +
-		((int)(((int)(isrc & _blueMask) -
-		(int)(idst & _blueMask)) * alpha) >> 8))) |
-		(idst & _alphaMask));
+	if (sizeof(PixelType) == 4) {
+		const byte sR = (color & _redMask) >> _format.rShift;
+		const byte sG = (color & _greenMask) >> _format.gShift;
+		const byte sB = (color & _blueMask) >> _format.bShift;
+
+		byte dR = (*ptr & _redMask) >> _format.rShift;
+		byte dG = (*ptr & _greenMask) >> _format.gShift;
+		byte dB = (*ptr & _blueMask) >> _format.bShift;
+
+		dR += ((sR - dR) * alpha) >> 8;
+		dG += ((sG - dG) * alpha) >> 8;
+		dB += ((sB - dB) * alpha) >> 8;
+
+		*ptr = ((dR << _format.rShift) & _redMask)
+		     | ((dG << _format.gShift) & _greenMask)
+		     | ((dB << _format.bShift) & _blueMask)
+		     | (*ptr & _alphaMask);
+	} else if (sizeof(PixelType) == 2) {
+		int idst = *ptr;
+		int isrc = color;
+
+		*ptr = (PixelType)(
+			(_redMask & ((idst & _redMask) +
+			((int)(((int)(isrc & _redMask) -
+			(int)(idst & _redMask)) * alpha) >> 8))) |
+			(_greenMask & ((idst & _greenMask) +
+			((int)(((int)(isrc & _greenMask) -
+			(int)(idst & _greenMask)) * alpha) >> 8))) |
+			(_blueMask & ((idst & _blueMask) +
+			((int)(((int)(isrc & _blueMask) -
+			(int)(idst & _blueMask)) * alpha) >> 8))) |
+			(idst & _alphaMask));
+	} else {
+		error("Unsupported BPP format: %u", (uint)sizeof(PixelType));
+	}
 }
 
 template<typename PixelType>
@@ -607,24 +711,45 @@ darkenFill(PixelType *ptr, PixelType *end) {
 template<typename PixelType>
 void VectorRendererSpec<PixelType>::
 drawString(const Graphics::Font *font, const Common::String &text, const Common::Rect &area,
-			Graphics::TextAlign alignH, GUI::ThemeEngine::TextAlignVertical alignV, int deltax, bool ellipsis) {
+			Graphics::TextAlign alignH, GUI::ThemeEngine::TextAlignVertical alignV, int deltax, bool ellipsis, const Common::Rect &textDrawableArea) {
 
 	int offset = area.top;
 
 	if (font->getFontHeight() < area.height()) {
 		switch (alignV) {
-			case GUI::ThemeEngine::kTextAlignVCenter:
-				offset = area.top + ((area.height() - font->getFontHeight()) >> 1);
-				break;
-			case GUI::ThemeEngine::kTextAlignVBottom:
-				offset = area.bottom - font->getFontHeight();
-				break;
-			default:
-				break;
+		case GUI::ThemeEngine::kTextAlignVCenter:
+			offset = area.top + ((area.height() - font->getFontHeight()) >> 1);
+			break;
+		case GUI::ThemeEngine::kTextAlignVBottom:
+			offset = area.bottom - font->getFontHeight();
+			break;
+		default:
+			break;
 		}
 	}
 
-	font->drawString(_activeSurface, text, area.left, offset, area.width() - deltax, _fgColor, alignH, deltax, ellipsis);
+	Common::Rect drawArea;
+	if (textDrawableArea.isEmpty()) {
+		// In case no special area to draw to is given we only draw in the
+		// area specified by the user.
+		drawArea = area;
+		// warning("there is no text drawable area. Please set this area for clipping");
+	} else {
+		// The area we can draw to is the intersection between the allowed
+		// drawing area (textDrawableArea) and the area where we try to draw
+		// the text (area).
+		drawArea = textDrawableArea.findIntersectingRect(area);
+	}
+
+	// Better safe than sorry. We intersect with the actual surface boundaries
+	// to avoid any ugly clipping in _activeSurface->getSubArea which messes
+	// up the calculation of the x and y coordinates where to draw the string.
+	drawArea = drawArea.findIntersectingRect(Common::Rect(0, 0, _activeSurface->w, _activeSurface->h));
+
+	if (!drawArea.isEmpty()) {
+		Surface textAreaSurface = _activeSurface->getSubArea(drawArea);
+		font->drawString(&textAreaSurface, text, area.left - drawArea.left, offset - drawArea.top, area.width() - deltax, _fgColor, alignH, deltax, ellipsis);
+	}
 }
 
 /** LINES **/
@@ -778,7 +903,8 @@ drawRoundedSquare(int x, int y, int r, int w, int h) {
 
 	if (Base::_fillMode != kFillDisabled && Base::_shadowOffset
 		&& x + w + Base::_shadowOffset + 1 < Base::_activeSurface->w
-		&& y + h + Base::_shadowOffset + 1 < Base::_activeSurface->h) {
+		&& y + h + Base::_shadowOffset + 1 < Base::_activeSurface->h
+		&& h > (Base::_shadowOffset + 1) * 2) {
 		drawRoundedSquareShadow(x, y, r, w, h, Base::_shadowOffset);
 	}
 
@@ -809,13 +935,14 @@ drawTab(int x, int y, int r, int w, int h) {
 			// FIXME: This is broken for the AA renderer.
 			// See the rounded rect alg for how to fix it. (The border should
 			// be drawn before the interior, both inside drawTabAlg.)
-			drawTabAlg(x, y, w, h, r, (Base::_fillMode == kFillBackground) ? _bgColor : _fgColor, Base::_fillMode);
+			drawTabShadow(x, y, w - 2, h, r);
+			drawTabAlg(x, y, w - 2, h, r, _bgColor, Base::_fillMode);
 			if (Base::_strokeWidth)
 				drawTabAlg(x, y, w, h, r, _fgColor, kFillDisabled, (Base::_dynamicData >> 16), (Base::_dynamicData & 0xFFFF));
 			break;
 
 		case kFillForeground:
-			drawTabAlg(x, y, w, h, r, (Base::_fillMode == kFillBackground) ? _bgColor : _fgColor, Base::_fillMode);
+			drawTabAlg(x, y, w, h, r, _fgColor, Base::_fillMode);
 			break;
 	}
 }
@@ -996,6 +1123,67 @@ drawTabAlg(int x1, int y1, int w, int h, int r, PixelType color, VectorRenderer:
 }
 
 
+template<typename PixelType>
+void VectorRendererSpec<PixelType>::
+drawTabShadow(int x1, int y1, int w, int h, int r) {
+	int offset = 3;
+	int pitch = _activeSurface->pitch / _activeSurface->format.bytesPerPixel;
+	
+	// "Harder" shadows when having lower BPP, since we will have artifacts (greenish tint on the modern theme)
+	uint8 expFactor = 3;
+	uint16 alpha = (_activeSurface->format.bytesPerPixel > 2) ? 4 : 8;
+
+	int xstart = x1;
+	int ystart = y1;
+	int width = w;
+	int height = h + offset + 1;
+
+	for (int i = offset; i >= 0; i--) {
+		int f, ddF_x, ddF_y;
+		int x, y, px, py;
+
+		PixelType *ptr_tl = (PixelType *)Base::_activeSurface->getBasePtr(xstart + r, ystart + r);
+		PixelType *ptr_tr = (PixelType *)Base::_activeSurface->getBasePtr(xstart + width - r, ystart + r);
+		PixelType *ptr_fill = (PixelType *)Base::_activeSurface->getBasePtr(xstart, ystart);
+
+		int short_h = height - (2 * r) + 2;
+		PixelType color = _format.RGBToColor(0, 0, 0);
+
+		BE_RESET();
+
+		// HACK: As we are drawing circles exploting 8-axis symmetry,
+		// there are 4 pixels on each circle which are drawn twice.
+		// this is ok on filled circles, but when blending on surfaces,
+		// we cannot let it blend twice. awful.
+		uint32 hb = 0;
+		
+		while (x++ < y) {
+			BE_ALGORITHM();
+
+			if (((1 << x) & hb) == 0) {
+				blendFill(ptr_tl - y - px, ptr_tr + y - px, color, (uint8)alpha);
+				hb |= (1 << x);
+			}
+
+			if (((1 << y) & hb) == 0) {
+				blendFill(ptr_tl - x - py, ptr_tr + x - py, color, (uint8)alpha);
+				hb |= (1 << y);
+			}
+		}
+	
+		ptr_fill += pitch * r;
+		while (short_h--) {
+			blendFill(ptr_fill, ptr_fill + width + 1, color, (uint8)alpha);
+			ptr_fill += pitch;
+		}
+
+		// Move shadow one pixel upward each iteration
+		xstart += 1;
+		// Multiply with expfactor
+		alpha = (alpha * (expFactor << 8)) >> 9;
+	}
+}
+	
 /** BEVELED TABS FOR CLASSIC THEME **/
 template<typename PixelType>
 void VectorRendererSpec<PixelType>::
@@ -1429,118 +1617,160 @@ drawTriangleFast(int x1, int y1, int size, bool inverted, PixelType color, Vecto
 /** ROUNDED SQUARE ALGORITHM **/
 template<typename PixelType>
 void VectorRendererSpec<PixelType>::
-drawRoundedSquareAlg(int x1, int y1, int r, int w, int h, PixelType color, VectorRenderer::FillMode fill_m) {
+drawBorderRoundedSquareAlg(int x1, int y1, int r, int w, int h, PixelType color, VectorRenderer::FillMode fill_m, uint8 alpha_t, uint8 alpha_r, uint8 alpha_b, uint8 alpha_l) {
 	int f, ddF_x, ddF_y;
 	int x, y, px, py;
 	int pitch = _activeSurface->pitch / _activeSurface->format.bytesPerPixel;
+	int sw = 0, sp = 0, hp = h * pitch;
 
-	// TODO: Split this up into border, bevel and interior functions
+	PixelType *ptr_tl = (PixelType *)Base::_activeSurface->getBasePtr(x1 + r, y1 + r);
+	PixelType *ptr_tr = (PixelType *)Base::_activeSurface->getBasePtr(x1 + w - r, y1 + r);
+	PixelType *ptr_bl = (PixelType *)Base::_activeSurface->getBasePtr(x1 + r, y1 + h - r);
+	PixelType *ptr_br = (PixelType *)Base::_activeSurface->getBasePtr(x1 + w - r, y1 + h - r);
+	PixelType *ptr_fill = (PixelType *)Base::_activeSurface->getBasePtr(x1, y1);
 
-	if (fill_m != kFillDisabled) {
-		PixelType *ptr_tl = (PixelType *)Base::_activeSurface->getBasePtr(x1 + r, y1 + r);
-		PixelType *ptr_tr = (PixelType *)Base::_activeSurface->getBasePtr(x1 + w - r, y1 + r);
-		PixelType *ptr_bl = (PixelType *)Base::_activeSurface->getBasePtr(x1 + r, y1 + h - r);
-		PixelType *ptr_br = (PixelType *)Base::_activeSurface->getBasePtr(x1 + w - r, y1 + h - r);
-		PixelType *ptr_fill = (PixelType *)Base::_activeSurface->getBasePtr(x1, y1);
-
-		int real_radius = r;
-		int short_h = h - (2 * r) + 2;
-		int long_h = h;
+	int real_radius = r;
+	int short_h = h - (2 * r) + 2;
 
-		BE_RESET();
+	PixelType color1 = color;
+	PixelType color2 = color;
 
-		PixelType color1 = color;
-		if (fill_m == kFillBackground)
-			color1 = _bgColor;
+	while (sw++ < Base::_strokeWidth) {
+		blendFill(ptr_fill + sp + r, ptr_fill + w + 1 + sp - r, color1, alpha_t); // top
+		blendFill(ptr_fill + hp - sp + r, ptr_fill + w + hp + 1 - sp - r, color2, alpha_b); // bottom
+		sp += pitch;
 
-		if (fill_m == kFillGradient) {
-			PixelType color2, color3, color4;
-			precalcGradient(long_h);
+		BE_RESET();
+		r--;
+		
+		int alphaStep_tr = ((alpha_t - alpha_r)/(y+1));
+		int alphaStep_br = ((alpha_r - alpha_b)/(y+1));
+		int alphaStep_bl = ((alpha_b - alpha_l)/(y+1));
+		int alphaStep_tl = ((alpha_l - alpha_t)/(y+1));
+
+		// Avoid blending the last pixels twice, since we have an alpha
+		while (x++ < (y - 2)) {
+			BE_ALGORITHM();
 
-			while (x++ < y) {
-				BE_ALGORITHM();
+			BE_DRAWCIRCLE_BCOLOR_TR_CW(ptr_tr, x, y, px, py, (uint8)(alpha_r + (alphaStep_tr * x))); 
+			BE_DRAWCIRCLE_BCOLOR_BR_CW(ptr_br, x, y, px, py, (uint8)(alpha_b + (alphaStep_br * x)));
+			BE_DRAWCIRCLE_BCOLOR_BL_CW(ptr_bl, x, y, px, py, (uint8)(alpha_l + (alphaStep_bl * x)));
+			BE_DRAWCIRCLE_BCOLOR_TL_CW(ptr_tl, x, y, px, py, (uint8)(alpha_t + (alphaStep_tl * x)));
+			
+			BE_DRAWCIRCLE_BCOLOR_TR_CCW(ptr_tr, x, y, px, py, (uint8)(alpha_t - (alphaStep_tr * x)));
+			BE_DRAWCIRCLE_BCOLOR_BR_CCW(ptr_br, x, y, px, py, (uint8)(alpha_r - (alphaStep_br * x)));
+			BE_DRAWCIRCLE_BCOLOR_BL_CCW(ptr_bl, x, y, px, py, (uint8)(alpha_b - (alphaStep_bl * x)));
+			BE_DRAWCIRCLE_BCOLOR_TL_CCW(ptr_tl, x, y, px, py, (uint8)(alpha_l - (alphaStep_tl * x)));
+			
+			if (Base::_strokeWidth > 1) {
+				BE_DRAWCIRCLE_BCOLOR(ptr_tr, ptr_tl, ptr_bl, ptr_br, x - 1, y, px, py);
+				BE_DRAWCIRCLE_BCOLOR(ptr_tr, ptr_tl, ptr_bl, ptr_br, x, y, px - pitch, py);
+			}
+		}
+	}
 
-				color1 = calcGradient(real_radius - x, long_h);
-				color2 = calcGradient(real_radius - y, long_h);
-				color3 = calcGradient(long_h - r + x, long_h);
-				color4 = calcGradient(long_h - r + y, long_h);
+	ptr_fill += pitch * real_radius;
+	while (short_h--) {
+		blendFill(ptr_fill, ptr_fill + Base::_strokeWidth, color1, alpha_l); // left
+		blendFill(ptr_fill + w - Base::_strokeWidth + 1, ptr_fill + w + 1, color2, alpha_r); // right
+		ptr_fill += pitch;
+	}
+}
 
-				gradientFill(ptr_tl - x - py, w - 2 * r + 2 * x, x1 + r - x - y, real_radius - y);
-				gradientFill(ptr_tl - y - px, w - 2 * r + 2 * y, x1 + r - y - x, real_radius - x);
+template<typename PixelType>
+void VectorRendererSpec<PixelType>::
+drawInteriorRoundedSquareAlg(int x1, int y1, int r, int w, int h, PixelType color, VectorRenderer::FillMode fill_m) {
+	int f, ddF_x, ddF_y;
+	int x, y, px, py;
+	int pitch = _activeSurface->pitch / _activeSurface->format.bytesPerPixel;
 
-				gradientFill(ptr_bl - x + py, w - 2 * r + 2 * x, x1 + r - x - y, long_h - r + y);
-				gradientFill(ptr_bl - y + px, w - 2 * r + 2 * y, x1 + r - y - x, long_h - r + x);
+	PixelType *ptr_tl = (PixelType *)Base::_activeSurface->getBasePtr(x1 + r, y1 + r);
+	PixelType *ptr_tr = (PixelType *)Base::_activeSurface->getBasePtr(x1 + w - r, y1 + r);
+	PixelType *ptr_bl = (PixelType *)Base::_activeSurface->getBasePtr(x1 + r, y1 + h - r);
+	PixelType *ptr_br = (PixelType *)Base::_activeSurface->getBasePtr(x1 + w - r, y1 + h - r);
+	PixelType *ptr_fill = (PixelType *)Base::_activeSurface->getBasePtr(x1, y1);
 
-				BE_DRAWCIRCLE_XCOLOR(ptr_tr, ptr_tl, ptr_bl, ptr_br, x, y, px, py);
-			}
-		} else {
-			while (x++ < y) {
-				BE_ALGORITHM();
+	int real_radius = r;
+	int short_h = h - (2 * r) + 2;
+	int long_h = h;
 
-				colorFill<PixelType>(ptr_tl - x - py, ptr_tr + x - py, color1);
-				colorFill<PixelType>(ptr_tl - y - px, ptr_tr + y - px, color1);
+	BE_RESET();
 
-				colorFill<PixelType>(ptr_bl - x + py, ptr_br + x + py, color1);
-				colorFill<PixelType>(ptr_bl - y + px, ptr_br + y + px, color1);
+	PixelType color1 = color;
 
-				// do not remove - messes up the drawing at lower resolutions
-				BE_DRAWCIRCLE(ptr_tr, ptr_tl, ptr_bl, ptr_br, x, y, px, py);
-			}
-		}
+	if (fill_m == kFillGradient) {
+		PixelType color2, color3, color4;
+		precalcGradient(long_h);
 
-		ptr_fill += pitch * r;
-		while (short_h--) {
-			if (fill_m == kFillGradient) {
-				gradientFill(ptr_fill, w + 1, x1, real_radius++);
-			} else {
-				colorFill<PixelType>(ptr_fill, ptr_fill + w + 1, color1);
-			}
-			ptr_fill += pitch;
-		}
-	}
+		while (x++ < y) {
+			BE_ALGORITHM();
 
+			color1 = calcGradient(real_radius - x, long_h);
+			color2 = calcGradient(real_radius - y, long_h);
+			color3 = calcGradient(long_h - r + x, long_h);
+			color4 = calcGradient(long_h - r + y, long_h);
 
-	if (Base::_strokeWidth) {
-		int sw = 0, sp = 0, hp = h * pitch;
+			gradientFill(ptr_tl - x - py, w - 2 * r + 2 * x, x1 + r - x - y, real_radius - y);
+			gradientFill(ptr_tl - y - px, w - 2 * r + 2 * y, x1 + r - y - x, real_radius - x);
 
-		PixelType *ptr_tl = (PixelType *)Base::_activeSurface->getBasePtr(x1 + r, y1 + r);
-		PixelType *ptr_tr = (PixelType *)Base::_activeSurface->getBasePtr(x1 + w - r, y1 + r);
-		PixelType *ptr_bl = (PixelType *)Base::_activeSurface->getBasePtr(x1 + r, y1 + h - r);
-		PixelType *ptr_br = (PixelType *)Base::_activeSurface->getBasePtr(x1 + w - r, y1 + h - r);
-		PixelType *ptr_fill = (PixelType *)Base::_activeSurface->getBasePtr(x1, y1);
+			gradientFill(ptr_bl - x + py, w - 2 * r + 2 * x, x1 + r - x - y, long_h - r + y);
+			gradientFill(ptr_bl - y + px, w - 2 * r + 2 * y, x1 + r - y - x, long_h - r + x);
 
-		int real_radius = r;
-		int short_h = h - (2 * r) + 2;
+			BE_DRAWCIRCLE_XCOLOR(ptr_tr, ptr_tl, ptr_bl, ptr_br, x, y, px, py);
+		}
+	} else {
+		while (x++ < y) {
+			BE_ALGORITHM();
 
-		// TODO: A gradient effect on the bevel
-		PixelType color1, color2;
-		color1 = Base::_bevel ? _bevelColor : color;
-		color2 = color;
+			colorFill<PixelType>(ptr_tl - x - py, ptr_tr + x - py, color1);
+			colorFill<PixelType>(ptr_tl - y - px, ptr_tr + y - px, color1);
 
-		while (sw++ < Base::_strokeWidth) {
-			colorFill<PixelType>(ptr_fill + sp + r, ptr_fill + w + 1 + sp - r, color1);
-			colorFill<PixelType>(ptr_fill + hp - sp + r, ptr_fill + w + hp + 1 - sp - r, color2);
-			sp += pitch;
+			colorFill<PixelType>(ptr_bl - x + py, ptr_br + x + py, color1);
+			colorFill<PixelType>(ptr_bl - y + px, ptr_br + y + px, color1);
 
-			BE_RESET();
-			r--;
-
-			while (x++ < y) {
-				BE_ALGORITHM();
-				BE_DRAWCIRCLE_BCOLOR(ptr_tr, ptr_tl, ptr_bl, ptr_br, x, y, px, py);
+			// do not remove - messes up the drawing at lower resolutions
+			BE_DRAWCIRCLE(ptr_tr, ptr_tl, ptr_bl, ptr_br, x, y, px, py);
+		}
+	}
 
-				if (Base::_strokeWidth > 1) {
-					BE_DRAWCIRCLE_BCOLOR(ptr_tr, ptr_tl, ptr_bl, ptr_br, x - 1, y, px, py);
-					BE_DRAWCIRCLE_BCOLOR(ptr_tr, ptr_tl, ptr_bl, ptr_br, x, y, px - pitch, py);
-				}
-			}
+	ptr_fill += pitch * r;
+	while (short_h--) {
+		if (fill_m == kFillGradient) {
+			gradientFill(ptr_fill, w + 1, x1, real_radius++);
+		} else {
+			colorFill<PixelType>(ptr_fill, ptr_fill + w + 1, color1);
 		}
+		ptr_fill += pitch;
+	}
+}
 
-		ptr_fill += pitch * real_radius;
-		while (short_h--) {
-			colorFill<PixelType>(ptr_fill, ptr_fill + Base::_strokeWidth, color1);
-			colorFill<PixelType>(ptr_fill + w - Base::_strokeWidth + 1, ptr_fill + w + 1, color2);
-			ptr_fill += pitch;
+template<typename PixelType>
+void VectorRendererSpec<PixelType>::
+drawRoundedSquareAlg(int x1, int y1, int r, int w, int h, PixelType color, VectorRenderer::FillMode fill_m) {
+	const uint8 borderAlpha_t = 0;
+	const uint8 borderAlpha_r = 127;
+	const uint8 borderAlpha_b = 255; 
+	const uint8 borderAlpha_l = 63;
+
+	const uint8 bevelAlpha_t = 255;
+	const uint8 bevelAlpha_r = 31;
+	const uint8 bevelAlpha_b = 0;
+	const uint8 bevelAlpha_l = 127;
+
+	// If only border is visible
+	if ((!(w <= 0 || h <= 0)) && (fill_m != Base::kFillDisabled)) {
+		if (fill_m == Base::kFillBackground)
+			drawInteriorRoundedSquareAlg(x1, y1, r, w, h, _bgColor, fill_m);
+		else
+			drawInteriorRoundedSquareAlg(x1, y1, r, w, h, color, fill_m);
+	}
+
+	if (Base::_strokeWidth) {
+		if (r != 0 && _bevel > 0) {
+			drawBorderRoundedSquareAlg(x1, y1, r, w, h, color, fill_m, borderAlpha_t, borderAlpha_r, borderAlpha_b, borderAlpha_l);
+			drawBorderRoundedSquareAlg(x1, y1, r, w, h, _bevelColor, fill_m, bevelAlpha_t, bevelAlpha_r, bevelAlpha_b, bevelAlpha_l);
+		} else {
+			drawBorderRoundedSquareAlg(x1, y1, r, w, h, color, fill_m, 255, 255, 255, 255);
 		}
 	}
 }
@@ -1599,85 +1829,112 @@ drawCircleAlg(int x1, int y1, int r, PixelType color, VectorRenderer::FillMode f
  ********************************************************************/
 template<typename PixelType>
 void VectorRendererSpec<PixelType>::
-drawSquareShadow(int x, int y, int w, int h, int blur) {
-	PixelType *ptr = (PixelType *)_activeSurface->getBasePtr(x + w - 1, y + blur);
+drawSquareShadow(int x, int y, int w, int h, int offset) {
+	PixelType *ptr = (PixelType *)_activeSurface->getBasePtr(x + w - 1, y + offset);
 	int pitch = _activeSurface->pitch / _activeSurface->format.bytesPerPixel;
 	int i, j;
 
-	i = h - blur;
+	i = h - offset;
 
 	while (i--) {
-		j = blur;
+		j = offset;
 		while (j--)
-			blendPixelPtr(ptr + j, 0, ((blur - j) << 8) / blur);
+			blendPixelPtr(ptr + j, 0, ((offset - j) << 8) / offset);
 		ptr += pitch;
 	}
 
-	ptr = (PixelType *)_activeSurface->getBasePtr(x + blur, y + h - 1);
+	ptr = (PixelType *)_activeSurface->getBasePtr(x + offset, y + h - 1);
 
-	while (i++ < blur) {
-		j = w - blur;
+	while (i++ < offset) {
+		j = w - offset;
 		while (j--)
-			blendPixelPtr(ptr + j, 0, ((blur - i) << 8) / blur);
+			blendPixelPtr(ptr + j, 0, ((offset - i) << 8) / offset);
 		ptr += pitch;
 	}
 
 	ptr = (PixelType *)_activeSurface->getBasePtr(x + w, y + h);
 
 	i = 0;
-	while (i++ < blur) {
-		j = blur - 1;
+	while (i++ < offset) {
+		j = offset - 1;
 		while (j--)
-			blendPixelPtr(ptr + j, 0, (((blur - j) * (blur - i)) << 8) / (blur * blur));
+			blendPixelPtr(ptr + j, 0, (((offset - j) * (offset - i)) << 8) / (offset * offset));
 		ptr += pitch;
 	}
 }
 
 template<typename PixelType>
 void VectorRendererSpec<PixelType>::
-drawRoundedSquareShadow(int x1, int y1, int r, int w, int h, int blur) {
-	int f, ddF_x, ddF_y;
-	int x, y, px, py;
+drawRoundedSquareShadow(int x1, int y1, int r, int w, int h, int offset) {
 	int pitch = _activeSurface->pitch / _activeSurface->format.bytesPerPixel;
-	int alpha = 102;
 
-	x1 += blur;
-	y1 += blur;
+	// "Harder" shadows when having lower BPP, since we will have artifacts (greenish tint on the modern theme)
+	uint8 expFactor = 3;
+	uint16 alpha = (_activeSurface->format.bytesPerPixel > 2) ? 4 : 8;
+	
+	// These constants ensure a border of 2px on the left and of each rounded square
+	int xstart = (x1 > 2) ? x1 - 2 : x1;
+	int ystart = y1;
+	int width = w + offset + 2;
+	int height = h + offset + 1;
+
+	for (int i = offset; i >= 0; i--) {
+		int f, ddF_x, ddF_y;
+		int x, y, px, py;
+		
+		PixelType *ptr_tl = (PixelType *)Base::_activeSurface->getBasePtr(xstart + r, ystart + r);
+		PixelType *ptr_tr = (PixelType *)Base::_activeSurface->getBasePtr(xstart + width - r, ystart + r);
+		PixelType *ptr_bl = (PixelType *)Base::_activeSurface->getBasePtr(xstart + r, ystart + height - r);
+		PixelType *ptr_br = (PixelType *)Base::_activeSurface->getBasePtr(xstart + width - r, ystart + height - r);
+		PixelType *ptr_fill = (PixelType *)Base::_activeSurface->getBasePtr(xstart, ystart);
+
+		int short_h = height - (2 * r) + 2;
+		PixelType color = _format.RGBToColor(0, 0, 0);
 
-	PixelType *ptr_tr = (PixelType *)Base::_activeSurface->getBasePtr(x1 + w - r, y1 + r);
-	PixelType *ptr_bl = (PixelType *)Base::_activeSurface->getBasePtr(x1 + r, y1 + h - r);
-	PixelType *ptr_br = (PixelType *)Base::_activeSurface->getBasePtr(x1 + w - r, y1 + h - r);
-	PixelType *ptr_fill = (PixelType *)Base::_activeSurface->getBasePtr(x1 + w - blur, y1 + r);
-
-	int short_h = h - (2 * r) + 1;
+		BE_RESET();
 
-	BE_RESET();
+		// HACK: As we are drawing circles exploting 8-axis symmetry,
+		// there are 4 pixels on each circle which are drawn twice.
+		// this is ok on filled circles, but when blending on surfaces,
+		// we cannot let it blend twice. awful.
+		uint32 hb = 0;
+		
+		while (x++ < y) {
+			BE_ALGORITHM();
 
-	// HACK: As we are drawing circles exploting 8-axis symmetry,
-	// there are 4 pixels on each circle which are drawn twice.
-	// this is ok on filled circles, but when blending on surfaces,
-	// we cannot let it blend twice. awful.
-	uint32 hb = 0;
 
-	while (x++ < y) {
-		BE_ALGORITHM();
+			if (((1 << x) & hb) == 0) {
+				blendFill(ptr_tl - y - px, ptr_tr + y - px, color, (uint8)alpha);
+				
+				// Will create a dark line of pixles if left out
+				if (hb > 0) {
+					blendFill(ptr_bl - y + px, ptr_br + y + px, color, (uint8)alpha);
+				}
+				hb |= (1 << x);
+			}
 
-		if (((1 << x) & hb) == 0) {
-			blendFill(ptr_tr - px - r, ptr_tr + y - px, 0, alpha);
-			blendFill(ptr_bl - y + px, ptr_br + y + px, 0, alpha);
-			hb |= (1 << x);
+			if (((1 << y) & hb) == 0) {
+				blendFill(ptr_tl - x - py, ptr_tr + x - py, color, (uint8)alpha);
+				blendFill(ptr_bl - x + py, ptr_br + x + py, color, (uint8)alpha);
+				hb |= (1 << y);
+			}
 		}
-
-		if (((1 << y) & hb) == 0) {
-			blendFill(ptr_tr - r - py, ptr_tr + x - py, 0, alpha);
-			blendFill(ptr_bl - x + py, ptr_br + x + py, 0, alpha);
-			hb |= (1 << y);
+	
+		ptr_fill += pitch * r;
+		while (short_h--) {
+			blendFill(ptr_fill, ptr_fill + width + 1, color, (uint8)alpha);
+			ptr_fill += pitch;
 		}
-	}
 
-	while (short_h--) {
-		blendFill(ptr_fill - r, ptr_fill + blur, 0, alpha);
-		ptr_fill += pitch;
+		// Make shadow smaller each iteration, and move it one pixel inward
+		xstart += 1;
+		ystart += 1;
+		width -= 2;
+		height -= 2;
+		
+		if (_shadowFillMode == kShadowExponential)
+			// Multiply with expfactor
+			alpha = (alpha * (expFactor << 8)) >> 9;
 	}
 }
 
@@ -1870,7 +2127,7 @@ drawTabAlg(int x1, int y1, int w, int h, int r, PixelType color, VectorRenderer:
 /** ROUNDED SQUARES **/
 template<typename PixelType>
 void VectorRendererAA<PixelType>::
-drawRoundedSquareAlg(int x1, int y1, int r, int w, int h, PixelType color, VectorRenderer::FillMode fill_m) {
+drawBorderRoundedSquareAlg(int x1, int y1, int r, int w, int h, PixelType color, VectorRenderer::FillMode fill_m, uint8 alpha_t, uint8 alpha_r, uint8 alpha_b, uint8 alpha_l) {
 	int x, y;
 	const int pitch = Base::_activeSurface->pitch / Base::_activeSurface->format.bytesPerPixel;
 	int px, py;
@@ -1879,65 +2136,89 @@ drawRoundedSquareAlg(int x1, int y1, int r, int w, int h, PixelType color, Vecto
 	frac_t T = 0, oldT;
 	uint8 a1, a2;
 
-	// TODO: Split this up into border, bevel and interior functions
+	PixelType *ptr_tl = (PixelType *)Base::_activeSurface->getBasePtr(x1 + r, y1 + r);
+	PixelType *ptr_tr = (PixelType *)Base::_activeSurface->getBasePtr(x1 + w - r, y1 + r);
+	PixelType *ptr_bl = (PixelType *)Base::_activeSurface->getBasePtr(x1 + r, y1 + h - r);
+	PixelType *ptr_br = (PixelType *)Base::_activeSurface->getBasePtr(x1 + w - r, y1 + h - r);
+	PixelType *ptr_fill = (PixelType *)Base::_activeSurface->getBasePtr(x1, y1);
 
-	if (Base::_strokeWidth) {
-		PixelType *ptr_tl = (PixelType *)Base::_activeSurface->getBasePtr(x1 + r, y1 + r);
-		PixelType *ptr_tr = (PixelType *)Base::_activeSurface->getBasePtr(x1 + w - r, y1 + r);
-		PixelType *ptr_bl = (PixelType *)Base::_activeSurface->getBasePtr(x1 + r, y1 + h - r);
-		PixelType *ptr_br = (PixelType *)Base::_activeSurface->getBasePtr(x1 + w - r, y1 + h - r);
-		PixelType *ptr_fill = (PixelType *)Base::_activeSurface->getBasePtr(x1, y1);
-
-		int sw = 0, sp = 0;
-		int short_h = h - 2 * r;
-		int hp = h * pitch;
-
-		int strokeWidth = Base::_strokeWidth;
-		// If we're going to fill the inside, draw a slightly thicker border
-		// so we can blend the inside on top of it.
-		if (fill_m != Base::kFillDisabled) strokeWidth++;
-
-		// TODO: A gradient effect on the bevel
-		PixelType color1, color2;
-		color1 = Base::_bevel ? Base::_bevelColor : color;
-		color2 = color;
+	int sw = 0, sp = 0;
+	int short_h = h - 2 * r;
+	int hp = h * pitch;
 
+	int strokeWidth = Base::_strokeWidth;
 
-		while (sw++ < strokeWidth) {
-			colorFill<PixelType>(ptr_fill + sp + r, ptr_fill + w + 1 + sp - r, color1);
-			colorFill<PixelType>(ptr_fill + hp - sp + r, ptr_fill + w + hp + 1 - sp - r, color2);
-			sp += pitch;
+	while (sw++ < strokeWidth) {
+		this->blendFill(ptr_fill + hp - sp + r, ptr_fill + w + hp + 1 - sp - r, color, alpha_b); // bottom
+		this->blendFill(ptr_fill + sp + r, ptr_fill + w + 1 + sp - r, color, alpha_t); // top
 
-			x = r - (sw - 1);
-			y = 0;
-			T = 0;
-			px = pitch * x;
-			py = 0;
+		sp += pitch;
 
-			while (x > y++) {
-				WU_ALGORITHM();
+		x = r - (sw - 1);
+		y = 0;
+		T = 0;
+		px = pitch * x;
+		py = 0;
 
-				// sw == 1: outside, sw = _strokeWidth: inside
-				// We always draw the outer edge AAed, but the inner edge
-				// only when the inside isn't filled
-				if (sw != strokeWidth || fill_m != Base::kFillDisabled)
-					a2 = 255;
+		int alphaStep_tr = ((alpha_t - alpha_r)/(x+1));
+		int alphaStep_br = ((alpha_r - alpha_b)/(x+1));
+		int alphaStep_bl = ((alpha_b - alpha_l)/(x+1));
+		int alphaStep_tl = ((alpha_l - alpha_t)/(x+1));
 
-				// inner arc
-				WU_DRAWCIRCLE_BCOLOR(ptr_tr, ptr_tl, ptr_bl, ptr_br, (x - 1), y, (px - pitch), py, a2);
+		while (x > y++) {
+			WU_ALGORITHM();
 
-				if (sw == 1) // outer arc
-					WU_DRAWCIRCLE_BCOLOR(ptr_tr, ptr_tl, ptr_bl, ptr_br, x, y, px, py, a1);
+			// sw == 1: outside, sw = _strokeWidth: inside
+			// We always draw the outer edge AAed, but the inner edge
+			// only when the inside isn't filled
+			if (sw != strokeWidth || fill_m != Base::kFillDisabled)
+				a2 = 255;
+			
+				// inner arc		
+				WU_DRAWCIRCLE_BCOLOR_TR_CW(ptr_tr, (x - 1), y, (px - pitch), py, (uint8)((uint32)(((alpha_t - (alphaStep_tr * y)) << 8) * a2) >> 16)); 
+				WU_DRAWCIRCLE_BCOLOR_BR_CW(ptr_br, (x - 1), y, (px - pitch), py, (uint8)((uint32)(((alpha_r - (alphaStep_br * y)) << 8) * a2) >> 16));
+				WU_DRAWCIRCLE_BCOLOR_BL_CW(ptr_bl, (x - 1), y, (px - pitch), py, (uint8)((uint32)(((alpha_b - (alphaStep_bl * y)) << 8) * a2) >> 16));
+				WU_DRAWCIRCLE_BCOLOR_TL_CW(ptr_tl, (x - 1), y, (px - pitch), py, (uint8)((uint32)(((alpha_l - (alphaStep_tl * y)) << 8) * a2) >> 16));
+				
+				WU_DRAWCIRCLE_BCOLOR_TR_CCW(ptr_tr, (x - 1), y, (px - pitch), py, (uint8)((uint32)(((alpha_r + (alphaStep_tr * y)) << 8) * a2) >> 16)); 
+				WU_DRAWCIRCLE_BCOLOR_BR_CCW(ptr_br, (x - 1), y, (px - pitch), py, (uint8)((uint32)(((alpha_b + (alphaStep_br * y)) << 8) * a2) >> 16));
+				WU_DRAWCIRCLE_BCOLOR_BL_CCW(ptr_bl, (x - 1), y, (px - pitch), py, (uint8)((uint32)(((alpha_l + (alphaStep_bl * y)) << 8) * a2) >> 16));
+				WU_DRAWCIRCLE_BCOLOR_TL_CCW(ptr_tl, (x - 1), y, (px - pitch), py, (uint8)((uint32)(((alpha_t + (alphaStep_tl * y)) << 8) * a2) >> 16));
+
+				// outer arc
+				if (sw == 1) {
+					WU_DRAWCIRCLE_BCOLOR_TR_CW(ptr_tr, x, y, px, py, (uint8)((uint32)(((alpha_t - (alphaStep_tr * y)) << 8) * a1) >> 16));
+					WU_DRAWCIRCLE_BCOLOR_BR_CW(ptr_br, x, y, px, py, (uint8)((uint32)(((alpha_r - (alphaStep_br * y)) << 8) * a1) >> 16));
+					WU_DRAWCIRCLE_BCOLOR_BL_CW(ptr_bl, x, y, px, py, (uint8)((uint32)(((alpha_b - (alphaStep_bl * y)) << 8) * a1) >> 16));
+					WU_DRAWCIRCLE_BCOLOR_TL_CW(ptr_tl, x, y, px, py, (uint8)((uint32)(((alpha_l - (alphaStep_tl * y)) << 8) * a1) >> 16));
+					
+					WU_DRAWCIRCLE_BCOLOR_TR_CCW(ptr_tr, x, y, px, py, (uint8)((uint32)(((alpha_r + (alphaStep_tr * y)) << 8) * a1) >> 16)); 
+					WU_DRAWCIRCLE_BCOLOR_BR_CCW(ptr_br, x, y, px, py, (uint8)((uint32)(((alpha_b + (alphaStep_br * y)) << 8) * a1) >> 16));
+					WU_DRAWCIRCLE_BCOLOR_BL_CCW(ptr_bl, x, y, px, py, (uint8)((uint32)(((alpha_l + (alphaStep_bl * y)) << 8) * a1) >> 16));
+					WU_DRAWCIRCLE_BCOLOR_TL_CCW(ptr_tl, x, y, px, py, (uint8)((uint32)(((alpha_t + (alphaStep_tl * y)) << 8) * a1) >> 16));
+				}
 			}
-		}
 
 		ptr_fill += pitch * r;
-		while (short_h-- >= 0) {
-			colorFill<PixelType>(ptr_fill, ptr_fill + Base::_strokeWidth, color1);
-			colorFill<PixelType>(ptr_fill + w - Base::_strokeWidth + 1, ptr_fill + w + 1, color2);
+
+		while (short_h-- >= -2) {
+			this->blendFill(ptr_fill, ptr_fill + Base::_strokeWidth, color, alpha_l); // left
+			this->blendFill(ptr_fill + w - Base::_strokeWidth + 1, ptr_fill + w + 1, color, alpha_r); // right
 			ptr_fill += pitch;
 		}
 	}
+}
+
+template<typename PixelType>
+void VectorRendererAA<PixelType>::
+drawInteriorRoundedSquareAlg(int x1, int y1, int r, int w, int h, PixelType color, VectorRenderer::FillMode fill_m) {
+	int x, y;
+	const int pitch = Base::_activeSurface->pitch / Base::_activeSurface->format.bytesPerPixel;
+	int px, py;
+	
+	uint32 rsq = r*r;
+	frac_t T = 0, oldT;
+	uint8 a1, a2;
 
 	r -= Base::_strokeWidth;
 	x1 += Base::_strokeWidth;
@@ -1946,93 +2227,116 @@ drawRoundedSquareAlg(int x1, int y1, int r, int w, int h, PixelType color, Vecto
 	h -= 2*Base::_strokeWidth;
 	rsq = r*r;
 
-	if (w <= 0 || h <= 0)
-		return; // Only border is visible
+	PixelType *ptr_tl = (PixelType *)Base::_activeSurface->getBasePtr(x1 + r, y1 + r);
+	PixelType *ptr_tr = (PixelType *)Base::_activeSurface->getBasePtr(x1 + w - r, y1 + r);
+	PixelType *ptr_bl = (PixelType *)Base::_activeSurface->getBasePtr(x1 + r, y1 + h - r);
+	PixelType *ptr_br = (PixelType *)Base::_activeSurface->getBasePtr(x1 + w - r, y1 + h - r);
+	PixelType *ptr_fill = (PixelType *)Base::_activeSurface->getBasePtr(x1, y1);
 
-	if (fill_m != Base::kFillDisabled) {
-		if (fill_m == Base::kFillBackground)
-			color = Base::_bgColor;
+	int short_h = h - 2 * r;
+	x = r;
+	y = 0;
+	T = 0;
+	px = pitch * x;
+	py = 0;
 
-		PixelType *ptr_tl = (PixelType *)Base::_activeSurface->getBasePtr(x1 + r, y1 + r);
-		PixelType *ptr_tr = (PixelType *)Base::_activeSurface->getBasePtr(x1 + w - r, y1 + r);
-		PixelType *ptr_bl = (PixelType *)Base::_activeSurface->getBasePtr(x1 + r, y1 + h - r);
-		PixelType *ptr_br = (PixelType *)Base::_activeSurface->getBasePtr(x1 + w - r, y1 + h - r);
-		PixelType *ptr_fill = (PixelType *)Base::_activeSurface->getBasePtr(x1, y1);
+	if (fill_m == Base::kFillGradient) {
 
-		int short_h = h - 2 * r;
-		x = r;
-		y = 0;
-		T = 0;
-		px = pitch * x;
-		py = 0;
+		Base::precalcGradient(h);
 
-		if (fill_m == Base::kFillGradient) {
+		PixelType color1, color2, color3, color4;
+		while (x > y++) {
+			WU_ALGORITHM();
 
-			Base::precalcGradient(h);
+			color1 = Base::calcGradient(r - x, h);
+			color2 = Base::calcGradient(r - y, h);
+			color3 = Base::calcGradient(h - r + x, h);
+			color4 = Base::calcGradient(h - r + y, h);
 
-			PixelType color1, color2, color3, color4;
-			while (x > y++) {
-				WU_ALGORITHM();
+			Base::gradientFill(ptr_tl - x - py + 1, w - 2 * r + 2 * x - 1, x1 + r - x - y + 1, r - y);
 
-				color1 = Base::calcGradient(r - x, h);
-				color2 = Base::calcGradient(r - y, h);
-				color3 = Base::calcGradient(h - r + x, h);
-				color4 = Base::calcGradient(h - r + y, h);
+			// Only fill each horizontal line once (or we destroy
+			// the gradient effect at the edges)
+			if (T < oldT || y == 1)
+				Base::gradientFill(ptr_tl - y - px + 1, w - 2 * r + 2 * y - 1, x1 + r - y - x + 1, r - x);
 
-				Base::gradientFill(ptr_tl - x - py + 1, w - 2 * r + 2 * x - 1, x1 + r - x - y + 1, r - y);
+			Base::gradientFill(ptr_bl - x + py + 1, w - 2 * r + 2 * x - 1, x1 + r - x - y + 1, h - r + y);
 
-				// Only fill each horizontal line once (or we destroy
-				// the gradient effect at the edges)
-				if (T < oldT || y == 1)
-					Base::gradientFill(ptr_tl - y - px + 1, w - 2 * r + 2 * y - 1, x1 + r - y - x + 1, r - x);
+			// Only fill each horizontal line once (or we destroy
+			// the gradient effect at the edges)
+			if (T < oldT || y == 1)
+				Base::gradientFill(ptr_bl - y + px + 1, w - 2 * r + 2 * y - 1, x1 + r - y - x + 1, h - r + x);
 
-				Base::gradientFill(ptr_bl - x + py + 1, w - 2 * r + 2 * x - 1, x1 + r - x - y + 1, h - r + y);
+			// This shape is used for dialog backgrounds.
+			// If we're drawing on top of an empty overlay background,
+			// and the overlay supports alpha, we have to do AA by
+			// setting the dest alpha channel, instead of blending with
+			// dest color channels.
+			if (!g_system->hasFeature(OSystem::kFeatureOverlaySupportsAlpha))
+				WU_DRAWCIRCLE_XCOLOR(ptr_tr, ptr_tl, ptr_bl, ptr_br, x, y, px, py, a1, blendPixelPtr);
+			else
+				WU_DRAWCIRCLE_XCOLOR(ptr_tr, ptr_tl, ptr_bl, ptr_br, x, y, px, py, a1, blendPixelDestAlphaPtr);
+		}
 
-				// Only fill each horizontal line once (or we destroy
-				// the gradient effect at the edges)
-				if (T < oldT || y == 1)
-					Base::gradientFill(ptr_bl - y + px + 1, w - 2 * r + 2 * y - 1, x1 + r - y - x + 1, h - r + x);
-
-				// This shape is used for dialog backgrounds.
-				// If we're drawing on top of an empty overlay background,
-				// and the overlay supports alpha, we have to do AA by
-				// setting the dest alpha channel, instead of blending with
-				// dest color channels.
-				if (!g_system->hasFeature(OSystem::kFeatureOverlaySupportsAlpha))
-					WU_DRAWCIRCLE_XCOLOR(ptr_tr, ptr_tl, ptr_bl, ptr_br, x, y, px, py, a1, blendPixelPtr);
-				else
-					WU_DRAWCIRCLE_XCOLOR(ptr_tr, ptr_tl, ptr_bl, ptr_br, x, y, px, py, a1, blendPixelDestAlphaPtr);
-			}
+		ptr_fill += pitch * r;
+		while (short_h-- >= 0) {
+			Base::gradientFill(ptr_fill, w + 1, x1, r++);
+			ptr_fill += pitch;
+		}
 
-			ptr_fill += pitch * r;
-			while (short_h-- >= 0) {
-				Base::gradientFill(ptr_fill, w + 1, x1, r++);
-				ptr_fill += pitch;
-			}
+	} else {
 
-		} else {
+		while (x > 1 + y++) {
+			WU_ALGORITHM();
 
-			while (x > 1 + y++) {
-				WU_ALGORITHM();
+			colorFill<PixelType>(ptr_tl - x - py + 1, ptr_tr + x - py, color);
+			if (T < oldT || y == 1)
+				colorFill<PixelType>(ptr_tl - y - px + 1, ptr_tr + y - px, color);
 
-				colorFill<PixelType>(ptr_tl - x - py + 1, ptr_tr + x - py, color);
-				if (T < oldT || y == 1)
-					colorFill<PixelType>(ptr_tl - y - px + 1, ptr_tr + y - px, color);
+			colorFill<PixelType>(ptr_bl - x + py + 1, ptr_br + x + py, color);
+			if (T < oldT || y == 1)
+				colorFill<PixelType>(ptr_bl - y + px + 1, ptr_br + y + px, color);
 
-				colorFill<PixelType>(ptr_bl - x + py + 1, ptr_br + x + py, color);
-				if (T < oldT || y == 1)
-					colorFill<PixelType>(ptr_bl - y + px + 1, ptr_br + y + px, color);
+			WU_DRAWCIRCLE(ptr_tr, ptr_tl, ptr_bl, ptr_br, x, y, px, py, a1);
+		}
 
-				WU_DRAWCIRCLE(ptr_tr, ptr_tl, ptr_bl, ptr_br, x, y, px, py, a1);
-			}
+		ptr_fill += pitch * r;
+		while (short_h-- >= 0) {
+			colorFill<PixelType>(ptr_fill, ptr_fill + w + 1, color);
+			ptr_fill += pitch;
+		}
+	}
+}
 
-			ptr_fill += pitch * r;
-			while (short_h-- >= 0) {
-				colorFill<PixelType>(ptr_fill, ptr_fill + w + 1, color);
-				ptr_fill += pitch;
-			}
+template<typename PixelType>
+void VectorRendererAA<PixelType>::
+drawRoundedSquareAlg(int x1, int y1, int r, int w, int h, PixelType color, VectorRenderer::FillMode fill_m) {
+	const uint8 borderAlpha_t = 0;
+	const uint8 borderAlpha_r = 127;
+	const uint8 borderAlpha_b = 255; 
+	const uint8 borderAlpha_l = 63;
+
+	const uint8 bevelAlpha_t = 255;
+	const uint8 bevelAlpha_r = 31;
+	const uint8 bevelAlpha_b = 0;
+	const uint8 bevelAlpha_l = 127;
+	
+	if (Base::_strokeWidth) {
+		if (r != 0 && Base::_bevel > 0) {
+			drawBorderRoundedSquareAlg(x1, y1, r, w, h, color, fill_m, borderAlpha_t, borderAlpha_r, borderAlpha_b, borderAlpha_l);
+			drawBorderRoundedSquareAlg(x1, y1, r, w, h, Base::_bevelColor, fill_m, bevelAlpha_t, bevelAlpha_r, bevelAlpha_b, bevelAlpha_l);
+		} else {
+			drawBorderRoundedSquareAlg(x1, y1, r, w, h, color, fill_m, 255, 255, 255, 255);
 		}
 	}
+	
+	// If only border is visible
+	if ((!(w <= 0 || h <= 0)) && (fill_m != Base::kFillDisabled)) {
+		if (fill_m == Base::kFillBackground)
+			drawInteriorRoundedSquareAlg(x1, y1, r, w, h, Base::_bgColor, fill_m);
+		else
+			drawInteriorRoundedSquareAlg(x1, y1, r, w, h, color, fill_m);
+	}
 }
 
 /** CIRCLES **/
diff --git a/graphics/VectorRendererSpec.h b/graphics/VectorRendererSpec.h
index 4ed80cb55f..c035ca0e19 100644
--- a/graphics/VectorRendererSpec.h
+++ b/graphics/VectorRendererSpec.h
@@ -61,7 +61,7 @@ public:
 	}
 	void drawString(const Graphics::Font *font, const Common::String &text,
 					const Common::Rect &area, Graphics::TextAlign alignH,
-					GUI::ThemeEngine::TextAlignVertical alignV, int deltax, bool elipsis);
+					GUI::ThemeEngine::TextAlignVertical alignV, int deltax, bool elipsis, const Common::Rect &textDrawableArea = Common::Rect(0, 0, 0, 0));
 
 	void setFgColor(uint8 r, uint8 g, uint8 b) { _fgColor = _format.RGBToColor(r, g, b); }
 	void setBgColor(uint8 r, uint8 g, uint8 b) { _bgColor = _format.RGBToColor(r, g, b); }
@@ -158,6 +158,12 @@ protected:
 	virtual void drawRoundedSquareAlg(int x1, int y1, int r, int w, int h,
 	    PixelType color, FillMode fill_m);
 
+	virtual void drawBorderRoundedSquareAlg(int x1, int y1, int r, int w, int h,
+	    PixelType color, FillMode fill_m, uint8 alpha_t, uint8 alpha_r, uint8 alpha_b, uint8 alpha_l);
+
+	virtual void drawInteriorRoundedSquareAlg(int x1, int y1, int r, int w, int h,
+	    PixelType color, FillMode fill_m);
+
 	virtual void drawSquareAlg(int x, int y, int w, int h,
 	    PixelType color, FillMode fill_m);
 
@@ -174,6 +180,8 @@ protected:
 	    PixelType color, VectorRenderer::FillMode fill_m,
 	    int baseLeft = 0, int baseRight = 0);
 
+	virtual void drawTabShadow(int x, int y, int w, int h, int r);
+
 	virtual void drawBevelTabAlg(int x, int y, int w, int h,
 	    int bevel, PixelType topColor, PixelType bottomColor,
 	    int baseLeft = 0, int baseRight = 0);
@@ -186,10 +194,10 @@ protected:
 	 * There functions may be overloaded in inheriting classes to improve performance
 	 * in the slowest platforms where pixel alpha blending just doesn't cut it.
 	 *
-	 * @param blur Intensity/size of the shadow.
+	 * @param offset Intensity/size of the shadow.
 	 */
-	virtual void drawSquareShadow(int x, int y, int w, int h, int blur);
-	virtual void drawRoundedSquareShadow(int x, int y, int r, int w, int h, int blur);
+	virtual void drawSquareShadow(int x, int y, int w, int h, int offset);
+	virtual void drawRoundedSquareShadow(int x, int y, int r, int w, int h, int offset);
 
 	/**
 	 * Calculates the color gradient on a given point.
@@ -292,10 +300,12 @@ protected:
 	 */
 	virtual void drawRoundedSquareAlg(int x1, int y1, int r, int w, int h, PixelType color, VectorRenderer::FillMode fill_m);
 
-	virtual void drawRoundedSquareShadow(int x, int y, int r, int w, int h, int blur) {
-		Base::drawRoundedSquareShadow(x, y, r, w, h, blur);
-//		VectorRenderer::applyConvolutionMatrix(VectorRenderer::kConvolutionHardBlur,
-//            Common::Rect(x, y, x + w + blur * 2, y + h + blur * 2));
+	virtual void drawBorderRoundedSquareAlg(int x1, int y1, int r, int w, int h, PixelType color, VectorRenderer::FillMode fill_m, uint8 alpha_t, uint8 alpha_l, uint8 alpha_r, uint8 alpha_b);
+	
+	virtual void drawInteriorRoundedSquareAlg(int x1, int y1, int r, int w, int h, PixelType color, VectorRenderer::FillMode fill_m);
+
+	virtual void drawRoundedSquareShadow(int x, int y, int r, int w, int h, int offset) {
+		Base::drawRoundedSquareShadow(x, y, r, w, h, offset);
 	}
 
 	virtual void drawTabAlg(int x, int y, int w, int h, int r,
diff --git a/graphics/cursorman.cpp b/graphics/cursorman.cpp
index c818101645..133ee5fd71 100644
--- a/graphics/cursorman.cpp
+++ b/graphics/cursorman.cpp
@@ -48,6 +48,9 @@ bool CursorManager::isVisible() {
 bool CursorManager::showMouse(bool visible) {
 	if (_cursorStack.empty())
 		return false;
+	if (_locked) {
+		return false;
+	}
 
 	_cursorStack.top()->_visible = visible;
 
@@ -225,6 +228,10 @@ void CursorManager::replaceCursorPalette(const byte *colors, uint start, uint nu
 	}
 }
 
+void CursorManager::lock(bool locked) {
+	_locked = locked;
+}
+
 CursorManager::Cursor::Cursor(const void *data, uint w, uint h, int hotspotX, int hotspotY, uint32 keycolor, bool dontScale, const Graphics::PixelFormat *format) {
 #ifdef USE_RGB_COLOR
 	if (!format)
@@ -246,6 +253,7 @@ CursorManager::Cursor::Cursor(const void *data, uint w, uint h, int hotspotX, in
 	_hotspotX = hotspotX;
 	_hotspotY = hotspotY;
 	_dontScale = dontScale;
+	_visible = false;
 }
 
 CursorManager::Cursor::~Cursor() {
diff --git a/graphics/cursorman.h b/graphics/cursorman.h
index 66e8d1ba56..b4d8ad94ce 100644
--- a/graphics/cursorman.h
+++ b/graphics/cursorman.h
@@ -160,12 +160,15 @@ public:
 	 */
 	void replaceCursorPalette(const byte *colors, uint start, uint num);
 
+	void lock(bool locked);
 private:
 	friend class Common::Singleton<SingletonBaseType>;
 	// Even though this is basically the default constructor we implement it
 	// ourselves, so it is private and thus there is no way to create this class
 	// except from the Singleton code.
-	CursorManager() {}
+	CursorManager() {
+		_locked = false;
+	}
 	~CursorManager();
 
 	struct Cursor {
@@ -198,6 +201,7 @@ private:
 	};
 	Common::Stack<Cursor *> _cursorStack;
 	Common::Stack<Palette *> _cursorPaletteStack;
+	bool _locked;
 };
 
 } // End of namespace Graphics
diff --git a/graphics/decoders/bmp.cpp b/graphics/decoders/bmp.cpp
index bcfd0abbda..2eabbb7631 100644
--- a/graphics/decoders/bmp.cpp
+++ b/graphics/decoders/bmp.cpp
@@ -130,14 +130,14 @@ bool BitmapDecoder::loadStream(Common::SeekableReadStream &stream) {
 	const int extraDataLength = (srcPitch % 4) ? 4 - (srcPitch % 4) : 0;
 
 	if (bitsPerPixel == 8) {
-		byte *dst = (byte *)_surface->pixels;
+		byte *dst = (byte *)_surface->getPixels();
 
 		for (int32 i = 0; i < height; i++) {
 			stream.read(dst + (height - i - 1) * width, width);
 			stream.skip(extraDataLength);
 		}
 	} else if (bitsPerPixel == 24) {
-		byte *dst = (byte *)_surface->pixels + (height - 1) * _surface->pitch;
+		byte *dst = (byte *)_surface->getBasePtr(0, height - 1);
 
 		for (int32 i = 0; i < height; i++) {
 			for (uint32 j = 0; j < width; j++) {
@@ -154,7 +154,7 @@ bool BitmapDecoder::loadStream(Common::SeekableReadStream &stream) {
 			dst -= _surface->pitch * 2;
 		}
 	} else { // 32 bpp
-		byte *dst = (byte *)_surface->pixels + (height - 1) * _surface->pitch;
+		byte *dst = (byte *)_surface->getBasePtr(0, height - 1);
 
 		for (int32 i = 0; i < height; i++) {
 			for (uint32 j = 0; j < width; j++) {
diff --git a/graphics/decoders/iff.cpp b/graphics/decoders/iff.cpp
index 50c7b4f7de..fe27258d28 100644
--- a/graphics/decoders/iff.cpp
+++ b/graphics/decoders/iff.cpp
@@ -30,6 +30,10 @@ namespace Graphics {
 IFFDecoder::IFFDecoder() {
 	_surface = 0;
 	_palette = 0;
+	
+	// these 2 properties are not reset by destroy(), so the default is set here.
+	_numRelevantPlanes = 8;
+	_pixelPacking = false;
 
 	destroy();
 }
@@ -54,8 +58,6 @@ void IFFDecoder::destroy() {
 	_paletteRanges.clear();
 	_type = TYPE_UNKNOWN;
 	_paletteColorCount = 0;
-	_numRelevantPlanes = 8;
-	_pixelPacking = false;
 }
 
 bool IFFDecoder::loadStream(Common::SeekableReadStream &stream) {
@@ -170,7 +172,7 @@ void IFFDecoder::loadBitmap(Common::SeekableReadStream &stream) {
 	if (_type == TYPE_ILBM) {
 		uint32 scanlinePitch = ((_header.width + 15) >> 4) << 1;
 		byte *scanlines = new byte[scanlinePitch * _header.numPlanes];
-		byte *data = (byte *)_surface->pixels;
+		byte *data = (byte *)_surface->getPixels();
 
 		for (uint16 i = 0; i < _header.height; ++i) {
 			byte *scanline = scanlines;
@@ -194,7 +196,7 @@ void IFFDecoder::loadBitmap(Common::SeekableReadStream &stream) {
 
 		delete[] scanlines;
 	} else if (_type == TYPE_PBM) {
-		byte *data = (byte *)_surface->pixels;
+		byte *data = (byte *)_surface->getPixels();
 		uint32 outSize = _header.width * _header.height;
 
 		if (_header.compression) {
diff --git a/graphics/decoders/iff.h b/graphics/decoders/iff.h
index beac62e519..37cb4b38c1 100644
--- a/graphics/decoders/iff.h
+++ b/graphics/decoders/iff.h
@@ -85,6 +85,8 @@ public:
 	/**
 	* The number of planes to decode, also determines the pixel packing if _packPixels is true.
 	* 8 == decode all planes, map 1 pixel in 1 byte. (default, no packing even if _packPixels is true)
+	*
+	* NOTE: this property must be reset manually, and is not reset by a call to destroy().
 	*/
 	void setNumRelevantPlanes(const uint8 numRelevantPlanes) { _numRelevantPlanes = numRelevantPlanes; }
 
@@ -94,6 +96,8 @@ public:
 	* 2 == decode first 2 planes, pack 4 pixels in 1 byte. This makes _surface->w 1/4th of _header.width
 	* 4 == decode first 4 planes, pack 2 pixels in 1 byte. This makes _surface->w half of _header.width
 	* Packed bitmaps won't have a proper surface format since there is no way to tell it to use 1, 2 or 4 bits per pixel
+	*
+	* NOTE: this property must be reset manually, and is not reset by a call to destroy().
 	*/
 	void setPixelPacking(const bool pixelPacking) { _pixelPacking = pixelPacking; }
 private:
diff --git a/graphics/decoders/image_decoder.h b/graphics/decoders/image_decoder.h
index 49e31c6e3a..a39a9a1493 100644
--- a/graphics/decoders/image_decoder.h
+++ b/graphics/decoders/image_decoder.h
@@ -44,6 +44,9 @@ public:
 
 	/**
 	 * Load an image from the specified stream
+	 * 
+	 * loadStream() should implicitly call destroy() to free the memory
+	 * of the last loadStream() call.
 	 *
 	 * @param stream the input stream
 	 * @return whether loading the file succeeded
@@ -54,6 +57,9 @@ public:
 
 	/**
 	 * Destroy this decoder's surface and palette
+	 *
+	 * This should be called by a loadStream() implementation as well
+	 * as the destructor.
 	 */
 	virtual void destroy() = 0;
 
diff --git a/graphics/decoders/jpeg.cpp b/graphics/decoders/jpeg.cpp
index 75fdcd6e5a..c858884095 100644
--- a/graphics/decoders/jpeg.cpp
+++ b/graphics/decoders/jpeg.cpp
@@ -20,8 +20,11 @@
  *
  */
 
+// libjpeg uses forbidden symbols in its header. Thus, we need to allow them
+// here.
+#define FORBIDDEN_SYMBOL_ALLOW_ALL
+
 #include "graphics/pixelformat.h"
-#include "graphics/yuv_to_rgb.h"
 #include "graphics/decoders/jpeg.h"
 
 #include "common/debug.h"
@@ -29,35 +32,19 @@
 #include "common/stream.h"
 #include "common/textconsole.h"
 
-namespace Graphics {
-
-// Order used to traverse the quantization tables
-static const uint8 _zigZagOrder[64] = {
-	0,   1,  8, 16,  9,  2,  3, 10,
-	17, 24, 32, 25, 18, 11,  4,  5,
-	12, 19, 26, 33, 40, 48, 41, 34,
-	27, 20, 13,  6,  7, 14, 21, 28,
-	35, 42, 49, 56, 57, 50, 43, 36,
-	29, 22, 15, 23, 30, 37, 44, 51,
-	58, 59, 52, 45, 38, 31, 39, 46,
-	53, 60, 61, 54, 47, 55, 62, 63
-};
-
-JPEGDecoder::JPEGDecoder() : ImageDecoder(),
-	_stream(NULL), _w(0), _h(0), _numComp(0), _components(NULL), _numScanComp(0),
-	_scanComp(NULL), _currentComp(NULL), _rgbSurface(0) {
+#ifdef USE_JPEG
+// The original release of libjpeg v6b did not contain any extern "C" in case
+// its header files are included in a C++ environment. To avoid any linking
+// issues we need to add it on our own.
+extern "C" {
+#include <jpeglib.h>
+#include <jerror.h>
+}
+#endif
 
-	// Initialize the quantization tables
-	for (int i = 0; i < JPEG_MAX_QUANT_TABLES; i++)
-		_quant[i] = NULL;
+namespace Graphics {
 
-	// Initialize the Huffman tables
-	for (int i = 0; i < 2 * JPEG_MAX_HUFF_TABLES; i++) {
-		_huff[i].count = 0;
-		_huff[i].values = NULL;
-		_huff[i].sizes = NULL;
-		_huff[i].codes = NULL;
-	}
+JPEGDecoder::JPEGDecoder() : ImageDecoder(), _surface(), _colorSpace(kColorSpaceRGBA) {
 }
 
 JPEGDecoder::~JPEGDecoder() {
@@ -65,723 +52,215 @@ JPEGDecoder::~JPEGDecoder() {
 }
 
 const Surface *JPEGDecoder::getSurface() const {
-	// Make sure we have loaded data
-	if (!isLoaded())
-		return 0;
-
-	if (_rgbSurface)
-		return _rgbSurface;
-
-	// Create an RGBA8888 surface
-	_rgbSurface = new Graphics::Surface();
-	_rgbSurface->create(_w, _h, Graphics::PixelFormat(4, 8, 8, 8, 0, 24, 16, 8, 0));
-
-	// Get our component surfaces
-	const Graphics::Surface *yComponent = getComponent(1);
-	const Graphics::Surface *uComponent = getComponent(2);
-	const Graphics::Surface *vComponent = getComponent(3);
-
-	YUVToRGBMan.convert444(_rgbSurface, Graphics::YUVToRGBManager::kScaleFull, (byte *)yComponent->pixels, (byte *)uComponent->pixels, (byte *)vComponent->pixels, yComponent->w, yComponent->h, yComponent->pitch, uComponent->pitch);
-
-	return _rgbSurface;
+	return &_surface;
 }
 
 void JPEGDecoder::destroy() {
-	// Reset member variables
-	_stream = NULL;
-	_w = _h = 0;
-	_restartInterval = 0;
-
-	// Free the components
-	for (int c = 0; c < _numComp; c++)
-		_components[c].surface.free();
-	delete[] _components; _components = NULL;
-	_numComp = 0;
-
-	// Free the scan components
-	delete[] _scanComp; _scanComp = NULL;
-	_numScanComp = 0;
-	_currentComp = NULL;
-
-	// Free the quantization tables
-	for (int i = 0; i < JPEG_MAX_QUANT_TABLES; i++) {
-		delete[] _quant[i];
-		_quant[i] = NULL;
-	}
-
-	// Free the Huffman tables
-	for (int i = 0; i < 2 * JPEG_MAX_HUFF_TABLES; i++) {
-		_huff[i].count = 0;
-		delete[] _huff[i].values; _huff[i].values = NULL;
-		delete[] _huff[i].sizes; _huff[i].sizes = NULL;
-		delete[] _huff[i].codes; _huff[i].codes = NULL;
-	}
-
-	if (_rgbSurface) {
-		_rgbSurface->free();
-		delete _rgbSurface;
-	}
+	_surface.free();
 }
 
-bool JPEGDecoder::loadStream(Common::SeekableReadStream &stream) {
-	// Reset member variables and tables from previous reads
-	destroy();
-
-	// Save the input stream
-	_stream = &stream;
-
-	bool ok = true;
-	bool done = false;
-	while (!_stream->eos() && ok && !done) {
-		// Read the marker
-
-		// WORKAROUND: While each and every JPEG file should end with
-		// an EOI (end of image) tag, in reality this may not be the
-		// case. For instance, at least one image in the Masterpiece
-		// edition of Myst doesn't, yet other programs are able to read
-		// the image without complaining.
-		//
-		// Apparently, the customary workaround is to insert a fake
-		// EOI tag.
-
-		uint16 marker = _stream->readByte();
-		bool fakeEOI = false;
-
-		if (_stream->eos()) {
-			fakeEOI = true;
-			marker = 0xFF;
-		}
-
-		if (marker != 0xFF) {
-			error("JPEG: Invalid marker[0]: 0x%02X", marker);
-			ok = false;
-			break;
-		}
-
-		while (marker == 0xFF && !_stream->eos())
-			marker = _stream->readByte();
-
-		if (_stream->eos()) {
-			fakeEOI = true;
-			marker = 0xD9;
-		}
-
-		if (fakeEOI)
-			warning("JPEG: Inserted fake EOI");
+#ifdef USE_JPEG
+namespace {
 
-		// Process the marker data
-		switch (marker) {
-		case 0xC0: // Start Of Frame
-			ok = readSOF0();
-			break;
-		case 0xC4: // Define Huffman Tables
-			ok = readDHT();
-			break;
-		case 0xD8: // Start Of Image
-			break;
-		case 0xD9: // End Of Image
-			done = true;
-			break;
-		case 0xDA: // Start Of Scan
-			ok = readSOS();
-			break;
-		case 0xDB: // Define Quantization Tables
-			ok = readDQT();
-			break;
-		case 0xE0: // JFIF/JFXX segment
-			ok = readJFIF();
-			break;
-		case 0xDD: // Define Restart Interval
-			ok = readDRI();
-			break;
-		case 0xFE: // Comment
-			_stream->seek(_stream->readUint16BE() - 2, SEEK_CUR);
-			break;
-		default: { // Unknown marker
-			uint16 size = _stream->readUint16BE();
+#define JPEG_BUFFER_SIZE 4096
 
-			if ((marker & 0xE0) != 0xE0)
-				warning("JPEG: Unknown marker %02X, skipping %d bytes", marker, size - 2);
-
-			_stream->seek(size - 2, SEEK_CUR);
-		}
-		}
-	}
+struct StreamSource : public jpeg_source_mgr {
+	Common::SeekableReadStream *stream;
+	bool startOfFile;
+	JOCTET buffer[JPEG_BUFFER_SIZE];
+};
 
-	_stream = 0;
-	return ok;
+void initSource(j_decompress_ptr cinfo) {
+	StreamSource *source = (StreamSource *)cinfo->src;
+	source->startOfFile = true;
 }
 
-bool JPEGDecoder::readJFIF() {
-	uint16 length = _stream->readUint16BE();
-	uint32 tag = _stream->readUint32BE();
+boolean fillInputBuffer(j_decompress_ptr cinfo) {
+	StreamSource *source = (StreamSource *)cinfo->src;
 
-	if (tag != MKTAG('J', 'F', 'I', 'F')) {
-		warning("JPEGDecoder::readJFIF() tag mismatch");
-		return false;
-	}
-
-	if (_stream->readByte() != 0)  { // NULL
-		warning("JPEGDecoder::readJFIF() NULL mismatch");
-		return false;
+	uint32 bufferSize = source->stream->read((byte *)source->buffer, sizeof(source->buffer));
+	if (bufferSize == 0) {
+		if (source->startOfFile) {
+			// An empty file is a fatal error
+			ERREXIT(cinfo, JERR_INPUT_EMPTY);
+		} else {
+			// Otherwise we insert an EOF marker
+			WARNMS(cinfo, JWRN_JPEG_EOF);
+			source->buffer[0] = (JOCTET)0xFF;
+			source->buffer[1] = (JOCTET)JPEG_EOI;
+			bufferSize = 2;
+		}
 	}
 
-	byte majorVersion = _stream->readByte();
-	byte minorVersion = _stream->readByte();
-	if (majorVersion != 1 || minorVersion > 2)
-		warning("JPEGDecoder::readJFIF(): v%d.%02d JPEGs may not be handled correctly", majorVersion, minorVersion);
-
-	/* byte densityUnits = */_stream->readByte();
-	/* uint16 xDensity = */_stream->readUint16BE();
-	/* uint16 yDensity = */_stream->readUint16BE();
-	byte thumbW = _stream->readByte();
-	byte thumbH = _stream->readByte();
-
-	_stream->seek(thumbW * thumbH * 3, SEEK_CUR); // Ignore thumbnail
-	if (length != (thumbW * thumbH * 3) + 16) {
-		warning("JPEGDecoder::readJFIF() length mismatch");
-		return false;
-	}
+	source->next_input_byte = source->buffer;
+	source->bytes_in_buffer = bufferSize;
+	source->startOfFile = false;
 
-	return true;
+	return TRUE;
 }
 
-// Marker 0xC0 (Start Of Frame, Baseline DCT)
-bool JPEGDecoder::readSOF0() {
-	debug(5, "JPEG: readSOF0");
-	uint16 size = _stream->readUint16BE();
-
-	// Read the sample precision
-	uint8 precision = _stream->readByte();
-	if (precision != 8) {
-		warning("JPEG: Just 8 bit precision supported at the moment");
-		return false;
-	}
+void skipInputData(j_decompress_ptr cinfo, long numBytes) {
+	StreamSource *source = (StreamSource *)cinfo->src;
 
-	// Image size
-	_h = _stream->readUint16BE();
-	_w = _stream->readUint16BE();
-
-	// Number of components
-	_numComp = _stream->readByte();
-	if (size != 8 + 3 * _numComp) {
-		warning("JPEG: Invalid number of components");
-		return false;
-	}
+	if (numBytes > 0) {
+		if (numBytes > (long)source->bytes_in_buffer) {
+			// In case we need to skip more bytes than there are in the buffer
+			// we will skip the remaining data and fill the buffer again
+			numBytes -= (long)source->bytes_in_buffer;
 
-	// Allocate the new components
-	delete[] _components;
-	_components = new Component[_numComp];
-
-	// Read the components details
-	for (int c = 0; c < _numComp; c++) {
-		_components[c].id = _stream->readByte();
-		_components[c].factorH = _stream->readByte();
-		_components[c].factorV = _components[c].factorH & 0xF;
-		_components[c].factorH >>= 4;
-		_components[c].quantTableSelector = _stream->readByte();
-	}
-
-	return true;
-}
+			// Skip the remaining bytes
+			source->stream->skip(numBytes);
 
-// Marker 0xC4 (Define Huffman Tables)
-bool JPEGDecoder::readDHT() {
-	debug(5, "JPEG: readDHT");
-	uint16 size = _stream->readUint16BE() - 2;
-	uint32 pos = _stream->pos();
-
-	while ((uint32)_stream->pos() < (size + pos)) {
-		// Read the table type and id
-		uint8 tableId = _stream->readByte();
-		uint8 tableType = tableId >> 4; // type 0: DC, 1: AC
-		tableId &= 0xF;
-		uint8 tableNum = (tableId << 1) + tableType;
-
-		// Free the Huffman table
-		delete[] _huff[tableNum].values; _huff[tableNum].values = NULL;
-		delete[] _huff[tableNum].sizes; _huff[tableNum].sizes = NULL;
-		delete[] _huff[tableNum].codes; _huff[tableNum].codes = NULL;
-
-		// Read the number of values for each length
-		uint8 numValues[16];
-		_huff[tableNum].count = 0;
-		for (int len = 0; len < 16; len++) {
-			numValues[len] = _stream->readByte();
-			_huff[tableNum].count += numValues[len];
-		}
-
-		// Allocate memory for the current table
-		_huff[tableNum].values = new uint8[_huff[tableNum].count];
-		_huff[tableNum].sizes = new uint8[_huff[tableNum].count];
-		_huff[tableNum].codes = new uint16[_huff[tableNum].count];
-
-		// Read the table contents
-		int cur = 0;
-		for (int len = 0; len < 16; len++) {
-			for (int i = 0; i < numValues[len]; i++) {
-				_huff[tableNum].values[cur] = _stream->readByte();
-				_huff[tableNum].sizes[cur] = len + 1;
-				cur++;
-			}
+			// Fill up the buffer again
+			(*source->fill_input_buffer)(cinfo);
+		} else {
+			source->next_input_byte += (size_t)numBytes;
+			source->bytes_in_buffer -= (size_t)numBytes;
 		}
 
-		// Fill the table of Huffman codes
-		cur = 0;
-		uint16 curCode = 0;
-		uint8 curCodeSize = _huff[tableNum].sizes[0];
-		while (cur < _huff[tableNum].count) {
-			// Increase the code size to fit the request
-			while (_huff[tableNum].sizes[cur] != curCodeSize) {
-				curCode <<= 1;
-				curCodeSize++;
-			}
-
-			// Assign the current code
-			_huff[tableNum].codes[cur] = curCode;
-			curCode++;
-			cur++;
-		}
 	}
-
-	return true;
 }
 
-// Marker 0xDA (Start Of Scan)
-bool JPEGDecoder::readSOS() {
-	debug(5, "JPEG: readSOS");
-	uint16 size = _stream->readUint16BE();
-
-	// Number of scan components
-	_numScanComp = _stream->readByte();
-	if (size != 6 + 2 * _numScanComp) {
-		warning("JPEG: Invalid number of components");
-		return false;
-	}
-
-	// Allocate the new scan components
-	delete[] _scanComp;
-	_scanComp = new Component *[_numScanComp];
-
-	// Reset the maximum sampling factors
-	_maxFactorV = 0;
-	_maxFactorH = 0;
-
-	// Component-specification parameters
-	for (int c = 0; c < _numScanComp; c++) {
-		// Read the desired component id
-		uint8 id = _stream->readByte();
-
-		// Search the component with the specified id
-		bool found = false;
-		for (int i = 0; !found && i < _numComp; i++) {
-			if (_components[i].id == id) {
-				// We found the desired component
-				found = true;
-
-				// Assign the found component to the c'th scan component
-				_scanComp[c] = &_components[i];
-			}
-		}
-
-		if (!found) {
-			warning("JPEG: Invalid component");
-			return false;
-		}
-
-		// Read the entropy table selectors
-		_scanComp[c]->DCentropyTableSelector = _stream->readByte();
-		_scanComp[c]->ACentropyTableSelector = _scanComp[c]->DCentropyTableSelector & 0xF;
-		_scanComp[c]->DCentropyTableSelector >>= 4;
-
-		// Calculate the maximum sampling factors
-		if (_scanComp[c]->factorV > _maxFactorV)
-			_maxFactorV = _scanComp[c]->factorV;
-
-		if (_scanComp[c]->factorH > _maxFactorH)
-			_maxFactorH = _scanComp[c]->factorH;
-
-		// Initialize the DC predictor
-		_scanComp[c]->DCpredictor = 0;
-	}
-
-	// Start of spectral selection
-	if (_stream->readByte() != 0) {
-		warning("JPEG: Progressive scanning not supported");
-		return false;
-	}
-
-	// End of spectral selection
-	if (_stream->readByte() != 63) {
-		warning("JPEG: Progressive scanning not supported");
-		return false;
-	}
-
-	// Successive approximation parameters
-	if (_stream->readByte() != 0) {
-		warning("JPEG: Progressive scanning not supported");
-		return false;
-	}
-
-	// Entropy coded sequence starts, initialize Huffman decoder
-	_bitsNumber = 0;
-
-	// Read all the scan MCUs
-	uint16 xMCU = _w / (_maxFactorH * 8);
-	uint16 yMCU = _h / (_maxFactorV * 8);
-
-	// Check for non- multiple-of-8 dimensions
-	if (_w % (_maxFactorH * 8) != 0)
-		xMCU++;
-	if (_h % (_maxFactorV * 8) != 0)
-		yMCU++;
-
-	// Initialize the scan surfaces
-	for (uint16 c = 0; c < _numScanComp; c++) {
-		_scanComp[c]->surface.create(xMCU * _maxFactorH * 8, yMCU * _maxFactorV * 8, PixelFormat::createFormatCLUT8());
-	}
-
-	bool ok = true;
-	uint16 interval = _restartInterval;
-
-	for (int y = 0; ok && (y < yMCU); y++) {
-		for (int x = 0; ok && (x < xMCU); x++) {
-			ok = readMCU(x, y);
-
-			// If we have a restart interval, we'll need to reset a couple
-			// variables
-			if (_restartInterval != 0) {
-				interval--;
-
-				if (interval == 0) {
-					interval = _restartInterval;
-					_bitsNumber = 0;
-
-					for (byte i = 0; i < _numScanComp; i++)
-						_scanComp[i]->DCpredictor = 0;
-				}
-			}
-		}
-	}
-
-	// Trim Component surfaces back to image height and width
-	// Note: Code using jpeg must use surface.pitch correctly...
-	for (uint16 c = 0; c < _numScanComp; c++) {
-		_scanComp[c]->surface.w = _w;
-		_scanComp[c]->surface.h = _h;
-	}
-
-	return ok;
+void termSource(j_decompress_ptr cinfo) {
 }
 
-// Marker 0xDB (Define Quantization Tables)
-bool JPEGDecoder::readDQT() {
-	debug(5, "JPEG: readDQT");
-	uint16 size = _stream->readUint16BE() - 2;
-	uint32 pos = _stream->pos();
-
-	while ((uint32)_stream->pos() < (pos + size)) {
-		// Read the table precision and id
-		uint8 tableId = _stream->readByte();
-		bool highPrecision = (tableId & 0xF0) != 0;
-
-		// Validate the table id
-		tableId &= 0xF;
-		if (tableId >= JPEG_MAX_QUANT_TABLES) {
-			warning("JPEG: Invalid quantization table");
-			return false;
-		}
-
-		// Create the new table if necessary
-		if (!_quant[tableId])
-			_quant[tableId] = new uint16[64];
+void jpeg_scummvm_src(j_decompress_ptr cinfo, Common::SeekableReadStream *stream) {
+	StreamSource *source;
 
-		// Read the table (stored in Zig-Zag order)
-		for (int i = 0; i < 64; i++)
-			_quant[tableId][i] = highPrecision ? _stream->readUint16BE() : _stream->readByte();
+	// Initialize the source in case it has not been done yet.
+	if (cinfo->src == NULL) {
+		cinfo->src = (jpeg_source_mgr *)(*cinfo->mem->alloc_small)((j_common_ptr)cinfo, JPOOL_PERMANENT, sizeof(StreamSource));
 	}
 
-	return true;
-}
+	source = (StreamSource *)cinfo->src;
+	source->init_source       = &initSource;
+	source->fill_input_buffer = &fillInputBuffer;
+	source->skip_input_data   = &skipInputData;
+	source->resync_to_restart = &jpeg_resync_to_restart;
+	source->term_source       = &termSource;
+	source->bytes_in_buffer   = 0;
+	source->next_input_byte   = NULL;
 
-// Marker 0xDD (Define Restart Interval)
-bool JPEGDecoder::readDRI() {
-	debug(5, "JPEG: readDRI");
-	uint16 size = _stream->readUint16BE() - 2;
-
-	if (size != 2) {
-		warning("JPEG: Invalid DRI size %d", size);
-		return false;
-	}
-
-	_restartInterval = _stream->readUint16BE();
-	debug(5, "Restart interval: %d", _restartInterval);
-	return true;
+	source->stream = stream;
 }
 
-bool JPEGDecoder::readMCU(uint16 xMCU, uint16 yMCU) {
-	bool ok = true;
-	for (int c = 0; ok && (c < _numComp); c++) {
-		// Set the current component
-		_currentComp = _scanComp[c];
-
-		// Read the data units of the current component
-		for (int y = 0; ok && (y < _scanComp[c]->factorV); y++)
-			for (int x = 0; ok && (x < _scanComp[c]->factorH); x++)
-				ok = readDataUnit(xMCU * _scanComp[c]->factorH + x, yMCU * _scanComp[c]->factorV + y);
-	}
-
-	return ok;
+void errorExit(j_common_ptr cinfo) {
+	char buffer[JMSG_LENGTH_MAX];
+	(*cinfo->err->format_message)(cinfo, buffer);
+	// This function is not allowed to return to the caller, thus we simply
+	// error out with our error handling here.
+	error("%s", buffer);
 }
 
-// triple-butterfly-add (and possible rounding)
-#define xadd3(xa, xb, xc, xd, h) \
-	p = xa + xb; \
-	n = xa - xb; \
-	xa = p + xc + h; \
-	xb = n + xd + h; \
-	xc = p - xc + h; \
-	xd = n - xd + h;
-
-// butterfly-mul
-#define xmul(xa, xb, k1, k2, sh) \
-	n = k1 * (xa + xb); \
-	p = xa; \
-	xa = (n + (k2 - k1) * xb) >> sh; \
-	xb = (n - (k2 + k1) * p) >> sh;
-
-// IDCT based on public domain code from http://halicery.com/jpeg/idct.html
-void JPEGDecoder::idct1D8x8(int32 src[8], int32 dest[64], int32 ps, int32 half) {
-	int p, n;
-
-	src[0] <<= 9;
-	src[1] <<= 7;
-	src[3] *= 181;
-	src[4] <<= 9;
-	src[5] *= 181;
-	src[7] <<= 7;
-
-	// Even part
-	xmul(src[6], src[2], 277, 669, 0)
-	xadd3(src[0], src[4], src[6], src[2], half)
-
-	// Odd part
-	xadd3(src[1], src[7], src[3], src[5], 0)
-	xmul(src[5], src[3], 251, 50, 6)
-	xmul(src[1], src[7], 213, 142, 6)
-
-	dest[0 * 8] = (src[0] + src[1]) >> ps;
-	dest[1 * 8] = (src[4] + src[5]) >> ps;
-	dest[2 * 8] = (src[2] + src[3]) >> ps;
-	dest[3 * 8] = (src[6] + src[7]) >> ps;
-	dest[4 * 8] = (src[6] - src[7]) >> ps;
-	dest[5 * 8] = (src[2] - src[3]) >> ps;
-	dest[6 * 8] = (src[4] - src[5]) >> ps;
-	dest[7 * 8] = (src[0] - src[1]) >> ps;
+void outputMessage(j_common_ptr cinfo) {
+	char buffer[JMSG_LENGTH_MAX];
+	(*cinfo->err->format_message)(cinfo, buffer);
+	// Is using debug here a good idea? Or do we want to ignore all libjpeg
+	// messages?
+	debug(3, "libjpeg: %s", buffer);
 }
 
-void JPEGDecoder::idct2D8x8(int32 block[64]) {
-	int32 tmp[64];
-
-	// Apply 1D IDCT to rows
-	for (int i = 0; i < 8; i++)
-		idct1D8x8(&block[i * 8], &tmp[i], 9, 1 << 8);
-
-	// Apply 1D IDCT to columns
-	for (int i = 0; i < 8; i++)
-		idct1D8x8(&tmp[i * 8], &block[i], 12, 1 << 11);
- }
-
-bool JPEGDecoder::readDataUnit(uint16 x, uint16 y) {
-	// Prepare an empty data array
-	int16 readData[64];
-	for (int i = 1; i < 64; i++)
-		readData[i] = 0;
-
-	// Read the DC component
-	readData[0] = _currentComp->DCpredictor + readDC();
-	_currentComp->DCpredictor = readData[0];
-
-	// Read the AC components (stored in Zig-Zag)
-	readAC(readData);
-
-	// Calculate the DCT coefficients from the input sequence
-	int32 block[64];
-	for (uint8 i = 0; i < 64; i++) {
-		// Dequantize
-		int32 val = readData[i];
-		int16 quant = _quant[_currentComp->quantTableSelector][i];
-		val *= quant;
-
-		// Store the normalized coefficients, undoing the Zig-Zag
-		block[_zigZagOrder[i]] = val;
-	}
+} // End of anonymous namespace
+#endif
 
-	// Apply the IDCT
-	idct2D8x8(block);
+bool JPEGDecoder::loadStream(Common::SeekableReadStream &stream) {
+#ifdef USE_JPEG
+	// Reset member variables from previous decodings
+	destroy();
 
-	// Level shift to make the values unsigned
-	for (int i = 0; i < 64; i++) {
-		block[i] = block[i] + 128;
+	jpeg_decompress_struct cinfo;
+	jpeg_error_mgr jerr;
 
-		if (block[i] < 0)
-			block[i] = 0;
+	// Initialize error handling callbacks
+	cinfo.err = jpeg_std_error(&jerr);
+	cinfo.err->error_exit = &errorExit;
+	cinfo.err->output_message = &outputMessage;
 
-		if (block[i] > 255)
-			block[i] = 255;
-	}
+	// Initialize the decompression structure
+	jpeg_create_decompress(&cinfo);
 
-	// Paint the component surface
-	uint8 scalingV = _maxFactorV / _currentComp->factorV;
-	uint8 scalingH = _maxFactorH / _currentComp->factorH;
+	// Initialize our buffer handling
+	jpeg_scummvm_src(&cinfo, &stream);
 
-	// Convert coordinates from MCU blocks to pixels
-	x <<= 3;
-	y <<= 3;
+	// Read the file header
+	jpeg_read_header(&cinfo, TRUE);
 
-	for (uint8 j = 0; j < 8; j++) {
-		for (uint16 sV = 0; sV < scalingV; sV++) {
-			// Get the beginning of the block line
-			byte *ptr = (byte *)_currentComp->surface.getBasePtr(x * scalingH, (y + j) * scalingV + sV);
+	// We can request YUV output because Groovie requires it
+	switch (_colorSpace) {
+	case kColorSpaceRGBA:
+		cinfo.out_color_space = JCS_RGB;
+		break;
 
-			for (uint8 i = 0; i < 8; i++) {
-				for (uint16 sH = 0; sH < scalingH; sH++) {
-					*ptr = (byte)(block[j * 8 + i]);
-					ptr++;
-				}
-			}
-		}
+	case kColorSpaceYUV:
+		cinfo.out_color_space = JCS_YCbCr;
+		break;
 	}
 
-	return true;
-}
-
-int16 JPEGDecoder::readDC() {
-	// DC is type 0
-	uint8 tableNum = _currentComp->DCentropyTableSelector << 1;
-
-	// Get the number of bits to read
-	uint8 numBits = readHuff(tableNum);
-
-	// Read the requested bits
-	return readSignedBits(numBits);
-}
-
-void JPEGDecoder::readAC(int16 *out) {
-	// AC is type 1
-	uint8 tableNum = (_currentComp->ACentropyTableSelector << 1) + 1;
-
-	// Start reading AC element 1
-	uint8 cur = 1;
-	while (cur < 64) {
-		uint8 s = readHuff(tableNum);
-		uint8 r = s >> 4;
-		s &= 0xF;
-
-		if (s == 0) {
-			if (r == 15) {
-				// Skip 16 values
-				cur += 16;
-			} else {
-				// EOB: end of block
-				cur = 64;
-			}
-		} else {
-			// Skip r values
-			cur += r;
-
-			// Read the next value
-			out[cur] = readSignedBits(s);
-			cur++;
-		}
+	// Actually start decompressing the image
+	jpeg_start_decompress(&cinfo);
+
+	// Allocate buffers for the output data
+	switch (_colorSpace) {
+	case kColorSpaceRGBA:
+		// We use RGBA8888 in this scenario
+		_surface.create(cinfo.output_width, cinfo.output_height, Graphics::PixelFormat(4, 8, 8, 8, 0, 24, 16, 8, 0));
+		break;
+
+	case kColorSpaceYUV:
+		// We use YUV with 3 bytes per pixel otherwise.
+		// This is pretty ugly since our PixelFormat cannot express YUV...
+		_surface.create(cinfo.output_width, cinfo.output_height, Graphics::PixelFormat(3, 0, 0, 0, 0, 0, 0, 0, 0));
+		break;
 	}
-}
-
-int16 JPEGDecoder::readSignedBits(uint8 numBits) {
-	uint16 ret = 0;
-	if (numBits > 16)
-		error("requested %d bits", numBits); //XXX
-
-	// MSB=0 for negatives, 1 for positives
-	for (int i = 0; i < numBits; i++)
-		ret = (ret << 1) + readBit();
-
-	// Extend sign bits (PAG109)
-	if (!(ret >> (numBits - 1))) {
-		uint16 tmp = ((uint16)-1 << numBits) + 1;
-		ret = ret + tmp;
-	}
-	return ret;
-}
-
-// TODO: optimize?
-uint8 JPEGDecoder::readHuff(uint8 table) {
-	bool foundCode = false;
-	uint8 val = 0;
-
-	uint8 cur = 0;
-	uint8 codeSize = 1;
-	uint16 code = readBit();
-	while (!foundCode) {
-		// Prepare a code of the current size
-		while (codeSize < _huff[table].sizes[cur]) {
-			code = (code << 1) + readBit();
-			codeSize++;
-		}
 
-		// Compare the codes of the current size
-		while (!foundCode && (codeSize == _huff[table].sizes[cur])) {
-			if (code == _huff[table].codes[cur]) {
-				// Found the code
-				val = _huff[table].values[cur];
-				foundCode = true;
-			} else {
-				// Continue reading
-				cur++;
+	// Allocate buffer for one scanline
+	assert(cinfo.output_components == 3);
+	JDIMENSION pitch = cinfo.output_width * cinfo.output_components;
+	assert(_surface.pitch >= pitch);
+	JSAMPARRAY buffer = (*cinfo.mem->alloc_sarray)((j_common_ptr)&cinfo, JPOOL_IMAGE, pitch, 1);
+
+	// Go through the image data scanline by scanline
+	while (cinfo.output_scanline < cinfo.output_height) {
+		byte *dst = (byte *)_surface.getBasePtr(0, cinfo.output_scanline);
+
+		jpeg_read_scanlines(&cinfo, buffer, 1);
+
+		const byte *src = buffer[0];
+		switch (_colorSpace) {
+		case kColorSpaceRGBA: {
+			for (int remaining = cinfo.output_width; remaining > 0; --remaining) {
+				byte r = *src++;
+				byte g = *src++;
+				byte b = *src++;
+				// We need to insert a alpha value of 255 (opaque) here.
+#ifdef SCUMM_BIG_ENDIAN
+				*dst++ = r;
+				*dst++ = g;
+				*dst++ = b;
+				*dst++ = 0xFF;
+#else
+				*dst++ = 0xFF;
+				*dst++ = b;
+				*dst++ = g;
+				*dst++ = r;
+#endif
 			}
-		}
-	}
+			} break;
 
-	return val;
-}
-
-uint8 JPEGDecoder::readBit() {
-	// Read a whole byte if necessary
-	if (_bitsNumber == 0) {
-		_bitsData = _stream->readByte();
-		_bitsNumber = 8;
-
-		// Detect markers
-		if (_bitsData == 0xFF) {
-			uint8 byte2 = _stream->readByte();
-
-			// A stuffed 0 validates the previous byte
-			if (byte2 != 0) {
-				if (byte2 == 0xDC) {
-					// DNL marker: Define Number of Lines
-					// TODO: terminate scan
-					warning("DNL marker detected: terminate scan");
-				} else if (byte2 >= 0xD0 && byte2 <= 0xD7) {
-					debug(7, "RST%d marker detected", byte2 & 7);
-					_bitsData = _stream->readByte();
-				} else {
-					warning("Error: marker 0x%02X read in entropy data", byte2);
-				}
-			}
+		case kColorSpaceYUV:
+			memcpy(dst, src, pitch);
+			break;
 		}
 	}
-	_bitsNumber--;
-
-	return (_bitsData & (1 << _bitsNumber)) ? 1 : 0;
-}
 
-const Surface *JPEGDecoder::getComponent(uint c) const {
-	for (int i = 0; i < _numComp; i++)
-		if (_components[i].id == c) // We found the desired component
-			return &_components[i].surface;
+	// We are done with decompressing, thus free all the data
+	jpeg_finish_decompress(&cinfo);
+	jpeg_destroy_decompress(&cinfo);
 
-	error("JPEGDecoder::getComponent: No component %d present", c);
-	return NULL;
+	return true;
+#else
+	return false;
+#endif
 }
 
 } // End of Graphics namespace
diff --git a/graphics/decoders/jpeg.h b/graphics/decoders/jpeg.h
index d59b72adf4..8460bc2698 100644
--- a/graphics/decoders/jpeg.h
+++ b/graphics/decoders/jpeg.h
@@ -40,100 +40,54 @@ class SeekableReadStream;
 
 namespace Graphics {
 
-struct PixelFormat;
-
-#define JPEG_MAX_QUANT_TABLES 4
-#define JPEG_MAX_HUFF_TABLES 2
-
 class JPEGDecoder : public ImageDecoder {
 public:
 	JPEGDecoder();
 	~JPEGDecoder();
 
 	// ImageDecoder API
-	void destroy();
-	bool loadStream(Common::SeekableReadStream &str);
-	const Surface *getSurface() const;
-
-	bool isLoaded() const { return _numComp && _w && _h; }
-	uint16 getWidth() const { return _w; }
-	uint16 getHeight() const { return _h; }
-	const Surface *getComponent(uint c) const;
-
-private:
-	Common::SeekableReadStream *_stream;
-	uint16 _w, _h;
-	uint16 _restartInterval;
-
-	// mutable so that we can convert to RGB only during
-	// a getSurface() call while still upholding the
-	// const requirement in other ImageDecoders
-	mutable Graphics::Surface *_rgbSurface;
-
-	// Image components
-	uint8 _numComp;
-	struct Component {
-		// Global values
-		uint8 id;
-		uint8 factorH;
-		uint8 factorV;
-		uint8 quantTableSelector;
-
-		// Scan specific values
-		uint8 DCentropyTableSelector;
-		uint8 ACentropyTableSelector;
-		int16 DCpredictor;
-
-		// Result image for this component
-		Surface surface;
+	virtual void destroy();
+	virtual bool loadStream(Common::SeekableReadStream &str);
+	virtual const Surface *getSurface() const;
+
+	// Special API for JPEG
+	enum ColorSpace {
+		/**
+		 * Output 32bit RGBA data.
+		 *
+		 * This is the default output.
+		 */
+		kColorSpaceRGBA,
+
+		/**
+		 * Output (interleaved) YUV data.
+		 *
+		 * Be aware that some images cannot be output in YUV mode.
+		 * These are (non-standard) JPEG images which are in RGB colorspace.
+		 *
+		 * The resulting Surface will have a PixelFormat with 3 bytes per
+		 * pixel and the remaining entries are completely zeroed. This works
+		 * around the fact that PixelFormat can only describe RGB formats.
+		 *
+		 * You should only use this when you are really aware of what you are
+		 * doing!
+		 */
+		kColorSpaceYUV
 	};
 
-	Component *_components;
+	/**
+	 * Request the output color space. This can be used to obtain raw YUV
+	 * data from the JPEG file. But this might not work for all files!
+	 *
+	 * The decoder itself defaults to RGBA.
+	 *
+	 * @param outSpace The color space to output.
+	 */
+	void setOutputColorSpace(ColorSpace outSpace) { _colorSpace = outSpace; }
 
-	// Scan components
-	uint8 _numScanComp;
-	Component **_scanComp;
-	Component *_currentComp;
-
-	// Maximum sampling factors, used to calculate the interleaving of the MCU
-	uint8 _maxFactorV;
-	uint8 _maxFactorH;
-
-	// Quantization tables
-	uint16 *_quant[JPEG_MAX_QUANT_TABLES];
-
-	// Huffman tables
-	struct HuffmanTable {
-		uint8 count;
-		uint8 *values;
-		uint8 *sizes;
-		uint16 *codes;
-	} _huff[2 * JPEG_MAX_HUFF_TABLES];
-
-	// Marker read functions
-	bool readJFIF();
-	bool readSOF0();
-	bool readDHT();
-	bool readSOS();
-	bool readDQT();
-	bool readDRI();
-
-	// Helper functions
-	bool readMCU(uint16 xMCU, uint16 yMCU);
-	bool readDataUnit(uint16 x, uint16 y);
-	int16 readDC();
-	void readAC(int16 *out);
-	int16 readSignedBits(uint8 numBits);
-
-	// Huffman decoding
-	uint8 readHuff(uint8 table);
-	uint8 readBit();
-	uint8 _bitsData;
-	uint8 _bitsNumber;
-
-	// Inverse Discrete Cosine Transformation
-	static void idct1D8x8(int32 src[8], int32 dest[64], int32 ps, int32 half);
-	static void idct2D8x8(int32 block[64]);
+private:
+	Graphics::Surface _surface;
+	ColorSpace _colorSpace;
 };
 
 } // End of Graphics namespace
diff --git a/graphics/decoders/pcx.cpp b/graphics/decoders/pcx.cpp
index 1250398c73..eb9b4c997d 100644
--- a/graphics/decoders/pcx.cpp
+++ b/graphics/decoders/pcx.cpp
@@ -117,7 +117,7 @@ bool PCXDecoder::loadStream(Common::SeekableReadStream &stream) {
 	if (nPlanes == 3 && bitsPerPixel == 8) {	// 24bpp
 		Graphics::PixelFormat format = Graphics::PixelFormat(4, 8, 8, 8, 8, 24, 16, 8, 0);
 		_surface->create(width, height, format);
-		dst = (byte *)_surface->pixels;
+		dst = (byte *)_surface->getPixels();
 		_paletteColorCount = 0;
 
 		for (y = 0; y < height; y++) {
@@ -135,7 +135,7 @@ bool PCXDecoder::loadStream(Common::SeekableReadStream &stream) {
 		}
 	} else if (nPlanes == 1 && bitsPerPixel == 8) {	// 8bpp indexed
 		_surface->create(width, height, Graphics::PixelFormat::createFormatCLUT8());
-		dst = (byte *)_surface->pixels;
+		dst = (byte *)_surface->getPixels();
 		_paletteColorCount = 16;
 
 		for (y = 0; y < height; y++, dst += _surface->pitch) {
@@ -163,7 +163,7 @@ bool PCXDecoder::loadStream(Common::SeekableReadStream &stream) {
 		}
 	} else if ((nPlanes == 2 || nPlanes == 3 || nPlanes == 4) && bitsPerPixel == 1) {	// planar, 4, 8 or 16 colors
 		_surface->create(width, height, Graphics::PixelFormat::createFormatCLUT8());
-		dst = (byte *)_surface->pixels;
+		dst = (byte *)_surface->getPixels();
 		_paletteColorCount = 16;
 
 		for (y = 0; y < height; y++, dst += _surface->pitch) {
diff --git a/graphics/decoders/pict.cpp b/graphics/decoders/pict.cpp
index b1d408ebc3..f3e17b33e2 100644
--- a/graphics/decoders/pict.cpp
+++ b/graphics/decoders/pict.cpp
@@ -364,7 +364,7 @@ void PICTDecoder::unpackBitsRect(Common::SeekableReadStream &stream, bool withPa
 	case 1:
 		// Just copy to the image
 		_outputSurface->create(width, height, PixelFormat::createFormatCLUT8());
-		memcpy(_outputSurface->pixels, buffer, _outputSurface->w * _outputSurface->h);
+		memcpy(_outputSurface->getPixels(), buffer, _outputSurface->w * _outputSurface->h);
 		break;
 	case 2:
 		// We have a 16-bit surface
diff --git a/graphics/decoders/png.cpp b/graphics/decoders/png.cpp
index 11e26162eb..c4fee46fec 100644
--- a/graphics/decoders/png.cpp
+++ b/graphics/decoders/png.cpp
@@ -38,7 +38,7 @@
 
 namespace Graphics {
 
-PNGDecoder::PNGDecoder() : _outputSurface(0), _palette(0), _paletteColorCount(0) {
+PNGDecoder::PNGDecoder() : _outputSurface(0), _palette(0), _paletteColorCount(0), _stream(0) {
 }
 
 PNGDecoder::~PNGDecoder() {
@@ -163,16 +163,20 @@ bool PNGDecoder::loadStream(Common::SeekableReadStream &stream) {
 		_outputSurface->create(width, height, Graphics::PixelFormat::createFormatCLUT8());
 		png_set_packing(pngPtr);
 	} else {
-		_outputSurface->create(width, height, Graphics::PixelFormat(4, 8, 8, 8, 8, 24, 16, 8, 0));
-		if (!_outputSurface->pixels) {
+		bool isAlpha = (colorType & PNG_COLOR_MASK_ALPHA);
+		if (png_get_valid(pngPtr, infoPtr, PNG_INFO_tRNS)) {
+			isAlpha = true;
+			png_set_expand(pngPtr);
+		}
+		_outputSurface->create(width, height, Graphics::PixelFormat(4,
+		                       8, 8, 8, isAlpha ? 8 : 0, 24, 16, 8, 0));
+		if (!_outputSurface->getPixels()) {
 			error("Could not allocate memory for output image.");
 		}
 		if (bitDepth == 16)
 			png_set_strip_16(pngPtr);
 		if (bitDepth < 8)
 			png_set_expand(pngPtr);
-		if (png_get_valid(pngPtr, infoPtr, PNG_INFO_tRNS))
-			png_set_expand(pngPtr);
 		if (colorType == PNG_COLOR_TYPE_GRAY ||
 			colorType == PNG_COLOR_TYPE_GRAY_ALPHA)
 			png_set_gray_to_rgb(pngPtr);
diff --git a/graphics/decoders/tga.cpp b/graphics/decoders/tga.cpp
index c3b9d84055..bc27f18a49 100644
--- a/graphics/decoders/tga.cpp
+++ b/graphics/decoders/tga.cpp
@@ -50,6 +50,8 @@ void TGADecoder::destroy() {
 }
 
 bool TGADecoder::loadStream(Common::SeekableReadStream &tga) {
+	destroy();
+
 	byte imageType, pixelDepth;
 	bool success;
 	success = readHeader(tga, imageType, pixelDepth);
@@ -272,7 +274,7 @@ bool TGADecoder::readData(Common::SeekableReadStream &tga, byte imageType, byte
 	} else if (imageType == TYPE_BW) {
 		_surface.create(_surface.w, _surface.h, _format);
 
-		byte *data  = (byte *)_surface.pixels;
+		byte *data  = (byte *)_surface.getPixels();
 		uint32 count = _surface.w * _surface.h;
 
 		while (count-- > 0) {
@@ -318,7 +320,7 @@ bool TGADecoder::readDataRLE(Common::SeekableReadStream &tga, byte imageType, by
 	if (imageType == TYPE_RLE_TRUECOLOR || imageType == TYPE_RLE_BW || imageType == TYPE_RLE_CMAP) {
 		_surface.create(_surface.w, _surface.h, _format);
 		uint32 count = _surface.w * _surface.h;
-		byte *data = (byte *)_surface.pixels;
+		byte *data = (byte *)_surface.getPixels();
 
 		while (count > 0) {
 			uint32 header = tga.readByte();
diff --git a/graphics/decoders/tga.h b/graphics/decoders/tga.h
index d8ccf8f766..5b1e87d4c5 100644
--- a/graphics/decoders/tga.h
+++ b/graphics/decoders/tga.h
@@ -26,6 +26,7 @@
 /*
  * TGA decoder used in engines:
  *  - wintermute
+ *  - zvision
  */
 
 #ifndef GRAPHICS_DECODERS_TGA_H
diff --git a/graphics/font.cpp b/graphics/font.cpp
index 3b00cd8568..e9f139c478 100644
--- a/graphics/font.cpp
+++ b/graphics/font.cpp
@@ -26,93 +26,32 @@
 
 namespace Graphics {
 
-int Font::getKerningOffset(byte left, byte right) const {
+int Font::getKerningOffset(uint32 left, uint32 right) const {
 	return 0;
 }
 
-int Font::getStringWidth(const Common::String &str) const {
+namespace {
+
+template<class StringType>
+int getStringWidthImpl(const Font &font, const StringType &str) {
 	int space = 0;
-	uint last = 0;
+	typename StringType::unsigned_type last = 0;
 
 	for (uint i = 0; i < str.size(); ++i) {
-		const uint cur = str[i];
-		space += getCharWidth(cur) + getKerningOffset(last, cur);
+		const typename StringType::unsigned_type cur = str[i];
+		space += font.getCharWidth(cur) + font.getKerningOffset(last, cur);
 		last = cur;
 	}
 
 	return space;
 }
 
-void Font::drawString(Surface *dst, const Common::String &sOld, int x, int y, int w, uint32 color, TextAlign align, int deltax, bool useEllipsis) const {
+template<class StringType>
+void drawStringImpl(const Font &font, Surface *dst, const StringType &str, int x, int y, int w, uint32 color, TextAlign align, int deltax) {
 	assert(dst != 0);
-	const int leftX = x, rightX = x + w;
-	uint i;
-	Common::String s = sOld;
-	int width = getStringWidth(s);
-	Common::String str;
-
-	if (useEllipsis && width > w && s.hasSuffix("...")) {
-		// String is too wide. Check whether it ends in an ellipsis
-		// ("..."). If so, remove that and try again!
-		s.deleteLastChar();
-		s.deleteLastChar();
-		s.deleteLastChar();
-		width = getStringWidth(s);
-	}
-
-	if (useEllipsis && width > w) {
-		// String is too wide. So we shorten it "intelligently" by
-		// replacing parts of the string by an ellipsis. There are
-		// three possibilities for this: replace the start, the end, or
-		// the middle of the string. What is best really depends on the
-		// context; but unless we want to make this configurable,
-		// replacing the middle seems to be a good compromise.
-
-		const int ellipsisWidth = getStringWidth("...");
-
-		// SLOW algorithm to remove enough of the middle. But it is good enough
-		// for now.
-		const int halfWidth = (w - ellipsisWidth) / 2;
-		int w2 = 0;
-		uint last = 0;
-
-		for (i = 0; i < s.size(); ++i) {
-			const uint cur = s[i];
-			int charWidth = getCharWidth(cur) + getKerningOffset(last, cur);
-			if (w2 + charWidth > halfWidth)
-				break;
-			last = cur;
-			w2 += charWidth;
-			str += cur;
-		}
-
-		// At this point we know that the first 'i' chars are together 'w2'
-		// pixels wide. We took the first i-1, and add "..." to them.
-		str += "...";
-		last = '.';
 
-		// The original string is width wide. Of those we already skipped past
-		// w2 pixels, which means (width - w2) remain.
-		// The new str is (w2+ellipsisWidth) wide, so we can accommodate about
-		// (w - (w2+ellipsisWidth)) more pixels.
-		// Thus we skip ((width - w2) - (w - (w2+ellipsisWidth))) =
-		// (width + ellipsisWidth - w)
-		int skip = width + ellipsisWidth - w;
-		for (; i < s.size() && skip > 0; ++i) {
-			const uint cur = s[i];
-			skip -= getCharWidth(cur) + getKerningOffset(last, cur);
-			last = cur;
-		}
-
-		// Append the remaining chars, if any
-		for (; i < s.size(); ++i) {
-			str += s[i];
-		}
-
-		width = getStringWidth(str);
-	} else {
-		str = s;
-	}
+	const int leftX = x, rightX = x + w;
+	int width = font.getStringWidth(str);
 
 	if (align == kTextAlignCenter)
 		x = x + (w - width)/2;
@@ -120,29 +59,29 @@ void Font::drawString(Surface *dst, const Common::String &sOld, int x, int y, in
 		x = x + w - width;
 	x += deltax;
 
-	uint last = 0;
-	for (i = 0; i < str.size(); ++i) {
-		const uint cur = str[i];
-		x += getKerningOffset(last, cur);
+	typename StringType::unsigned_type last = 0;
+	for (typename StringType::const_iterator i = str.begin(), end = str.end(); i != end; ++i) {
+		const typename StringType::unsigned_type cur = *i;
+		x += font.getKerningOffset(last, cur);
 		last = cur;
-		w = getCharWidth(cur);
+		w = font.getCharWidth(cur);
 		if (x+w > rightX)
 			break;
-		if (x >= leftX)
-			drawChar(dst, str[i], x, y, color);
+		if (x+w >= leftX)
+			font.drawChar(dst, cur, x, y, color);
 		x += w;
 	}
 }
 
-
+template<class StringType>
 struct WordWrapper {
-	Common::Array<Common::String> &lines;
+	Common::Array<StringType> &lines;
 	int actualMaxLineWidth;
 
-	WordWrapper(Common::Array<Common::String> &l) : lines(l), actualMaxLineWidth(0) {
+	WordWrapper(Common::Array<StringType> &l) : lines(l), actualMaxLineWidth(0) {
 	}
 
-	void add(Common::String &line, int &w) {
+	void add(StringType &line, int &w) {
 		if (actualMaxLineWidth < w)
 			actualMaxLineWidth = w;
 
@@ -153,10 +92,11 @@ struct WordWrapper {
 	}
 };
 
-int Font::wordWrapText(const Common::String &str, int maxWidth, Common::Array<Common::String> &lines) const {
-	WordWrapper wrapper(lines);
-	Common::String line;
-	Common::String tmpStr;
+template<class StringType>
+int wordWrapTextImpl(const Font &font, const StringType &str, int maxWidth, Common::Array<StringType> &lines) {
+	WordWrapper<StringType> wrapper(lines);
+	StringType line;
+	StringType tmpStr;
 	int lineWidth = 0;
 	int tmpWidth = 0;
 
@@ -173,10 +113,10 @@ int Font::wordWrapText(const Common::String &str, int maxWidth, Common::Array<Co
 	// of a line. If we encounter such a word, we have to wrap it over multiple
 	// lines.
 
-	uint last = 0;
-	for (Common::String::const_iterator x = str.begin(); x != str.end(); ++x) {
-		const byte c = *x;
-		const int w = getCharWidth(c) + getKerningOffset(last, c);
+	typename StringType::unsigned_type last = 0;
+	for (typename StringType::const_iterator x = str.begin(); x != str.end(); ++x) {
+		const typename StringType::unsigned_type c = *x;
+		const int w = font.getCharWidth(c) + font.getKerningOffset(last, c);
 		last = c;
 		const bool wouldExceedWidth = (lineWidth + tmpWidth + w > maxWidth);
 
@@ -212,7 +152,7 @@ int Font::wordWrapText(const Common::String &str, int maxWidth, Common::Array<Co
 					tmpStr.deleteChar(0);
 					// This is not very fast, but it is the simplest way to
 					// assure we do not mess something up because of kerning.
-					tmpWidth = getStringWidth(tmpStr);
+					tmpWidth = font.getStringWidth(tmpStr);
 				}
 			} else {
 				wrapper.add(tmpStr, tmpWidth);
@@ -232,5 +172,98 @@ int Font::wordWrapText(const Common::String &str, int maxWidth, Common::Array<Co
 	return wrapper.actualMaxLineWidth;
 }
 
+} // End of anonymous namespace
+
+int Font::getStringWidth(const Common::String &str) const {
+	return getStringWidthImpl(*this, str);
+}
+
+int Font::getStringWidth(const Common::U32String &str) const {
+	return getStringWidthImpl(*this, str);
+}
+
+void Font::drawString(Surface *dst, const Common::String &sOld, int x, int y, int w, uint32 color, TextAlign align, int deltax, bool useEllipsis) const {
+	Common::String s = sOld;
+	int width = getStringWidth(s);
+	Common::String str;
+
+	if (useEllipsis && width > w && s.hasSuffix("...")) {
+		// String is too wide. Check whether it ends in an ellipsis
+		// ("..."). If so, remove that and try again!
+		s.deleteLastChar();
+		s.deleteLastChar();
+		s.deleteLastChar();
+		width = getStringWidth(s);
+	}
+
+	if (useEllipsis && width > w) {
+		// String is too wide. So we shorten it "intelligently" by
+		// replacing parts of the string by an ellipsis. There are
+		// three possibilities for this: replace the start, the end, or
+		// the middle of the string. What is best really depends on the
+		// context; but unless we want to make this configurable,
+		// replacing the middle seems to be a good compromise.
+
+		const int ellipsisWidth = getStringWidth("...");
+
+		// SLOW algorithm to remove enough of the middle. But it is good enough
+		// for now.
+		const int halfWidth = (w - ellipsisWidth) / 2;
+		int w2 = 0;
+		Common::String::unsigned_type last = 0;
+		uint i;
+
+		for (i = 0; i < s.size(); ++i) {
+			const Common::String::unsigned_type cur = s[i];
+			int charWidth = getCharWidth(cur) + getKerningOffset(last, cur);
+			if (w2 + charWidth > halfWidth)
+				break;
+			last = cur;
+			w2 += charWidth;
+			str += cur;
+		}
+
+		// At this point we know that the first 'i' chars are together 'w2'
+		// pixels wide. We took the first i-1, and add "..." to them.
+		str += "...";
+		last = '.';
+
+		// The original string is width wide. Of those we already skipped past
+		// w2 pixels, which means (width - w2) remain.
+		// The new str is (w2+ellipsisWidth) wide, so we can accommodate about
+		// (w - (w2+ellipsisWidth)) more pixels.
+		// Thus we skip ((width - w2) - (w - (w2+ellipsisWidth))) =
+		// (width + ellipsisWidth - w)
+		int skip = width + ellipsisWidth - w;
+		for (; i < s.size() && skip > 0; ++i) {
+			const Common::String::unsigned_type cur = s[i];
+			skip -= getCharWidth(cur) + getKerningOffset(last, cur);
+			last = cur;
+		}
+
+		// Append the remaining chars, if any
+		for (; i < s.size(); ++i) {
+			str += s[i];
+		}
+
+		width = getStringWidth(str);
+	} else {
+		str = s;
+	}
+
+	drawStringImpl(*this, dst, str, x, y, w, color, align, deltax);
+}
+
+void Font::drawString(Surface *dst, const Common::U32String &str, int x, int y, int w, uint32 color, TextAlign align) const {
+	drawStringImpl(*this, dst, str, x, y, w, color, align, 0);
+}
+
+int Font::wordWrapText(const Common::String &str, int maxWidth, Common::Array<Common::String> &lines) const {
+	return wordWrapTextImpl(*this, str, maxWidth, lines);
+}
+
+int Font::wordWrapText(const Common::U32String &str, int maxWidth, Common::Array<Common::U32String> &lines) const {
+	return wordWrapTextImpl(*this, str, maxWidth, lines);
+}
 
 } // End of namespace Graphics
diff --git a/graphics/font.h b/graphics/font.h
index 6819b42f52..77b7623a85 100644
--- a/graphics/font.h
+++ b/graphics/font.h
@@ -23,6 +23,7 @@
 #define GRAPHICS_FONT_H
 
 #include "common/str.h"
+#include "common/ustr.h"
 
 namespace Common {
 template<class T> class Array;
@@ -70,7 +71,7 @@ public:
 	 * @param chr The character to query the width of.
 	 * @return The character's width.
 	 */
-	virtual int getCharWidth(byte chr) const = 0;
+	virtual int getCharWidth(uint32 chr) const = 0;
 
 	/**
 	 * Query the kerning offset between two characters.
@@ -79,7 +80,7 @@ public:
 	 * @param right The right character. May be 0.
 	 * @return The horizontal displacement.
 	 */
-	virtual int getKerningOffset(byte left, byte right) const;
+	virtual int getKerningOffset(uint32 left, uint32 right) const;
 
 	/**
 	 * Draw a character at a specific point on a surface.
@@ -96,15 +97,17 @@ public:
 	 * @param y   The y coordinate where to draw the character.
 	 * @param color The color of the character.
 	 */
-	virtual void drawChar(Surface *dst, byte chr, int x, int y, uint32 color) const = 0;
+	virtual void drawChar(Surface *dst, uint32 chr, int x, int y, uint32 color) const = 0;
 
 	// TODO: Add doxygen comments to this
 	void drawString(Surface *dst, const Common::String &str, int x, int y, int w, uint32 color, TextAlign align = kTextAlignLeft, int deltax = 0, bool useEllipsis = true) const;
+	void drawString(Surface *dst, const Common::U32String &str, int x, int y, int w, uint32 color, TextAlign align = kTextAlignLeft) const;
 
 	/**
 	 * Compute and return the width the string str has when rendered using this font.
 	 */
 	int getStringWidth(const Common::String &str) const;
+	int getStringWidth(const Common::U32String &str) const;
 
 	/**
 	 * Take a text (which may contain newline characters) and word wrap it so that
@@ -120,6 +123,7 @@ public:
 	 * @return the maximal width of any of the lines added to lines
 	 */
 	int wordWrapText(const Common::String &str, int maxWidth, Common::Array<Common::String> &lines) const;
+	int wordWrapText(const Common::U32String &str, int maxWidth, Common::Array<Common::U32String> &lines) const;
 };
 
 } // End of namespace Graphics
diff --git a/graphics/fonts/bdf.cpp b/graphics/fonts/bdf.cpp
index 6d4befa37c..7b2290cc3a 100644
--- a/graphics/fonts/bdf.cpp
+++ b/graphics/fonts/bdf.cpp
@@ -51,7 +51,7 @@ int BdfFont::getMaxCharWidth() const {
 	return _data.maxAdvance;
 }
 
-int BdfFont::getCharWidth(byte chr) const {
+int BdfFont::getCharWidth(uint32 chr) const {
 	// In case all font have the same advance value, we use the maximum.
 	if (!_data.advances)
 		return _data.maxAdvance;
@@ -85,9 +85,9 @@ void drawCharIntern(byte *ptr, uint pitch, const byte *src, int h, int width, in
 	}
 }
 
-int BdfFont::mapToIndex(byte ch) const {
+int BdfFont::mapToIndex(uint32 ch) const {
 	// Check whether the character is included
-	if (_data.firstCharacter <= ch && ch <= _data.firstCharacter + _data.numCharacters) {
+	if (_data.firstCharacter <= (int)ch && (int)ch <= _data.firstCharacter + _data.numCharacters) {
 		if (_data.bitmaps[ch - _data.firstCharacter])
 			return ch - _data.firstCharacter;
 	}
@@ -95,14 +95,14 @@ int BdfFont::mapToIndex(byte ch) const {
 	return _data.defaultCharacter - _data.firstCharacter;
 }
 
-void BdfFont::drawChar(Surface *dst, byte chr, const int tx, const int ty, const uint32 color) const {
+void BdfFont::drawChar(Surface *dst, uint32 chr, const int tx, const int ty, const uint32 color) const {
 	assert(dst != 0);
 
 	// TODO: Where is the relation between the max advance being smaller or
 	// equal to 50 and the decision of the theme designer?
 	// asserting _data.maxAdvance <= 50: let the theme designer decide what looks best
 	assert(_data.maxAdvance <= 50);
-	assert(dst->format.bytesPerPixel == 1 || dst->format.bytesPerPixel == 2);
+	assert(dst->format.bytesPerPixel == 1 || dst->format.bytesPerPixel == 2 || dst->format.bytesPerPixel == 4);
 
 	const int idx = mapToIndex(chr);
 	if (idx < 0)
@@ -165,6 +165,8 @@ void BdfFont::drawChar(Surface *dst, byte chr, const int tx, const int ty, const
 		drawCharIntern<byte>(ptr, dst->pitch, src, height, originalWidth, xStart, xEnd, color);
 	else if (dst->format.bytesPerPixel == 2)
 		drawCharIntern<uint16>(ptr, dst->pitch, src, height, originalWidth, xStart, xEnd, color);
+	else if (dst->format.bytesPerPixel == 4)
+		drawCharIntern<uint32>(ptr, dst->pitch, src, height, originalWidth, xStart, xEnd, color);
 }
 
 namespace {
diff --git a/graphics/fonts/bdf.h b/graphics/fonts/bdf.h
index b0166a2095..842e54f851 100644
--- a/graphics/fonts/bdf.h
+++ b/graphics/fonts/bdf.h
@@ -61,14 +61,14 @@ public:
 	virtual int getFontHeight() const;
 	virtual int getMaxCharWidth() const;
 
-	virtual int getCharWidth(byte chr) const;
-	virtual void drawChar(Surface *dst, byte chr, int x, int y, uint32 color) const;
+	virtual int getCharWidth(uint32 chr) const;
+	virtual void drawChar(Surface *dst, uint32 chr, int x, int y, uint32 color) const;
 
 	static BdfFont *loadFont(Common::SeekableReadStream &stream);
 	static bool cacheFontData(const BdfFont &font, const Common::String &filename);
 	static BdfFont *loadFromCache(Common::SeekableReadStream &stream);
 private:
-	int mapToIndex(byte ch) const;
+	int mapToIndex(uint32 ch) const;
 
 	const BdfFontData _data;
 	const DisposeAfterUse::Flag _dispose;
diff --git a/graphics/fonts/ttf.cpp b/graphics/fonts/ttf.cpp
index 2b1dca1eae..93f119f028 100644
--- a/graphics/fonts/ttf.cpp
+++ b/graphics/fonts/ttf.cpp
@@ -107,11 +107,11 @@ public:
 
 	virtual int getMaxCharWidth() const;
 
-	virtual int getCharWidth(byte chr) const;
+	virtual int getCharWidth(uint32 chr) const;
 
-	virtual int getKerningOffset(byte left, byte right) const;
+	virtual int getKerningOffset(uint32 left, uint32 right) const;
 
-	virtual void drawChar(Surface *dst, byte chr, int x, int y, uint32 color) const;
+	virtual void drawChar(Surface *dst, uint32 chr, int x, int y, uint32 color) const;
 private:
 	bool _initialized;
 	FT_Face _face;
@@ -126,13 +126,14 @@ private:
 		Surface image;
 		int xOffset, yOffset;
 		int advance;
+		FT_UInt slot;
 	};
 
-	bool cacheGlyph(Glyph &glyph, FT_UInt &slot, uint chr);
-	typedef Common::HashMap<byte, Glyph> GlyphCache;
-	GlyphCache _glyphs;
-
-	FT_UInt _glyphSlots[256];
+	bool cacheGlyph(Glyph &glyph, uint32 chr) const;
+	typedef Common::HashMap<uint32, Glyph> GlyphCache;
+	mutable GlyphCache _glyphs;
+	bool _allowLateCaching;
+	void assureCached(uint32 chr) const;
 
 	bool _monochrome;
 	bool _hasKerning;
@@ -140,7 +141,7 @@ private:
 
 TTFFont::TTFFont()
     : _initialized(false), _face(), _ttfFile(0), _size(0), _width(0), _height(0), _ascent(0),
-      _descent(0), _glyphs(), _glyphSlots(), _monochrome(false), _hasKerning(false) {
+      _descent(0), _glyphs(), _monochrome(false), _hasKerning(false), _allowLateCaching(false) {
 }
 
 TTFFont::~TTFFont() {
@@ -212,19 +213,26 @@ bool TTFFont::load(Common::SeekableReadStream &stream, int size, uint dpi, bool
 	_height = _ascent - _descent + 1;
 
 	if (!mapping) {
+		// Allow loading of all unicode characters.
+		_allowLateCaching = true;
+
 		// Load all ISO-8859-1 characters.
 		for (uint i = 0; i < 256; ++i) {
-			if (!cacheGlyph(_glyphs[i], _glyphSlots[i], i))
-				_glyphSlots[i] = 0;
+			if (!cacheGlyph(_glyphs[i], i)) {
+				_glyphs.erase(i);
+			}
 		}
 	} else {
+		// We have a fixed map of characters do not load more later.
+		_allowLateCaching = false;
+
 		for (uint i = 0; i < 256; ++i) {
 			const uint32 unicode = mapping[i] & 0x7FFFFFFF;
 			const bool isRequired = (mapping[i] & 0x80000000) != 0;
 			// Check whether loading an important glyph fails and error out if
 			// that is the case.
-			if (!cacheGlyph(_glyphs[i], _glyphSlots[i], unicode)) {
-				_glyphSlots[i] = 0;
+			if (!cacheGlyph(_glyphs[i], unicode)) {
+				_glyphs.erase(i);
 				if (isRequired)
 					return false;
 			}
@@ -243,7 +251,8 @@ int TTFFont::getMaxCharWidth() const {
 	return _width;
 }
 
-int TTFFont::getCharWidth(byte chr) const {
+int TTFFont::getCharWidth(uint32 chr) const {
+	assureCached(chr);
 	GlyphCache::const_iterator glyphEntry = _glyphs.find(chr);
 	if (glyphEntry == _glyphs.end())
 		return 0;
@@ -251,12 +260,29 @@ int TTFFont::getCharWidth(byte chr) const {
 		return glyphEntry->_value.advance;
 }
 
-int TTFFont::getKerningOffset(byte left, byte right) const {
+int TTFFont::getKerningOffset(uint32 left, uint32 right) const {
 	if (!_hasKerning)
 		return 0;
 
-	FT_UInt leftGlyph = _glyphSlots[left];
-	FT_UInt rightGlyph = _glyphSlots[right];
+	assureCached(left);
+	assureCached(right);
+
+	FT_UInt leftGlyph, rightGlyph;
+	GlyphCache::const_iterator glyphEntry;
+
+	glyphEntry = _glyphs.find(left);
+	if (glyphEntry != _glyphs.end()) {
+		leftGlyph = glyphEntry->_value.slot;
+	} else {
+		return 0;
+	}
+
+	glyphEntry = _glyphs.find(right);
+	if (glyphEntry != _glyphs.end()) {
+		rightGlyph = glyphEntry->_value.slot;
+	} else {
+		return 0;
+	}
 
 	if (!leftGlyph || !rightGlyph)
 		return 0;
@@ -304,7 +330,8 @@ void renderGlyph(uint8 *dstPos, const int dstPitch, const uint8 *srcPos, const i
 
 } // End of anonymous namespace
 
-void TTFFont::drawChar(Surface *dst, byte chr, int x, int y, uint32 color) const {
+void TTFFont::drawChar(Surface *dst, uint32 chr, int x, int y, uint32 color) const {
+	assureCached(chr);
 	GlyphCache::const_iterator glyphEntry = _glyphs.find(chr);
 	if (glyphEntry == _glyphs.end())
 		return;
@@ -322,7 +349,7 @@ void TTFFont::drawChar(Surface *dst, byte chr, int x, int y, uint32 color) const
 	int w = glyph.image.w;
 	int h = glyph.image.h;
 
-	const uint8 *srcPos = (const uint8 *)glyph.image.getBasePtr(0, 0);
+	const uint8 *srcPos = (const uint8 *)glyph.image.getPixels();
 
 	// Make sure we are not drawing outside the screen bounds
 	if (x < 0) {
@@ -376,11 +403,13 @@ void TTFFont::drawChar(Surface *dst, byte chr, int x, int y, uint32 color) const
 	}
 }
 
-bool TTFFont::cacheGlyph(Glyph &glyph, FT_UInt &slot, uint chr) {
-	slot = FT_Get_Char_Index(_face, chr);
+bool TTFFont::cacheGlyph(Glyph &glyph, uint32 chr) const {
+	FT_UInt slot = FT_Get_Char_Index(_face, chr);
 	if (!slot)
 		return false;
 
+	glyph.slot = slot;
+
 	// We use the light target and render mode to improve the looks of the
 	// glyphs. It is most noticable in FreeSansBold.ttf, where otherwise the
 	// 't' glyph looks like it is cut off on the right side.
@@ -422,7 +451,7 @@ bool TTFFont::cacheGlyph(Glyph &glyph, FT_UInt &slot, uint chr) {
 		srcPitch = -srcPitch;
 	}
 
-	uint8 *dst = (uint8 *)glyph.image.getBasePtr(0, 0);
+	uint8 *dst = (uint8 *)glyph.image.getPixels();
 	memset(dst, 0, glyph.image.h * glyph.image.pitch);
 
 	switch (bitmap.pixel_mode) {
@@ -456,12 +485,24 @@ bool TTFFont::cacheGlyph(Glyph &glyph, FT_UInt &slot, uint chr) {
 
 	default:
 		warning("TTFFont::cacheGlyph: Unsupported pixel mode %d", bitmap.pixel_mode);
+		glyph.image.free();
 		return false;
 	}
 
 	return true;
 }
 
+void TTFFont::assureCached(uint32 chr) const {
+	if (!chr || !_allowLateCaching || _glyphs.contains(chr)) {
+		return;
+	}
+
+	Glyph newGlyph;
+	if (cacheGlyph(newGlyph, chr)) {
+		_glyphs[chr] = newGlyph;
+	}
+}
+
 Font *loadTTFFont(Common::SeekableReadStream &stream, int size, uint dpi, bool monochrome, const uint32 *mapping) {
 	TTFFont *font = new TTFFont();
 
diff --git a/graphics/fonts/winfont.cpp b/graphics/fonts/winfont.cpp
index 3bad92236d..c0ebab19ba 100644
--- a/graphics/fonts/winfont.cpp
+++ b/graphics/fonts/winfont.cpp
@@ -200,7 +200,7 @@ char WinFont::indexToCharacter(uint16 index) const {
 	return index + _firstChar;
 }
 
-uint16 WinFont::characterToIndex(byte character) const {
+uint16 WinFont::characterToIndex(uint32 character) const {
 	// Go to the default character if we didn't find a mapping
 	if (character < _firstChar || character > _lastChar)
 		character = _defaultChar;
@@ -208,7 +208,7 @@ uint16 WinFont::characterToIndex(byte character) const {
 	return character - _firstChar;
 }
 
-int WinFont::getCharWidth(byte chr) const {
+int WinFont::getCharWidth(uint32 chr) const {
 	return _glyphs[characterToIndex(chr)].charWidth;
 }
 
@@ -324,7 +324,7 @@ bool WinFont::loadFromFNT(Common::SeekableReadStream &stream) {
 	return true;
 }
 
-void WinFont::drawChar(Surface *dst, byte chr, int x, int y, uint32 color) const {
+void WinFont::drawChar(Surface *dst, uint32 chr, int x, int y, uint32 color) const {
 	assert(dst);
 	assert(dst->format.bytesPerPixel == 1 || dst->format.bytesPerPixel == 2 || dst->format.bytesPerPixel == 4);
 	assert(_glyphs);
diff --git a/graphics/fonts/winfont.h b/graphics/fonts/winfont.h
index 4382d7ed6b..2c7ba07b41 100644
--- a/graphics/fonts/winfont.h
+++ b/graphics/fonts/winfont.h
@@ -62,8 +62,8 @@ public:
 	// Font API
 	int getFontHeight() const { return _pixHeight; }
 	int getMaxCharWidth() const { return _maxWidth; }
-	int getCharWidth(byte chr) const;
-	void drawChar(Surface *dst, byte chr, int x, int y, uint32 color) const;
+	int getCharWidth(uint32 chr) const;
+	void drawChar(Surface *dst, uint32 chr, int x, int y, uint32 color) const;
 
 private:
 	bool loadFromPE(const Common::String &fileName, const WinFontDirEntry &dirEntry);
@@ -72,7 +72,7 @@ private:
 	uint32 getFontIndex(Common::SeekableReadStream &stream, const WinFontDirEntry &dirEntry);
 	bool loadFromFNT(Common::SeekableReadStream &stream);
 	char indexToCharacter(uint16 index) const;
-	uint16 characterToIndex(byte character) const;
+	uint16 characterToIndex(uint32 character) const;
 
 	uint16 _pixHeight;
 	uint16 _maxWidth;
diff --git a/graphics/scaler/thumbnail_intern.cpp b/graphics/scaler/thumbnail_intern.cpp
index 88f3cc2077..c30fc3b6fd 100644
--- a/graphics/scaler/thumbnail_intern.cpp
+++ b/graphics/scaler/thumbnail_intern.cpp
@@ -42,8 +42,10 @@ uint16 quadBlockInterpolate(const uint8 *src, uint32 srcPitch) {
 
 template<int bitFormat>
 void createThumbnail_2(const uint8 *src, uint32 srcPitch, uint8 *dstPtr, uint32 dstPitch, int width, int height) {
-	assert(width % 2 == 0);
-	assert(height % 2 == 0);
+	// Make sure the width and height is a multiple of 2.
+	width &= ~1;
+	height &= ~1;
+
 	for (int y = 0; y < height; y += 2) {
 		for (int x = 0; x < width; x += 2, dstPtr += 2) {
 			*((uint16 *)dstPtr) = quadBlockInterpolate<bitFormat>(src + 2 * x, srcPitch);
@@ -55,8 +57,10 @@ void createThumbnail_2(const uint8 *src, uint32 srcPitch, uint8 *dstPtr, uint32
 
 template<int bitFormat>
 void createThumbnail_4(const uint8 *src, uint32 srcPitch, uint8 *dstPtr, uint32 dstPitch, int width, int height) {
-	assert(width % 4 == 0);
-	assert(height % 4 == 0);
+	// Make sure the width and height is a multiple of 4
+	width &= ~3;
+	height &= ~3;
+
 	for (int y = 0; y < height; y += 4) {
 		for (int x = 0; x < width; x += 4, dstPtr += 2) {
 			uint16 upleft = quadBlockInterpolate<bitFormat>(src + 2 * x, srcPitch);
@@ -71,17 +75,87 @@ void createThumbnail_4(const uint8 *src, uint32 srcPitch, uint8 *dstPtr, uint32
 	}
 }
 
-static void createThumbnail(const uint8 *src, uint32 srcPitch, uint8 *dstPtr, uint32 dstPitch, int width, int height) {
-	// only 1/2 and 1/4 downscale supported
-	if (width != 320 && width != 640)
-		return;
+static void scaleThumbnail(Graphics::Surface &in, Graphics::Surface &out) {
+	while (in.w / out.w >= 4 || in.h / out.h >= 4) {
+		createThumbnail_4<565>((const uint8 *)in.getPixels(), in.pitch, (uint8 *)in.getPixels(), in.pitch, in.w, in.h);
+		in.w /= 4;
+		in.h /= 4;
+	}
+
+	while (in.w / out.w >= 2 || in.h / out.h >= 2) {
+		createThumbnail_2<565>((const uint8 *)in.getPixels(), in.pitch, (uint8 *)in.getPixels(), in.pitch, in.w, in.h);
+		in.w /= 2;
+		in.h /= 2;
+	}
+
+	if ((in.w == out.w && in.h < out.h) || (in.w < out.w && in.h == out.h)) {
+		// In this case we simply center the input surface in the output
+		uint8 *dst = (uint8 *)out.getBasePtr((out.w - in.w) / 2, (out.h - in.h) / 2);
+		const uint8 *src = (const uint8 *)in.getPixels();
 
-	int downScaleMode = (width == 320) ? 2 : 4;
+		for (int y = 0; y < in.h; ++y) {
+			memcpy(dst, src, in.w * in.format.bytesPerPixel);
+			src += in.pitch;
+			dst += out.pitch;
+		}
+	} else {
+		// Assure the aspect of the scaled image still matches the original.
+		int targetWidth = out.w, targetHeight = out.h;
 
-	if (downScaleMode == 2) {
-		createThumbnail_2<565>(src, srcPitch, dstPtr, dstPitch, width, height);
-	} else if (downScaleMode == 4) {
-		createThumbnail_4<565>(src, srcPitch, dstPtr, dstPitch, width, height);
+		const float inputAspect = (float)in.w / in.h;
+		const float outputAspect = (float)out.w / out.h;
+
+		if (inputAspect > outputAspect) {
+			targetHeight = int(targetWidth / inputAspect);
+		} else if (inputAspect < outputAspect) {
+			targetWidth = int(targetHeight * inputAspect);
+		}
+
+		// Make sure we are still in the bounds of the output
+		assert(targetWidth <= out.w);
+		assert(targetHeight <= out.h);
+
+		// Center the image on the output surface
+		byte *dst = (byte *)out.getBasePtr((out.w - targetWidth) / 2, (out.h - targetHeight) / 2);
+		const uint dstLineIncrease = out.pitch - targetWidth * out.format.bytesPerPixel;
+
+		const float scaleFactorX = (float)targetWidth / in.w;
+		const float scaleFactorY = (float)targetHeight / in.h;
+
+		for (int y = 0; y < targetHeight; ++y) {
+			const float yFrac = (y / scaleFactorY);
+			const int y1 = (int)yFrac;
+			const int y2 = (y1 + 1 < in.h) ? (y1 + 1) : (in.h - 1);
+
+			for (int x = 0; x < targetWidth; ++x) {
+				const float xFrac = (x / scaleFactorX);
+				const int x1 = (int)xFrac;
+				const int x2 = (x1 + 1 < in.w) ? (x1 + 1) : (in.w - 1);
+
+				// Look up colors at the points
+				uint8 p1R, p1G, p1B;
+				Graphics::colorToRGB<Graphics::ColorMasks<565> >(READ_UINT16(in.getBasePtr(x1, y1)), p1R, p1G, p1B);
+				uint8 p2R, p2G, p2B;
+				Graphics::colorToRGB<Graphics::ColorMasks<565> >(READ_UINT16(in.getBasePtr(x2, y1)), p2R, p2G, p2B);
+				uint8 p3R, p3G, p3B;
+				Graphics::colorToRGB<Graphics::ColorMasks<565> >(READ_UINT16(in.getBasePtr(x1, y2)), p3R, p3G, p3B);
+				uint8 p4R, p4G, p4B;
+				Graphics::colorToRGB<Graphics::ColorMasks<565> >(READ_UINT16(in.getBasePtr(x2, y2)), p4R, p4G, p4B);
+
+				const float xDiff = xFrac - x1;
+				const float yDiff = yFrac - y1;
+
+				uint8 pR = (uint8)((1 - yDiff) * ((1 - xDiff) * p1R + xDiff * p2R) + yDiff * ((1 - xDiff) * p3R + xDiff * p4R));
+				uint8 pG = (uint8)((1 - yDiff) * ((1 - xDiff) * p1G + xDiff * p2G) + yDiff * ((1 - xDiff) * p3G + xDiff * p4G));
+				uint8 pB = (uint8)((1 - yDiff) * ((1 - xDiff) * p1B + xDiff * p2B) + yDiff * ((1 - xDiff) * p3B + xDiff * p4B));
+
+				WRITE_UINT16(dst, Graphics::RGBToColor<Graphics::ColorMasks<565> >(pR, pG, pB));
+				dst += 2;
+			}
+
+			// Move to the next line
+			dst = (byte *)dst + dstLineIncrease;
+		}
 	}
 }
 
@@ -90,7 +164,7 @@ static void createThumbnail(const uint8 *src, uint32 srcPitch, uint8 *dstPtr, ui
  * Copies the current screen contents to a new surface, using RGB565 format.
  * WARNING: surf->free() must be called by the user to avoid leaking.
  *
- * @param surf		the surface to store the data in it
+ * @param surf      the surface to store the data in it
  */
 static bool grabScreen565(Graphics::Surface *surf) {
 	Graphics::Surface *screen = g_system->lockScreen();
@@ -98,7 +172,7 @@ static bool grabScreen565(Graphics::Surface *surf) {
 		return false;
 
 	assert(screen->format.bytesPerPixel == 1 || screen->format.bytesPerPixel == 2);
-	assert(screen->pixels != 0);
+	assert(screen->getPixels() != 0);
 
 	Graphics::PixelFormat screenFormat = g_system->getScreenFormat();
 
@@ -116,15 +190,16 @@ static bool grabScreen565(Graphics::Surface *surf) {
 			byte r = 0, g = 0, b = 0;
 
 			if (screenFormat.bytesPerPixel == 1) {
-				r = palette[((uint8 *)screen->pixels)[y * screen->pitch + x] * 3];
-				g = palette[((uint8 *)screen->pixels)[y * screen->pitch + x] * 3 + 1];
-				b = palette[((uint8 *)screen->pixels)[y * screen->pitch + x] * 3 + 2];
+				uint8 pixel = *(uint8 *)screen->getBasePtr(x, y);
+				r = palette[pixel * 3 + 0];
+				g = palette[pixel * 3 + 1];
+				b = palette[pixel * 3 + 2];
 			} else if (screenFormat.bytesPerPixel == 2) {
 				uint16 col = READ_UINT16(screen->getBasePtr(x, y));
 				screenFormat.colorToRGB(col, r, g, b);
 			}
 
-			((uint16 *)surf->pixels)[y * surf->w + x] = Graphics::RGBToColor<Graphics::ColorMasks<565> >(r, g, b);
+			*((uint16 *)surf->getBasePtr(x, y)) = Graphics::RGBToColor<Graphics::ColorMasks<565> >(r, g, b);
 		}
 	}
 
@@ -135,78 +210,20 @@ static bool grabScreen565(Graphics::Surface *surf) {
 }
 
 static bool createThumbnail(Graphics::Surface &out, Graphics::Surface &in) {
-	uint16 width = in.w;
-	uint16 inHeight = in.h;
-
-	if (width < 320) {
-		// Special case to handle MM NES (uses a screen width of 256)
-		width = 320;
-
-		// center MM NES screen
-		Graphics::Surface newscreen;
-		newscreen.create(width, in.h, in.format);
-
-		uint8 *dst = (uint8 *)newscreen.getBasePtr((320 - in.w) / 2, 0);
-		const uint8 *src = (const uint8 *)in.getBasePtr(0, 0);
-		uint16 height = in.h;
-
-		while (height--) {
-			memcpy(dst, src, in.pitch);
-			dst += newscreen.pitch;
-			src += in.pitch;
-		}
-
-		in.free();
-		in = newscreen;
-	} else if (width == 720) {
-		// Special case to handle Hercules mode
-		//
-		// NOTE: This code is pretty SCUMM specific.
-		// For other games this code might cut off
-		// not only the menu, but also other graphics.
-		width = 640;
-		inHeight = 400;
-
-		// cut off menu and so on..
-		Graphics::Surface newscreen;
-		newscreen.create(width, 400, in.format);
-
-		uint8 *dst = (uint8 *)newscreen.getBasePtr(0, (400 - 240) / 2);
-		const uint8 *src = (const uint8 *)in.getBasePtr(41, 28);
-
-		for (int y = 0; y < 240; ++y) {
-			memcpy(dst, src, 640 * in.format.bytesPerPixel);
-			dst += newscreen.pitch;
-			src += in.pitch;
-		}
-
-		in.free();
-		in = newscreen;
-	} else if (width == 640 && inHeight == 440) {
-		// Special case to handle KQ6 Windows: resize the screen to 640x480,
-		// adding a black band in the bottom.
-		inHeight = 480;
-
-		Graphics::Surface newscreen;
-		newscreen.create(width, 480, in.format);
-
-		memcpy(newscreen.getBasePtr(0, 0), in.getBasePtr(0, 0), width * 440 * in.format.bytesPerPixel);
-
-		in.free();
-		in = newscreen;
+	int height;
+	if ((in.w == 320 && in.h == 200) || (in.w == 640 && in.h == 400)) {
+		height = kThumbnailHeight1;
+	} else {
+		height = kThumbnailHeight2;
 	}
 
-	uint16 newHeight = !(inHeight % 240) ? kThumbnailHeight2 : kThumbnailHeight1;
-
-	out.create(kThumbnailWidth, newHeight, Graphics::PixelFormat(2, 5, 6, 5, 0, 11, 5, 0, 0));
-	createThumbnail((const uint8 *)in.pixels, width * sizeof(uint16), (uint8 *)out.pixels, out.pitch, width, inHeight);
-
+	out.create(kThumbnailWidth, height, Graphics::PixelFormat(2, 5, 6, 5, 0, 11, 5, 0, 0));
+	scaleThumbnail(in, out);
 	in.free();
-
 	return true;
 }
 
-bool createThumbnailFromScreen(Graphics::Surface* surf) {
+bool createThumbnailFromScreen(Graphics::Surface *surf) {
 	assert(surf);
 
 	Graphics::Surface screen;
@@ -230,9 +247,37 @@ bool createThumbnail(Graphics::Surface *surf, const uint8 *pixels, int w, int h,
 			g = palette[pixels[y * w + x] * 3 + 1];
 			b = palette[pixels[y * w + x] * 3 + 2];
 
-			((uint16 *)screen.pixels)[y * screen.w + x] = Graphics::RGBToColor<Graphics::ColorMasks<565> >(r, g, b);
+			*((uint16 *)screen.getBasePtr(x, y)) = Graphics::RGBToColor<Graphics::ColorMasks<565> >(r, g, b);
 		}
 	}
 
 	return createThumbnail(*surf, screen);
 }
+
+// this is somewhat awkward, but createScreenShot should logically be in graphics,
+// but moving other functions in this file into that namespace breaks several engines
+namespace Graphics {
+bool createScreenShot(Graphics::Surface &surf) {
+	Graphics::PixelFormat screenFormat = g_system->getScreenFormat();
+	//convert surface to 2 bytes pixel format to avoid problems with palette saving and loading
+	if ((screenFormat.bytesPerPixel == 1) || (screenFormat.bytesPerPixel == 2)) {
+		return grabScreen565(&surf);
+	} else {
+		Graphics::Surface *screen = g_system->lockScreen();
+		if (!screen) {
+			return false;
+		}
+		surf.create(screen->w, screen->h, Graphics::PixelFormat(4, 8, 8, 8, 8, 24, 16, 8, 0));
+		for (uint y = 0; y < screen->h; ++y) {
+			for (uint x = 0; x < screen->w; ++x) {
+				byte r = 0, g = 0, b = 0, a = 0;
+				uint32 col = READ_UINT32(screen->getBasePtr(x, y));
+				screenFormat.colorToARGB(col, a, r, g, b);
+				*((uint32 *)surf.getBasePtr(x, y)) = Graphics::ARGBToColor<Graphics::ColorMasks<8888> >(a, r, g, b);
+			}
+		}
+		g_system->unlockScreen();
+		return true;
+	}
+}
+} // End of namespace Graphics
diff --git a/graphics/surface.cpp b/graphics/surface.cpp
index 41ae8dcebb..777c1058fb 100644
--- a/graphics/surface.cpp
+++ b/graphics/surface.cpp
@@ -82,9 +82,79 @@ void Surface::free() {
 	format = PixelFormat();
 }
 
+void Surface::init(uint16 width, uint16 height, uint16 newPitch, void *newPixels, const PixelFormat &f) {
+	w = width;
+	h = height;
+	pitch = newPitch;
+	pixels = newPixels;
+	format = f;
+}
+
 void Surface::copyFrom(const Surface &surf) {
 	create(surf.w, surf.h, surf.format);
-	memcpy(pixels, surf.pixels, h * pitch);
+	if (surf.pitch == pitch) {
+		memcpy(pixels, surf.pixels, h * pitch);
+	} else {
+		const byte *src = (const byte *)surf.pixels;
+		byte *dst = (byte *)pixels;
+		for (int y = h; y > 0; --y) {
+			memcpy(dst, src, w * format.bytesPerPixel);
+			src += surf.pitch;
+			dst += pitch;
+		}
+	}
+}
+
+Surface Surface::getSubArea(const Common::Rect &area) {
+	Common::Rect effectiveArea(area);
+	effectiveArea.clip(w, h);
+
+	Surface subSurface;
+	subSurface.w = effectiveArea.width();
+	subSurface.h = effectiveArea.height();
+	subSurface.pitch = pitch;
+	subSurface.pixels = getBasePtr(area.left, area.top);
+	subSurface.format = format;
+	return subSurface;
+}
+
+const Surface Surface::getSubArea(const Common::Rect &area) const {
+	Common::Rect effectiveArea(area);
+	effectiveArea.clip(w, h);
+
+	Surface subSurface;
+	subSurface.w = effectiveArea.width();
+	subSurface.h = effectiveArea.height();
+	subSurface.pitch = pitch;
+	// We need to cast the const away here because a Surface always has a
+	// pointer to modifiable pixel data.
+	subSurface.pixels = const_cast<void *>(getBasePtr(area.left, area.top));
+	subSurface.format = format;
+	return subSurface;
+}
+
+void Surface::copyRectToSurface(const void *buffer, int srcPitch, int destX, int destY, int width, int height) {
+	assert(buffer);
+
+	assert(destX >= 0 && destX < w);
+	assert(destY >= 0 && destY < h);
+	assert(height > 0 && destY + height <= h);
+	assert(width > 0 && destX + width <= w);
+
+	// Copy buffer data to internal buffer
+	const byte *src = (const byte *)buffer;
+	byte *dst = (byte *)getBasePtr(destX, destY);
+	for (int i = 0; i < height; i++) {
+		memcpy(dst, src, width * format.bytesPerPixel);
+		src += srcPitch;
+		dst += pitch;
+	}
+}
+
+void Surface::copyRectToSurface(const Graphics::Surface &srcSurface, int destX, int destY, const Common::Rect subRect) {
+	assert(srcSurface.format == format);
+
+	copyRectToSurface(srcSurface.getBasePtr(subRect.left, subRect.top), srcSurface.pitch, destX, destY, subRect.width(), subRect.height());
 }
 
 void Surface::hLine(int x, int y, int x2, uint32 color) {
diff --git a/graphics/surface.h b/graphics/surface.h
index 6c9e464657..f1b2aa64ab 100644
--- a/graphics/surface.h
+++ b/graphics/surface.h
@@ -61,11 +61,13 @@ struct Surface {
 	 */
 	uint16 pitch;
 
+protected:
 	/**
 	 * The surface's pixel data.
 	 */
 	void *pixels;
 
+public:
 	/**
 	 * The pixel format of the surface.
 	 */
@@ -78,6 +80,33 @@ struct Surface {
 	}
 
 	/**
+	 * Return a pointer to the pixel data.
+	 *
+	 * @return Pointer to the pixel data.
+	 */
+	inline const void *getPixels() const {
+		return pixels;
+	}
+
+	/**
+	 * Return a pointer to the pixel data.
+	 *
+	 * @return Pointer to the pixel data.
+	 */
+	inline void *getPixels() {
+		return pixels;
+	}
+
+	/**
+	 * Sets the pixel data.
+	 *
+	 * Note that this is a simply a setter. Be careful what you are doing!
+	 *
+	 * @param newPixels The new pixel data.
+	 */
+	void setPixels(void *newPixels) { pixels = newPixels; }
+
+	/**
 	 * Return a pointer to the pixel at the specified point.
 	 *
 	 * @param x The x coordinate of the pixel.
@@ -122,6 +151,20 @@ struct Surface {
 	void free();
 
 	/**
+	 * Set up the Surface with user specified data.
+	 *
+	 * Note that this simply sets the 'internal' attributes of the Surface. It
+	 * will not take care of freeing old data via free or similar!
+	 *
+	 * @param width Width of the pixel data.
+	 * @param height Height of the pixel data.
+	 * @param pitch The pitch of the pixel data.
+	 * @param pixels The pixel data itself.
+	 * @param format The pixel format of the pixel data.
+	 */
+	void init(uint16 width, uint16 height, uint16 pitch, void *pixels, const PixelFormat &format);
+
+	/**
 	 * Copy the data from another Surface.
 	 *
 	 * Note that this calls free on the current surface, to assure it being
@@ -135,10 +178,66 @@ struct Surface {
 	void copyFrom(const Surface &surf);
 
 	/**
+	 * Creates a Surface which represents a sub-area of this Surface object.
+	 *
+	 * The pixel (0, 0) of the returned Surface will be the same as Pixel
+	 * (area.x, area.y) of this Surface. Changes to any of the Surface objects
+	 * will change the shared pixel data.
+	 *
+	 * Note that the Surface returned is only valid as long as this Surface
+	 * object is still alive (i.e. its pixel data is not destroyed or
+	 * reallocated). Do *never* try to free the returned Surface.
+	 *
+	 * @param area The area which should be represented. Note that the area
+	 *             will get clipped in case it does not fit!
+	 */
+	Surface getSubArea(const Common::Rect &area);
+
+	/**
+	 * Creates a Surface which represents a sub-area of this Surface object.
+	 *
+	 * The pixel (0, 0) of the returned Surface will be the same as Pixel
+	 * (area.x, area.y) of this Surface.
+	 *
+	 * Note that the Surface returned is only valid as long as this Surface
+	 * object is still alive (i.e. its pixel data is not destroyed or
+	 * reallocated). Do *never* try to free the returned Surface.
+	 *
+	 * @param area The area which should be represented. Note that the area
+	 *             will get clipped in case it does not fit!
+	 */
+	const Surface getSubArea(const Common::Rect &area) const;
+
+	/**
+	 * Copies a bitmap to the Surface internal buffer. The pixel format
+	 * of buffer must match the pixel format of the Surface.
+	 *
+	 * @param buffer    The buffer containing the graphics data source
+	 * @param pitch     The pitch of the buffer (number of bytes in a scanline)
+	 * @param destX     The x coordinate of the destination rectangle
+	 * @param destY     The y coordinate of the destination rectangle
+	 * @param width     The width of the destination rectangle
+	 * @param height    The height of the destination rectangle
+	 */
+	void copyRectToSurface(const void *buffer, int srcPitch, int destX, int destY, int width, int height);
+	/**
+	 * Copies a bitmap to the Surface internal buffer. The pixel format
+	 * of buffer must match the pixel format of the Surface.
+	 *
+	 * @param srcSurface    The source of the bitmap data
+	 * @param destX         The x coordinate of the destination rectangle
+	 * @param destY         The y coordinate of the destination rectangle
+	 * @param subRect       The subRect of surface to be blitted
+	 * @return                
+	 */
+	void copyRectToSurface(const Graphics::Surface &srcSurface, int destX, int destY, const Common::Rect subRect);
+
+	/**
 	 * Convert the data to another pixel format.
 	 *
 	 * This works in-place. This means it will not create an additional buffer
-	 * for the conversion process. The value of pixels might change though.
+	 * for the conversion process. The value of 'pixels' might change though
+	 * (that means it might realloc the pixel data).
 	 *
 	 * Note that you should only use this, when you created the Surface data via
 	 * create! Otherwise this function has undefined behavior.
@@ -235,7 +334,9 @@ struct Surface {
  */
 struct SharedPtrSurfaceDeleter {
 	void operator()(Surface *ptr) {
-		ptr->free();
+		if (ptr) {
+			ptr->free();
+		}
 		delete ptr;
 	}
 };
diff --git a/graphics/thumbnail.cpp b/graphics/thumbnail.cpp
index ddb377306d..e3f368dffa 100644
--- a/graphics/thumbnail.cpp
+++ b/graphics/thumbnail.cpp
@@ -23,6 +23,7 @@
 #include "graphics/scaler.h"
 #include "graphics/colormasks.h"
 #include "common/endian.h"
+#include "common/algorithm.h"
 #include "common/system.h"
 #include "common/stream.h"
 #include "common/textconsole.h"
@@ -42,7 +43,16 @@ struct ThumbnailHeader {
 
 #define ThumbnailHeaderSize (4+4+1+2+2+(1+4+4))
 
-bool loadHeader(Common::SeekableReadStream &in, ThumbnailHeader &header, bool outputWarnings) {
+enum HeaderState {
+	/// There is no header present
+	kHeaderNone,
+	/// The header present only has reliable values for version and size
+	kHeaderUnsupported,
+	/// The header is present and the version is supported
+	kHeaderPresent
+};
+
+HeaderState loadHeader(Common::SeekableReadStream &in, ThumbnailHeader &header, bool outputWarnings) {
 	header.type = in.readUint32BE();
 	// We also accept the bad 'BMHT' header here, for the sake of compatibility
 	// with some older savegames which were written incorrectly due to a bug in
@@ -50,16 +60,28 @@ bool loadHeader(Common::SeekableReadStream &in, ThumbnailHeader &header, bool ou
 	if (header.type != MKTAG('T','H','M','B') && header.type != MKTAG('B','M','H','T')) {
 		if (outputWarnings)
 			warning("couldn't find thumbnail header type");
-		return false;
+		return kHeaderNone;
 	}
 
 	header.size = in.readUint32BE();
 	header.version = in.readByte();
 
+	// Do a check whether any read errors had occured. If so we cannot use the
+	// values obtained for size and version because they might be bad.
+	if (in.err() || in.eos()) {
+		// TODO: We fake that there is no header. This is actually not quite
+		// correct since we found the start of the header and then things
+		// started to break. Right no we leave detection of this to the client.
+		// Since this case is caused by broken files, the client code should
+		// catch it anyway... If there is a nicer solution here, we should
+		// implement it.
+		return kHeaderNone;
+	}
+
 	if (header.version > THMB_VERSION) {
 		if (outputWarnings)
 			warning("trying to load a newer thumbnail version: %d instead of %d", header.version, THMB_VERSION);
-		return false;
+		return kHeaderUnsupported;
 	}
 
 	header.width = in.readUint16BE();
@@ -81,7 +103,15 @@ bool loadHeader(Common::SeekableReadStream &in, ThumbnailHeader &header, bool ou
 		header.format = createPixelFormat<565>();
 	}
 
-	return true;
+	if (in.err() || in.eos()) {
+		// When we reached this point we know that at least the size and
+		// version field was loaded successfully, thus we tell this header
+		// is not supported and silently hope that the client code is
+		// prepared to handle read errors.
+		return kHeaderUnsupported;
+	} else {
+		return kHeaderPresent;
+	}
 }
 } // end of anonymous namespace
 
@@ -89,7 +119,12 @@ bool checkThumbnailHeader(Common::SeekableReadStream &in) {
 	uint32 position = in.pos();
 	ThumbnailHeader header;
 
-	bool hasHeader = loadHeader(in, header, false);
+	// TODO: It is not clear whether this is the best semantics. Now
+	// checkThumbnailHeader will return true even when the thumbnail header
+	// found is actually not usable. However, most engines seem to use this
+	// to detect the presence of any header and if there is none it wont even
+	// try to skip it. Thus, this looks like the best solution for now...
+	bool hasHeader = (loadHeader(in, header, false) != kHeaderNone);
 
 	in.seek(position, SEEK_SET);
 
@@ -100,7 +135,9 @@ bool skipThumbnail(Common::SeekableReadStream &in) {
 	uint32 position = in.pos();
 	ThumbnailHeader header;
 
-	if (!loadHeader(in, header, false)) {
+	// We can skip unsupported and supported headers. So we only seek back
+	// to the old position in case there is no header at all.
+	if (loadHeader(in, header, false) == kHeaderNone) {
 		in.seek(position, SEEK_SET);
 		return false;
 	}
@@ -110,10 +147,23 @@ bool skipThumbnail(Common::SeekableReadStream &in) {
 }
 
 Graphics::Surface *loadThumbnail(Common::SeekableReadStream &in) {
+	const uint32 position = in.pos();
 	ThumbnailHeader header;
-
-	if (!loadHeader(in, header, true))
+	HeaderState headerState = loadHeader(in, header, true);
+
+	// Try to handle unsupported/broken headers gracefully. If there is no
+	// header at all, we seek back and return at this point. If there is an
+	// unsupported/broken header, we skip the actual data and return. The
+	// downside is that we might reset the end of stream flag with this and
+	// the client code would not be able to notice a read past the end of the
+	// stream at this point then.
+	if (headerState == kHeaderNone) {
+		in.seek(position, SEEK_SET);
+		return 0;
+	} else if (headerState == kHeaderUnsupported) {
+		in.seek(header.size - (in.pos() - position), SEEK_CUR);
 		return 0;
+	}
 
 	if (header.format.bytesPerPixel != 2 && header.format.bytesPerPixel != 4) {
 		warning("trying to load thumbnail with unsupported bit depth %d", header.format.bytesPerPixel);
@@ -143,7 +193,6 @@ Graphics::Surface *loadThumbnail(Common::SeekableReadStream &in) {
 			assert(0);
 		}
 	}
-
 	return to;
 }
 
@@ -216,4 +265,55 @@ bool saveThumbnail(Common::WriteStream &out, const Graphics::Surface &thumb) {
 	return true;
 }
 
+
+/**
+ * Returns an array indicating which pixels of a source image horizontally or vertically get
+ * included in a scaled image
+ */
+int *scaleLine(int size, int srcSize) {
+	int scale = 100 * size / srcSize;
+	assert(scale > 0);
+	int *v = new int[size];
+	Common::fill(v, &v[size], 0);
+
+	int distCtr = 0;
+	int *destP = v;
+	for (int distIndex = 0; distIndex < srcSize; ++distIndex) {
+		distCtr += scale;
+		while (distCtr >= 100) {
+			assert(destP < &v[size]);
+			*destP++ = distIndex;
+			distCtr -= 100;
+		}
+	}
+
+	return v;
+}
+
+Graphics::Surface *scale(const Graphics::Surface &srcImage, int xSize, int ySize) {
+	Graphics::Surface *s = new Graphics::Surface();
+	s->create(xSize, ySize, srcImage.format);
+
+	int *horizUsage = scaleLine(xSize, srcImage.w);
+	int *vertUsage = scaleLine(ySize, srcImage.h);
+
+	// Loop to create scaled version
+	for (int yp = 0; yp < ySize; ++yp) {
+		const byte *srcP = (const byte *)srcImage.getBasePtr(0, vertUsage[yp]);
+		byte *destP = (byte *)s->getBasePtr(0, yp);
+
+		for (int xp = 0; xp < xSize; ++xp) {
+			const byte *tempSrcP = srcP + (horizUsage[xp] * srcImage.format.bytesPerPixel);
+			for (int byteCtr = 0; byteCtr < srcImage.format.bytesPerPixel; ++byteCtr) {
+				*destP++ = *tempSrcP++;
+			}
+		}
+	}
+
+	// Delete arrays and return surface
+	delete[] horizUsage;
+	delete[] vertUsage;
+	return s;
+}
+
 } // End of namespace Graphics
diff --git a/graphics/thumbnail.h b/graphics/thumbnail.h
index 45a0fdbf07..c857809c91 100644
--- a/graphics/thumbnail.h
+++ b/graphics/thumbnail.h
@@ -64,6 +64,24 @@ bool saveThumbnail(Common::WriteStream &out);
  */
 bool saveThumbnail(Common::WriteStream &out, const Graphics::Surface &thumb);
 
+/**
+ * Grabs framebuffer into surface
+ *
+ * @param surf	a surface
+ * @return		false if a error occurred
+ */
+bool createScreenShot(Graphics::Surface &surf);
+
+/**
+ * Scales a passed surface, creating a new surface with the result
+ * @param srcImage		Source image to scale
+ * @param xSize			New surface width
+ * @param ySize			New surface height
+ * @remarks Caller is responsible for freeing the returned surface
+ */
+Graphics::Surface *scale(const Graphics::Surface &srcImage, int xSize, int ySize);
+
+
 } // End of namespace Graphics
 
 #endif
diff --git a/graphics/yuv_to_rgb.cpp b/graphics/yuv_to_rgb.cpp
index 6043315a13..2a485fa664 100644
--- a/graphics/yuv_to_rgb.cpp
+++ b/graphics/yuv_to_rgb.cpp
@@ -229,7 +229,7 @@ void convertYUV444ToRGB(byte *dstPtr, int dstPitch, const YUVToRGBLookup *lookup
 
 void YUVToRGBManager::convert444(Graphics::Surface *dst, YUVToRGBManager::LuminanceScale scale, const byte *ySrc, const byte *uSrc, const byte *vSrc, int yWidth, int yHeight, int yPitch, int uvPitch) {
 	// Sanity checks
-	assert(dst && dst->pixels);
+	assert(dst && dst->getPixels());
 	assert(dst->format.bytesPerPixel == 2 || dst->format.bytesPerPixel == 4);
 	assert(ySrc && uSrc && vSrc);
 
@@ -237,9 +237,9 @@ void YUVToRGBManager::convert444(Graphics::Surface *dst, YUVToRGBManager::Lumina
 
 	// Use a templated function to avoid an if check on every pixel
 	if (dst->format.bytesPerPixel == 2)
-		convertYUV444ToRGB<uint16>((byte *)dst->pixels, dst->pitch, lookup, _colorTab, ySrc, uSrc, vSrc, yWidth, yHeight, yPitch, uvPitch);
+		convertYUV444ToRGB<uint16>((byte *)dst->getPixels(), dst->pitch, lookup, _colorTab, ySrc, uSrc, vSrc, yWidth, yHeight, yPitch, uvPitch);
 	else
-		convertYUV444ToRGB<uint32>((byte *)dst->pixels, dst->pitch, lookup, _colorTab, ySrc, uSrc, vSrc, yWidth, yHeight, yPitch, uvPitch);
+		convertYUV444ToRGB<uint32>((byte *)dst->getPixels(), dst->pitch, lookup, _colorTab, ySrc, uSrc, vSrc, yWidth, yHeight, yPitch, uvPitch);
 }
 
 template<typename PixelInt>
@@ -283,7 +283,7 @@ void convertYUV420ToRGB(byte *dstPtr, int dstPitch, const YUVToRGBLookup *lookup
 
 void YUVToRGBManager::convert420(Graphics::Surface *dst, YUVToRGBManager::LuminanceScale scale, const byte *ySrc, const byte *uSrc, const byte *vSrc, int yWidth, int yHeight, int yPitch, int uvPitch) {
 	// Sanity checks
-	assert(dst && dst->pixels);
+	assert(dst && dst->getPixels());
 	assert(dst->format.bytesPerPixel == 2 || dst->format.bytesPerPixel == 4);
 	assert(ySrc && uSrc && vSrc);
 	assert((yWidth & 1) == 0);
@@ -293,9 +293,9 @@ void YUVToRGBManager::convert420(Graphics::Surface *dst, YUVToRGBManager::Lumina
 
 	// Use a templated function to avoid an if check on every pixel
 	if (dst->format.bytesPerPixel == 2)
-		convertYUV420ToRGB<uint16>((byte *)dst->pixels, dst->pitch, lookup, _colorTab, ySrc, uSrc, vSrc, yWidth, yHeight, yPitch, uvPitch);
+		convertYUV420ToRGB<uint16>((byte *)dst->getPixels(), dst->pitch, lookup, _colorTab, ySrc, uSrc, vSrc, yWidth, yHeight, yPitch, uvPitch);
 	else
-		convertYUV420ToRGB<uint32>((byte *)dst->pixels, dst->pitch, lookup, _colorTab, ySrc, uSrc, vSrc, yWidth, yHeight, yPitch, uvPitch);
+		convertYUV420ToRGB<uint32>((byte *)dst->getPixels(), dst->pitch, lookup, _colorTab, ySrc, uSrc, vSrc, yWidth, yHeight, yPitch, uvPitch);
 }
 
 #define READ_QUAD(ptr, prefix) \
@@ -368,7 +368,7 @@ void convertYUV410ToRGB(byte *dstPtr, int dstPitch, const YUVToRGBLookup *lookup
 
 void YUVToRGBManager::convert410(Graphics::Surface *dst, YUVToRGBManager::LuminanceScale scale, const byte *ySrc, const byte *uSrc, const byte *vSrc, int yWidth, int yHeight, int yPitch, int uvPitch) {
 	// Sanity checks
-	assert(dst && dst->pixels);
+	assert(dst && dst->getPixels());
 	assert(dst->format.bytesPerPixel == 2 || dst->format.bytesPerPixel == 4);
 	assert(ySrc && uSrc && vSrc);
 	assert((yWidth & 3) == 0);
@@ -378,9 +378,9 @@ void YUVToRGBManager::convert410(Graphics::Surface *dst, YUVToRGBManager::Lumina
 
 	// Use a templated function to avoid an if check on every pixel
 	if (dst->format.bytesPerPixel == 2)
-		convertYUV410ToRGB<uint16>((byte *)dst->pixels, dst->pitch, lookup, _colorTab, ySrc, uSrc, vSrc, yWidth, yHeight, yPitch, uvPitch);
+		convertYUV410ToRGB<uint16>((byte *)dst->getPixels(), dst->pitch, lookup, _colorTab, ySrc, uSrc, vSrc, yWidth, yHeight, yPitch, uvPitch);
 	else
-		convertYUV410ToRGB<uint32>((byte *)dst->pixels, dst->pitch, lookup, _colorTab, ySrc, uSrc, vSrc, yWidth, yHeight, yPitch, uvPitch);
+		convertYUV410ToRGB<uint32>((byte *)dst->getPixels(), dst->pitch, lookup, _colorTab, ySrc, uSrc, vSrc, yWidth, yHeight, yPitch, uvPitch);
 }
 
 } // End of namespace Graphics
diff --git a/graphics/yuv_to_rgb.h b/graphics/yuv_to_rgb.h
index f785422c5a..a1e61ec705 100644
--- a/graphics/yuv_to_rgb.h
+++ b/graphics/yuv_to_rgb.h
@@ -22,10 +22,17 @@
 
 /**
  * @file
- * YUV to RGB conversion used in engines:
- * - mohawk
- * - scumm (he)
- * - sword25
+ * YUV to RGB conversion.
+ *
+ * Used in graphics:
+ * - JPEGDecoder
+ *
+ * Used in video:
+ * - BinkDecoder
+ * - Indeo3Decoder
+ * - PSXStreamDecoder
+ * - TheoraDecoder
+ * - SVQ1Decoder
  */
 
 #ifndef GRAPHICS_YUV_TO_RGB_H