Draw images with anti-aliased edges.

7 files changed, 217 insertions, 487 deletions

diff --git a/draw/archx86.c b/draw/archx86.c
index a32e54be..b75fb28f 100644
--- a/draw/archx86.c
+++ b/draw/archx86.c
@@ -77,6 +77,8 @@ static void duff_4i1o4mmx(byte *sp0, int sw, byte *mp0, int mw, byte *dp0, int d
 	_mm_empty();
 }
 
+#if 0 /* TODO */
+
 static inline unsigned
 getargb(unsigned *s, int w, int h, int u, int v)
 {
@@ -212,6 +214,8 @@ static void img_4o4mmx(FZ_PSRC, FZ_PDST, FZ_PCTM)
 	_mm_empty();
 }
 
+#endif
+
 #endif /* HAVE_MMX */
 
 #if defined (ARCH_X86) || defined(ARCH_X86_64)
@@ -222,7 +226,7 @@ fz_accelerate(void)
 	if (fz_cpuflags & HAVE_MMX)
 	{
 		fz_duff_4i1o4 = duff_4i1o4mmx;
-		fz_img_4o4 = img_4o4mmx;
+// TODO		fz_img_4o4 = img_4o4mmx;
 	}
 #  endif
 }
diff --git a/draw/imagedraw.c b/draw/imagedraw.c
index 7ef4964f..7bdba6e0 100644
--- a/draw/imagedraw.c
+++ b/draw/imagedraw.c
@@ -1,59 +1,54 @@
 #include "fitz.h"
 
-#define CLAMPUV
-
 typedef unsigned char byte;
 
-#define lerp(a,b,t) (a + (((b - a) * t) >> 16))
-
-static inline byte getcomp(byte *s, int w, int h, int u, int v, int n, int k)
+static inline byte
+getmask(byte *s, int w, int h, int u, int v)
 {
-#ifdef CLAMPUV
 	if (u < 0) u = 0;
 	if (v < 0) v = 0;
 	if (u >= w) u = w - 1;
 	if (v >= h) v = h - 1;
-#else
-	if (u < 0) return 0;
-	if (v < 0) return 0;
-	if (u >= w) return 0;
-	if (v >= h) return 0;
-#endif
-	return s[(w * v + u) * n + k];
+	return s[w * v + u];
 }
 
-static inline int samplecomp(byte *s, int w, int h, int u, int v, int n, int k)
+static inline byte *
+getargb(byte *s, int w, int h, int u, int v)
 {
-	int ui = u >> 16;
-	int vi = v >> 16;
-	int ud = u & 0xFFFF;
-	int vd = v & 0xFFFF;
-	int a = getcomp(s, w, h, ui, vi, n, k);
-	int b = getcomp(s, w, h, ui+1, vi, n, k);
-	int c = getcomp(s, w, h, ui, vi+1, n, k);
-	int d = getcomp(s, w, h, ui+1, vi+1, n, k);
-	int ab = lerp(a, b, ud);
-	int cd = lerp(c, d, ud);
-	return lerp(ab, cd, vd);
+	if (u < 0) u = 0;
+	if (v < 0) v = 0;
+	if (u >= w) u = w - 1;
+	if (v >= h) v = h - 1;
+	return s + ((w * v + u) << 2);
 }
 
-static inline byte getmask(byte *s, int w, int h, int u, int v)
+static inline int
+getcolor(byte *s, int w, int h, int n, int u, int v, int k)
 {
-#ifdef CLAMPUV
 	if (u < 0) u = 0;
 	if (v < 0) v = 0;
 	if (u >= w) u = w - 1;
 	if (v >= h) v = h - 1;
-#else
-	if (u < 0) return 0;
-	if (v < 0) return 0;
-	if (u >= w) return 0;
-	if (v >= h) return 0;
-#endif
-	return s[w * v + u];
+	return s[w * v * n + u + k];
+}
+
+static inline int
+lerp(int a, int b, int t)
+{
+	return a + (((b - a) * t) >> 16);
+}
+
+static inline void
+lerpargb(byte *dst, byte *a, byte *b, int t)
+{
+	dst[0] = lerp(a[0], b[0], t);
+	dst[1] = lerp(a[1], b[1], t);
+	dst[2] = lerp(a[2], b[2], t);
+	dst[3] = lerp(a[3], b[3], t);
 }
 
-static inline int samplemask(byte *s, int w, int h, int u, int v)
+static inline int
+samplemask(byte *s, int w, int h, int u, int v)
 {
 	int ui = u >> 16;
 	int vi = v >> 16;
@@ -68,32 +63,8 @@ static inline int samplemask(byte *s, int w, int h, int u, int v)
 	return lerp(ab, cd, vd);
 }
 
-static inline void lerpargb(byte *dst, byte *a, byte *b, int t)
-{
-	dst[0] = lerp(a[0], b[0], t);
-	dst[1] = lerp(a[1], b[1], t);
-	dst[2] = lerp(a[2], b[2], t);
-	dst[3] = lerp(a[3], b[3], t);
-}
-
-static inline byte *getargb(byte *s, int w, int h, int u, int v)
-{
-#ifdef CLAMPUV
-	if (u < 0) u = 0;
-	if (v < 0) v = 0;
-	if (u >= w) u = w - 1;
-	if (v >= h) v = h - 1;
-#else
-	static byte t[4] = { 0, 0, 0, 0 };
-	if (u < 0) return t;
-	if (v < 0) return t;
-	if (u >= w) return t;
-	if (v >= h) return t;
-#endif
-	return s + ((w * v + u) << 2);
-}
-
-static inline void sampleargb(byte *s, int w, int h, int u, int v, byte *abcd)
+static inline void
+sampleargb(byte *s, int w, int h, int u, int v, byte *out)
 {
 	byte ab[4];
 	byte cd[4];
@@ -107,163 +78,101 @@ static inline void sampleargb(byte *s, int w, int h, int u, int v, byte *abcd)
 	byte *d = getargb(s, w, h, ui+1, vi+1);
 	lerpargb(ab, a, b, ud);
 	lerpargb(cd, c, d, ud);
-	lerpargb(abcd, ab, cd, vd);
+	lerpargb(out, ab, cd, vd);
 }
 
-static void img_ncn(FZ_PSRC, int srcn, FZ_PDST, FZ_PCTM)
+static inline void
+samplecolor(byte *s, int w, int h, int n, int u, int v, byte *out)
 {
+	int ui = u >> 16;
+	int vi = v >> 16;
+	int ud = u & 0xFFFF;
+	int vd = v & 0xFFFF;
 	int k;
-	while (h--)
-	{
-		byte *dstp = dst0;
-		int u = u0;
-		int v = v0;
-		int w = w0;
-		while (w--)
-		{
-			for (k = 0; k < srcn; k++)
-			{
-				dstp[k] = samplecomp(src, srcw, srch, u, v, srcn, k);
-				dstp += srcn;
-				u += fa;
-				v += fb;
-			}
-		}
-		dst0 += dstw;
-		u0 += fc;
-		v0 += fd;
-	}
-}
-
-static void img_1c1(FZ_PSRC, FZ_PDST, FZ_PCTM)
-{
-	while (h--)
+	for (k = 0; k < n; k++)
 	{
-		byte *dstp = dst0;
-		int u = u0;
-		int v = v0;
-		int w = w0;
-		while (w--)
-		{
-			dstp[0] = samplemask(src, srcw, srch, u, v);
-			dstp ++;
-			u += fa;
-			v += fb;
-		}
-		dst0 += dstw;
-		u0 += fc;
-		v0 += fd;
+		int a = getcolor(s, w, h, n, ui, vi, k);
+		int b = getcolor(s, w, h, n, ui+1, vi, k);
+		int c = getcolor(s, w, h, n, ui, vi+1, k);
+		int d = getcolor(s, w, h, n, ui+1, vi+1, k);
+		int ab = lerp(a, b, ud);
+		int cd = lerp(c, d, ud);
+		out[k] = lerp(ab, cd, vd);
 	}
 }
 
-static void img_4c4(FZ_PSRC, FZ_PDST, FZ_PCTM)
+static void
+img_1o1(byte * restrict src, byte cov, int len, byte * restrict dst,
+	fz_pixmap *image, int u, int v, int fa, int fb)
 {
-	while (h--)
+	byte *samples = image->samples;
+	int w = image->w;
+	int h = image->h;
+	byte sa;
+	while (len--)
 	{
-		byte *dstp = dst0;
-		int u = u0;
-		int v = v0;
-		int w = w0;
-		while (w--)
-		{
-			sampleargb(src, srcw, srch, u, v, dstp);
-			dstp += 4;
-			u += fa;
-			v += fb;
-		}
-		dst0 += dstw;
-		u0 += fc;
-		v0 += fd;
+		cov += *src; *src = 0; src++;
+		sa = fz_mul255(cov, samplemask(samples, w, h, u, v));
+		dst[0] = sa + fz_mul255(dst[0], 255 - sa);
+		dst++;
+		u += fa;
+		v += fb;
 	}
 }
 
-static void img_1o1(FZ_PSRC, FZ_PDST, FZ_PCTM)
-{
-	byte srca;
-	while (h--)
-	{
-		byte *dstp = dst0;
-		int u = u0;
-		int v = v0;
-		int w = w0;
-		while (w--)
-		{
-			srca = samplemask(src, srcw, srch, u, v);
-			dstp[0] = srca + fz_mul255(dstp[0], 255 - srca);
-			dstp ++;
-			u += fa;
-			v += fb;
-		}
-		dst0 += dstw;
-		u0 += fc;
-		v0 += fd;
-	}
-}
-
-static void img_4o4(FZ_PSRC, FZ_PDST, FZ_PCTM)
+static void
+img_4o4(byte * restrict src, byte cov, int len, byte * restrict dst,
+	fz_pixmap *image, int u, int v, int fa, int fb)
 {
+	byte *samples = image->samples;
+	int w = image->w;
+	int h = image->h;
 	byte argb[4];
-	byte ssa;
-	while (h--)
+	byte sa, ssa;
+	while (len--)
 	{
-		byte *dstp = dst0;
-		int u = u0;
-		int v = v0;
-		int w = w0;
-		while (w--)
-		{
-			sampleargb(src, srcw, srch, u, v, argb);
-			ssa = 255 - argb[0];
-			dstp[0] = argb[0] + fz_mul255(dstp[0], ssa);
-			dstp[1] = argb[1] + fz_mul255(dstp[1], ssa);
-			dstp[2] = argb[2] + fz_mul255(dstp[2], ssa);
-			dstp[3] = argb[3] + fz_mul255(dstp[3], ssa);
-			dstp += 4;
-			u += fa;
-			v += fb;
-		}
-		dst0 += dstw;
-		u0 += fc;
-		v0 += fd;
+		cov += *src; *src = 0; src++;
+		sampleargb(samples, w, h, u, v, argb);
+		sa = fz_mul255(argb[0], cov);
+		ssa = 255 - sa;
+		dst[0] = sa + fz_mul255(dst[0], ssa);
+		dst[1] = fz_mul255(argb[1], sa) + fz_mul255(dst[1], ssa);
+		dst[2] = fz_mul255(argb[2], sa) + fz_mul255(dst[2], ssa);
+		dst[3] = fz_mul255(argb[3], sa) + fz_mul255(dst[3], ssa);
+		dst += 4;
+		u += fa;
+		v += fb;
 	}
 }
 
-static void img_w4i1o4(byte *argb, FZ_PSRC, FZ_PDST, FZ_PCTM)
+static void
+img_w4i1o4(byte *argb, byte * restrict src, byte cov, int len, byte * restrict dst,
+	fz_pixmap *image, int u, int v, int fa, int fb)
 {
+	byte *samples = image->samples;
+	int w = image->w;
+	int h = image->h;
 	byte alpha = argb[0];
 	byte r = argb[1];
 	byte g = argb[2];
 	byte b = argb[3];
-	byte cov;
-	byte ca;
-	while (h--)
+	byte ca, cca;
+	while (len--)
 	{
-		byte *dstp = dst0;
-		int u = u0;
-		int v = v0;
-		int w = w0;
-		while (w--)
-		{
-			cov = samplemask(src, srcw, srch, u, v);
-			ca = fz_mul255(cov, alpha);
-			dstp[0] = ca + fz_mul255(dstp[0], 255 - ca);
-			dstp[1] = fz_mul255((short)r - dstp[1], ca) + dstp[1];
-			dstp[2] = fz_mul255((short)g - dstp[2], ca) + dstp[2];
-			dstp[3] = fz_mul255((short)b - dstp[3], ca) + dstp[3];
-			dstp += 4;
-			u += fa;
-			v += fb;
-		}
-		dst0 += dstw;
-		u0 += fc;
-		v0 += fd;
+		cov += *src; *src = 0; src++;
+		ca = fz_mul255(cov, samplemask(samples, w, h, u, v));
+		ca = fz_mul255(ca, alpha);
+		cca = 255 - ca;
+		dst[0] = ca + fz_mul255(dst[0], cca);
+		dst[1] = fz_mul255(r, ca) + fz_mul255(dst[1], cca);
+		dst[2] = fz_mul255(g, ca) + fz_mul255(dst[2], cca);
+		dst[3] = fz_mul255(b, ca) + fz_mul255(dst[3], cca);
+		dst += 4;
+		u += fa;
+		v += fb;
 	}
 }
 
-void (*fz_img_ncn)(FZ_PSRC, int sn, FZ_PDST, FZ_PCTM) = img_ncn;
-void (*fz_img_1c1)(FZ_PSRC, FZ_PDST, FZ_PCTM) = img_1c1;
-void (*fz_img_4c4)(FZ_PSRC, FZ_PDST, FZ_PCTM) = img_4c4;
-void (*fz_img_1o1)(FZ_PSRC, FZ_PDST, FZ_PCTM) = img_1o1;
-void (*fz_img_4o4)(FZ_PSRC, FZ_PDST, FZ_PCTM) = img_4o4;
-void (*fz_img_w4i1o4)(byte*,FZ_PSRC,FZ_PDST,FZ_PCTM) = img_w4i1o4;
-
+void (*fz_img_1o1)(byte*, byte, int, byte*, fz_pixmap *image, int u, int v, int fa, int fb) = img_1o1;
+void (*fz_img_4o4)(byte*, byte, int, byte*, fz_pixmap *image, int u, int v, int fa, int fb) = img_4o4;
+void (*fz_img_w4i1o4)(byte*, byte*, byte, int, byte*, fz_pixmap *image, int u, int v, int fa, int fb) = img_w4i1o4;
diff --git a/draw/pathscan.c b/draw/pathscan.c
index ebf9502a..b60eddd6 100644
--- a/draw/pathscan.c
+++ b/draw/pathscan.c
@@ -422,7 +422,7 @@ static inline void toalpha(unsigned char *list, int n)
 
 static inline void blit(fz_pixmap *pix, int x, int y,
 	unsigned char *list, int skipx, int len,
-	unsigned char *argb, int over)
+	unsigned char *argb, fz_pixmap *image, fz_matrix *invmat)
 {
 	unsigned char *dst;
 	unsigned char cov;
@@ -437,17 +437,31 @@ static inline void blit(fz_pixmap *pix, int x, int y,
 		++list;
 	}
 
-	if (argb)
-		fz_path_w4i1o4(argb, list, cov, len, dst);
-	else if (over)
-		fz_path_1o1(list, cov, len, dst);
+	if (image)
+	{
+		int u = (invmat->a * (x + 0.5) + invmat->c * (y + 0.5) + invmat->e) * 65536;
+		int v = (invmat->b * (x + 0.5) + invmat->d * (y + 0.5) + invmat->f) * 65536;
+		int fa = invmat->a * 65536;
+		int fb = invmat->b * 65536;
+		if (argb)
+			fz_img_w4i1o4(argb, list, cov, len, dst, image, u, v, fa, fb);
+		else if (image->colorspace)
+			fz_img_4o4(list, cov, len, dst, image, u, v, fa, fb);
+		else
+			fz_img_1o1(list, cov, len, dst, image, u, v, fa, fb);
+	}
 	else
-		fz_path_1c1(list, cov, len, dst);
+	{
+		if (argb)
+			fz_path_w4i1o4(argb, list, cov, len, dst);
+		else
+			fz_path_1o1(list, cov, len, dst);
+	}
 }
 
 fz_error
 fz_scanconvert(fz_gel *gel, fz_ael *ael, int eofill, fz_bbox clip,
-	fz_pixmap *pix, unsigned char *argb, int over)
+	fz_pixmap *pix, unsigned char *argb, fz_pixmap *image, fz_matrix *invmat)
 {
 	fz_error error;
 	unsigned char *deltas;
@@ -483,7 +497,7 @@ fz_scanconvert(fz_gel *gel, fz_ael *ael, int eofill, fz_bbox clip,
 		{
 			if (yd >= clip.y0 && yd < clip.y1)
 			{
-				blit(pix, xmin + skipx, yd, deltas, skipx, clipn, argb, over);
+				blit(pix, xmin + skipx, yd, deltas, skipx, clipn, argb, image, invmat);
 			}
 		}
 		yd = yc;
@@ -512,7 +526,7 @@ fz_scanconvert(fz_gel *gel, fz_ael *ael, int eofill, fz_bbox clip,
 
 	if (yd >= clip.y0 && yd < clip.y1)
 	{
-		blit(pix, xmin + skipx, yd, deltas, skipx, clipn, argb, over);
+		blit(pix, xmin + skipx, yd, deltas, skipx, clipn, argb, image, invmat);
 	}
 
 	fz_free(deltas);
diff --git a/draw/porterduff.c b/draw/porterduff.c
index 64fa842f..1441a358 100644
--- a/draw/porterduff.c
+++ b/draw/porterduff.c
@@ -32,32 +32,6 @@ duff_non(byte * restrict sp0, int sw, int sn, byte * restrict dp0, int dw, int w
 	}
 }
 
-/* dst = src in msk */
-static void
-duff_nimcn(byte * restrict sp0, int sw, int sn, byte * restrict mp0, int mw, int mn, byte * restrict dp0, int dw, int w0, int h)
-{
-	int k;
-	while (h--)
-	{
-		byte *sp = sp0;
-		byte *mp = mp0;
-		byte *dp = dp0;
-		int w = w0;
-		while (w--)
-		{
-			byte ma = mp[0];
-			for (k = 0; k < sn; k++)
-				dp[k] = fz_mul255(sp[k], ma);
-			sp += sn;
-			mp += mn;
-			dp += sn;
-		}
-		sp0 += sw;
-		mp0 += mw;
-		dp0 += dw;
-	}
-}
-
 /* dst = src in msk over dst */
 static void
 duff_nimon(byte * restrict sp0, int sw, int sn, byte * restrict mp0, int mw, int mn, byte * restrict dp0, int dw, int w0, int h)
@@ -89,7 +63,8 @@ duff_nimon(byte * restrict sp0, int sw, int sn, byte * restrict mp0, int mw, int
 	}
 }
 
-static void duff_1o1(byte * restrict sp0, int sw, byte * restrict dp0, int dw, int w0, int h)
+static void
+duff_1o1(byte * restrict sp0, int sw, byte * restrict dp0, int dw, int w0, int h)
 {
 	/* duff_non(sp0, sw, 1, dp0, dw, w0, h); */
 	while (h--)
@@ -108,7 +83,8 @@ static void duff_1o1(byte * restrict sp0, int sw, byte * restrict dp0, int dw, i
 	}
 }
 
-static void duff_4o4(byte *sp0, int sw, byte *dp0, int dw, int w0, int h)
+static void
+duff_4o4(byte *sp0, int sw, byte *dp0, int dw, int w0, int h)
 {
 	/* duff_non(sp0, sw, 4, dp0, dw, w0, h); */
 	while (h--)
@@ -131,55 +107,8 @@ static void duff_4o4(byte *sp0, int sw, byte *dp0, int dw, int w0, int h)
 	}
 }
 
-static void duff_1i1c1(byte * restrict sp0, int sw, byte * restrict mp0, int mw, byte * restrict dp0, int dw, int w0, int h)
-{
-	/* duff_nimcn(sp0, sw, 1, mp0, mw, 1, dp0, dw, w0, h); */
-	while (h--)
-	{
-		byte *sp = sp0;
-		byte *mp = mp0;
-		byte *dp = dp0;
-		int w = w0;
-		while (w--)
-		{
-			dp[0] = fz_mul255(sp[0], mp[0]);
-			sp ++;
-			mp ++;
-			dp ++;
-		}
-		sp0 += sw;
-		mp0 += mw;
-		dp0 += dw;
-	}
-}
-
-static void duff_4i1c4(byte * restrict sp0, int sw, byte * restrict mp0, int mw, byte * restrict dp0, int dw, int w0, int h)
-{
-	/* duff_nimcn(sp0, sw, 4, mp0, mw, 1, dp0, dw, w0, h); */
-	while (h--)
-	{
-		byte *sp = sp0;
-		byte *mp = mp0;
-		byte *dp = dp0;
-		int w = w0;
-		while (w--)
-		{
-			byte ma = mp[0];
-			dp[0] = fz_mul255(sp[0], ma);
-			dp[1] = fz_mul255(sp[1], ma);
-			dp[2] = fz_mul255(sp[2], ma);
-			dp[3] = fz_mul255(sp[3], ma);
-			sp += 4;
-			mp += 1;
-			dp += 4;
-		}
-		sp0 += sw;
-		mp0 += mw;
-		dp0 += dw;
-	}
-}
-
-static void duff_1i1o1(byte * restrict sp0, int sw, byte * restrict mp0, int mw, byte * restrict dp0, int dw, int w0, int h)
+static void
+duff_1i1o1(byte * restrict sp0, int sw, byte * restrict mp0, int mw, byte * restrict dp0, int dw, int w0, int h)
 {
 	/* duff_nimon(sp0, sw, 1, mp0, mw, 1, dp0, dw, w0, h); */
 	while (h--)
@@ -204,7 +133,8 @@ static void duff_1i1o1(byte * restrict sp0, int sw, byte * restrict mp0, int mw,
 	}
 }
 
-static void duff_4i1o4(byte * restrict sp0, int sw, byte * restrict mp0, int mw, byte * restrict dp0, int dw, int w0, int h)
+static void
+duff_4i1o4(byte * restrict sp0, int sw, byte * restrict mp0, int mw, byte * restrict dp0, int dw, int w0, int h)
 {
 	/* duff_nimon(sp0, sw, 4, mp0, mw, 1, dp0, dw, w0, h); */
 	while (h--)
@@ -233,19 +163,11 @@ static void duff_4i1o4(byte * restrict sp0, int sw, byte * restrict mp0, int mw,
 }
 
 /*
- * Path and text masks
+ * Path masks
  */
 
-static void path_1c1(byte * restrict src, byte cov, int len, byte * restrict dst)
-{
-	while (len--)
-	{
-		cov += *src; *src = 0; src++;
-		*dst++ = cov;
-	}
-}
-
-static void path_1o1(byte * restrict src, byte cov, int len, byte * restrict dst)
+static void
+path_1o1(byte * restrict src, byte cov, int len, byte * restrict dst)
 {
 	while (len--)
 	{
@@ -255,8 +177,8 @@ static void path_1o1(byte * restrict src, byte cov, int len, byte * restrict dst
 	}
 }
 
-// With 4 In 1 Over 4
-static void path_w4i1o4(byte * restrict argb, byte * restrict src, byte cov, int len, byte * restrict dst)
+static void
+path_w4i1o4(byte * restrict argb, byte * restrict src, byte cov, int len, byte * restrict dst)
 {
 	byte alpha = argb[0];
 	byte r = argb[1];
@@ -268,7 +190,7 @@ static void path_w4i1o4(byte * restrict argb, byte * restrict src, byte cov, int
 		cov += *src; *src = 0; src++;
 		ca = fz_mul255(cov, alpha);
 		cca = 255 - ca;
-		dst[0] = ca + fz_mul255(dst[0], 255 - ca);
+		dst[0] = ca + fz_mul255(dst[0], cca);
 		dst[1] = fz_mul255(r, ca) + fz_mul255(dst[1], cca);
 		dst[2] = fz_mul255(g, ca) + fz_mul255(dst[2], cca);
 		dst[3] = fz_mul255(b, ca) + fz_mul255(dst[3], cca);
@@ -276,23 +198,12 @@ static void path_w4i1o4(byte * restrict argb, byte * restrict src, byte cov, int
 	}
 }
 
-static void text_1c1(byte * restrict src0, int srcw, byte * restrict dst0, int dstw, int w0, int h)
-{
-	while (h--)
-	{
-		byte * restrict src = src0;
-		byte * restrict dst = dst0;
-		int w = w0;
-		while (w--)
-		{
-			*dst++ = *src++;
-		}
-		src0 += srcw;
-		dst0 += dstw;
-	}
-}
+/*
+ * Text masks
+ */
 
-static void text_1o1(byte * restrict src0, int srcw, byte * restrict dst0, int dstw, int w0, int h)
+static void
+text_1o1(byte * restrict src0, int srcw, byte * restrict dst0, int dstw, int w0, int h)
 {
 	while (h--)
 	{
@@ -310,7 +221,8 @@ static void text_1o1(byte * restrict src0, int srcw, byte * restrict dst0, int d
 	}
 }
 
-static void text_w4i1o4(byte * restrict argb, byte * restrict src0, int srcw, byte * restrict dst0, int dstw, int w0, int h)
+static void
+text_w4i1o4(byte * restrict argb, byte * restrict src0, int srcw, byte * restrict dst0, int dstw, int w0, int h)
 {
 	unsigned char alpha = argb[0];
 	unsigned char r = argb[1];
@@ -325,7 +237,7 @@ static void text_w4i1o4(byte * restrict argb, byte * restrict src0, int srcw, by
 		{
 			byte ca = fz_mul255(src[0], alpha);
 			byte cca = 255 - ca;
-			dst[0] = ca + fz_mul255(dst[0], 255 - ca);
+			dst[0] = ca + fz_mul255(dst[0], cca);
 			dst[1] = fz_mul255(r, ca) + fz_mul255(dst[1], cca);
 			dst[2] = fz_mul255(g, ca) + fz_mul255(dst[2], cca);
 			dst[3] = fz_mul255(b, ca) + fz_mul255(dst[3], cca);
@@ -342,20 +254,15 @@ static void text_w4i1o4(byte * restrict argb, byte * restrict src0, int srcw, by
  */
 
 void (*fz_duff_non)(byte*,int,int,byte*,int,int,int) = duff_non;
-void (*fz_duff_nimcn)(byte*,int,int,byte*,int,int,byte*,int,int,int) = duff_nimcn;
 void (*fz_duff_nimon)(byte*,int,int,byte*,int,int,byte*,int,int,int) = duff_nimon;
 void (*fz_duff_1o1)(byte*,int,byte*,int,int,int) = duff_1o1;
 void (*fz_duff_4o4)(byte*,int,byte*,int,int,int) = duff_4o4;
-void (*fz_duff_1i1c1)(byte*,int,byte*,int,byte*,int,int,int) = duff_1i1c1;
-void (*fz_duff_4i1c4)(byte*,int,byte*,int,byte*,int,int,int) = duff_4i1c4;
 void (*fz_duff_1i1o1)(byte*,int,byte*,int,byte*,int,int,int) = duff_1i1o1;
 void (*fz_duff_4i1o4)(byte*,int,byte*,int,byte*,int,int,int) = duff_4i1o4;
 
-void (*fz_path_1c1)(byte*,byte,int,byte*) = path_1c1;
 void (*fz_path_1o1)(byte*,byte,int,byte*) = path_1o1;
 void (*fz_path_w4i1o4)(byte*,byte*,byte,int,byte*) = path_w4i1o4;
 
-void (*fz_text_1c1)(byte*,int,byte*,int,int,int) = text_1c1;
 void (*fz_text_1o1)(byte*,int,byte*,int,int,int) = text_1o1;
 void (*fz_text_w4i1o4)(byte*,byte*,int,byte*,int,int,int) = text_w4i1o4;
 
diff --git a/fitz/dev_draw.c b/fitz/dev_draw.c
index cc7bdc0c..1b2b4ded 100644
--- a/fitz/dev_draw.c
+++ b/fitz/dev_draw.c
@@ -133,11 +133,11 @@ fz_drawfillpath(void *user, fz_path *path, int evenodd, fz_matrix ctm,
 		argb[1] = rgb[0] * 255;
 		argb[2] = rgb[1] * 255;
 		argb[3] = rgb[2] * 255;
-		fz_scanconvert(dev->gel, dev->ael, evenodd, bbox, dev->dest, argb, 1);
+		fz_scanconvert(dev->gel, dev->ael, evenodd, bbox, dev->dest, argb, nil, nil);
 	}
 	else
 	{
-		fz_scanconvert(dev->gel, dev->ael, evenodd, bbox, dev->dest, nil, 1);
+		fz_scanconvert(dev->gel, dev->ael, evenodd, bbox, dev->dest, nil, nil, nil);
 	}
 }
 
@@ -176,11 +176,11 @@ fz_drawstrokepath(void *user, fz_path *path, fz_strokestate *stroke, fz_matrix c
 		argb[1] = rgb[0] * 255;
 		argb[2] = rgb[1] * 255;
 		argb[3] = rgb[2] * 255;
-		fz_scanconvert(dev->gel, dev->ael, 0, bbox, dev->dest, argb, 1);
+		fz_scanconvert(dev->gel, dev->ael, 0, bbox, dev->dest, argb, nil, nil);
 	}
 	else
 	{
-		fz_scanconvert(dev->gel, dev->ael, 0, bbox, dev->dest, nil, 1);
+		fz_scanconvert(dev->gel, dev->ael, 0, bbox, dev->dest, nil, nil, nil);
 	}
 }
 
@@ -222,7 +222,7 @@ fz_drawclippath(void *user, fz_path *path, int evenodd, fz_matrix ctm)
 	memset(mask->samples, 0, mask->w * mask->h * mask->n);
 	memset(dest->samples, 0, dest->w * dest->h * dest->n);
 
-	fz_scanconvert(dev->gel, dev->ael, evenodd, bbox, mask, nil, 1);
+	fz_scanconvert(dev->gel, dev->ael, evenodd, bbox, mask, nil, nil, nil);
 
 	dev->clipstack[dev->cliptop].scissor = dev->scissor;
 	dev->clipstack[dev->cliptop].mask = mask;
@@ -268,7 +268,7 @@ fz_drawclipstrokepath(void *user, fz_path *path, fz_strokestate *stroke, fz_matr
 	memset(dest->samples, 0, dest->w * dest->h * dest->n);
 
 	if (!fz_isemptyrect(bbox))
-		fz_scanconvert(dev->gel, dev->ael, 0, bbox, mask, nil, 1);
+		fz_scanconvert(dev->gel, dev->ael, 0, bbox, mask, nil, nil, nil);
 
 	dev->clipstack[dev->cliptop].scissor = dev->scissor;
 	dev->clipstack[dev->cliptop].mask = mask;
@@ -515,49 +515,66 @@ fz_drawfillshade(void *user, fz_shade *shade, fz_matrix ctm)
 }
 
 static inline void
-calcimagescale(fz_matrix ctm, int w, int h, int *odx, int *ody)
+calcimagestate(fz_drawdevice *dev, fz_pixmap *image, fz_matrix ctm,
+	fz_bbox *bbox, fz_matrix *invmat, int *dx, int *dy)
 {
-	float sx = w / sqrt(ctm.a * ctm.a + ctm.b * ctm.b);
-	float sy = h / sqrt(ctm.c * ctm.c + ctm.d * ctm.d);
+	float sx, sy;
+	fz_path *path;
+	fz_matrix mat;
+	int w, h;
+
+	sx = image->w / sqrt(ctm.a * ctm.a + ctm.b * ctm.b);
+	sy = image->h / sqrt(ctm.c * ctm.c + ctm.d * ctm.d);
 
 	if (sx < 1.0)
-		*odx = 1;
+		*dx = 1;
 	else
-		*odx = sx;
+		*dx = sx;
 
 	if (sy < 1.0)
-		*ody = 1;
+		*dy = 1;
 	else
-		*ody = sy;
-}
+		*dy = sy;
+
+	w = image->w / *dx;
+	h = image->h / *dy;
+
+	path = fz_newpath();
+	fz_moveto(path, 0.0, 0.0);
+	fz_lineto(path, 1.0, 0.0);
+	fz_lineto(path, 1.0, 1.0);
+	fz_lineto(path, 0.0, 1.0);
+	fz_closepath(path);
 
-#define PDST(p) p->samples + ((y0-p->y) * p->w + (x0-p->x)) * p->n, p->w * p->n
+	fz_resetgel(dev->gel, dev->scissor);
+	fz_fillpath(dev->gel, path, ctm, 1.0);
+	fz_sortgel(dev->gel);
+
+	fz_freepath(path);
+
+	*bbox = fz_boundgel(dev->gel);
+	*bbox = fz_intersectbbox(*bbox, dev->scissor);
+	
+	mat.a =  1.0 / w;
+	mat.b = 0.0;
+	mat.c = 0.0;
+	mat.d =  -1.0 / h;
+	mat.e = 0.0;
+	mat.f = 1.0;
+	*invmat = fz_invertmatrix(fz_concat(mat, ctm));
+	invmat->e -= 0.5;
+	invmat->f -= 0.5;
+}
 
 static void
 fz_drawfillimage(void *user, fz_pixmap *image, fz_matrix ctm)
 {
 	fz_drawdevice *dev = user;
-	fz_rect bounds;
 	fz_bbox bbox;
 	int dx, dy;
 	fz_pixmap *scaled = nil;
 	fz_pixmap *converted = nil;
-	fz_matrix imgmat;
 	fz_matrix invmat;
-	int fa, fb, fc, fd;
-	int u0, v0;
-	int x0, y0;
-	int w, h;
-
-	bounds = fz_transformrect(ctm, fz_unitrect);
-	bbox = fz_roundrect(bounds);
-	bbox = fz_intersectbbox(bbox, dev->scissor);
-
-	if (fz_isemptyrect(bbox))
-		return;
-
-	if (image->w == 0 || image->h == 0)
-		return;
 
 	if (!dev->model)
 	{
@@ -565,7 +582,10 @@ fz_drawfillimage(void *user, fz_pixmap *image, fz_matrix ctm)
 		return;
 	}
 
-	calcimagescale(ctm, image->w, image->h, &dx, &dy);
+	calcimagestate(dev, image, ctm, &bbox, &invmat, &dx, &dy);
+
+	if (fz_isemptyrect(bbox) || image->w == 0 || image->h == 0)
+		return;
 
 	if (dx != 1 || dy != 1)
 	{
@@ -580,34 +600,7 @@ fz_drawfillimage(void *user, fz_pixmap *image, fz_matrix ctm)
 		image = converted;
 	}
 
-	imgmat.a = 1.0 / image->w;
-	imgmat.b = 0.0;
-	imgmat.c = 0.0;
-	imgmat.d = -1.0 / image->h;
-	imgmat.e = 0.0;
-	imgmat.f = 1.0;
-	invmat = fz_invertmatrix(fz_concat(imgmat, ctm));
-
-	invmat.e -= 0.5;
-	invmat.f -= 0.5;
-
-	w = bbox.x1 - bbox.x0;
-	h = bbox.y1 - bbox.y0;
-	x0 = bbox.x0;
-	y0 = bbox.y0;
-	u0 = (invmat.a * (x0+0.5) + invmat.c * (y0+0.5) + invmat.e) * 65536;
-	v0 = (invmat.b * (x0+0.5) + invmat.d * (y0+0.5) + invmat.f) * 65536;
-	fa = invmat.a * 65536;
-	fb = invmat.b * 65536;
-	fc = invmat.c * 65536;
-	fd = invmat.d * 65536;
-
-	if (dev->dest->colorspace)
-		fz_img_4o4(image->samples, image->w, image->h, PDST(dev->dest),
-			u0, v0, fa, fb, fc, fd, w, h);
-	else
-		fz_img_1o1(image->samples, image->w, image->h, PDST(dev->dest),
-			u0, v0, fa, fb, fc, fd, w, h);
+	fz_scanconvert(dev->gel, dev->ael, 0, bbox, dev->dest, nil, image, &invmat);
 
 	if (scaled)
 		fz_droppixmap(scaled);
@@ -620,57 +613,22 @@ fz_drawfillimagemask(void *user, fz_pixmap *image, fz_matrix ctm,
 	fz_colorspace *colorspace, float *color, float alpha)
 {
 	fz_drawdevice *dev = user;
-	fz_rect bounds;
 	fz_bbox bbox;
 	int dx, dy;
 	fz_pixmap *scaled = nil;
-	fz_matrix imgmat;
 	fz_matrix invmat;
-	int fa, fb, fc, fd;
-	int u0, v0;
-	int x0, y0;
-	int w, h;
 
-	bounds = fz_transformrect(ctm, fz_unitrect);
-	bbox = fz_roundrect(bounds);
-	bbox = fz_intersectbbox(bbox, dev->scissor);
+	calcimagestate(dev, image, ctm, &bbox, &invmat, &dx, &dy);
 
-	if (fz_isemptyrect(bbox))
+	if (fz_isemptyrect(bbox) || image->w == 0 || image->h == 0)
 		return;
 
-	if (image->w == 0 || image->h == 0)
-		return;
-
-	calcimagescale(ctm, image->w, image->h, &dx, &dy);
-
 	if (dx != 1 || dy != 1)
 	{
 		scaled = fz_scalepixmap(image, dx, dy);
 		image = scaled;
 	}
 
-	imgmat.a = 1.0 / image->w;
-	imgmat.b = 0.0;
-	imgmat.c = 0.0;
-	imgmat.d = -1.0 / image->h;
-	imgmat.e = 0.0;
-	imgmat.f = 1.0;
-	invmat = fz_invertmatrix(fz_concat(imgmat, ctm));
-
-	invmat.e -= 0.5;
-	invmat.f -= 0.5;
-
-	w = bbox.x1 - bbox.x0;
-	h = bbox.y1 - bbox.y0;
-	x0 = bbox.x0;
-	y0 = bbox.y0;
-	u0 = (invmat.a * (x0+0.5) + invmat.c * (y0+0.5) + invmat.e) * 65536;
-	v0 = (invmat.b * (x0+0.5) + invmat.d * (y0+0.5) + invmat.f) * 65536;
-	fa = invmat.a * 65536;
-	fb = invmat.b * 65536;
-	fc = invmat.c * 65536;
-	fd = invmat.d * 65536;
-
 	if (dev->dest->colorspace)
 	{
 		unsigned char argb[4];
@@ -680,13 +638,11 @@ fz_drawfillimagemask(void *user, fz_pixmap *image, fz_matrix ctm,
 		argb[1] = rgb[0] * 255;
 		argb[2] = rgb[1] * 255;
 		argb[3] = rgb[2] * 255;
-		fz_img_w4i1o4(argb, image->samples, image->w, image->h, PDST(dev->dest),
-			u0, v0, fa, fb, fc, fd, w, h);
+		fz_scanconvert(dev->gel, dev->ael, 0, bbox, dev->dest, argb, image, &invmat);
 	}
 	else
 	{
-		fz_img_1o1(image->samples, image->w, image->h, PDST(dev->dest),
-			u0, v0, fa, fb, fc, fd, w, h);
+		fz_scanconvert(dev->gel, dev->ael, 0, bbox, dev->dest, nil, image, &invmat);
 	}
 
 	if (scaled)
@@ -697,17 +653,11 @@ static void
 fz_drawclipimagemask(void *user, fz_pixmap *image, fz_matrix ctm)
 {
 	fz_drawdevice *dev = user;
-	fz_rect bounds;
 	fz_bbox bbox;
 	fz_pixmap *mask, *dest;
 	int dx, dy;
 	fz_pixmap *scaled = nil;
-	fz_matrix imgmat;
 	fz_matrix invmat;
-	int fa, fb, fc, fd;
-	int u0, v0;
-	int x0, y0;
-	int w, h;
 
 	if (dev->cliptop == MAXCLIP)
 	{
@@ -715,11 +665,9 @@ fz_drawclipimagemask(void *user, fz_pixmap *image, fz_matrix ctm)
 		return;
 	}
 
-	bounds = fz_transformrect(ctm, fz_unitrect);
-	bbox = fz_roundrect(bounds);
-	bbox = fz_intersectbbox(bbox, dev->scissor);
+	calcimagestate(dev, image, ctm, &bbox, &invmat, &dx, &dy);
 
-	if (fz_isemptyrect(bbox) || (image->w == 0 || image->h == 0))
+	if (fz_isemptyrect(bbox) || image->w == 0 || image->h == 0)
 	{
 		dev->clipstack[dev->cliptop].scissor = dev->scissor;
 		dev->clipstack[dev->cliptop].mask = nil;
@@ -729,44 +677,19 @@ fz_drawclipimagemask(void *user, fz_pixmap *image, fz_matrix ctm)
 		return;
 	}
 
-	calcimagescale(ctm, image->w, image->h, &dx, &dy);
-
 	if (dx != 1 || dy != 1)
 	{
 		scaled = fz_scalepixmap(image, dx, dy);
 		image = scaled;
 	}
 
-	imgmat.a = 1.0 / image->w;
-	imgmat.b = 0.0;
-	imgmat.c = 0.0;
-	imgmat.d = -1.0 / image->h;
-	imgmat.e = 0.0;
-	imgmat.f = 1.0;
-	invmat = fz_invertmatrix(fz_concat(imgmat, ctm));
-
-	invmat.e -= 0.5;
-	invmat.f -= 0.5;
-
-	w = bbox.x1 - bbox.x0;
-	h = bbox.y1 - bbox.y0;
-	x0 = bbox.x0;
-	y0 = bbox.y0;
-	u0 = (invmat.a * (x0+0.5) + invmat.c * (y0+0.5) + invmat.e) * 65536;
-	v0 = (invmat.b * (x0+0.5) + invmat.d * (y0+0.5) + invmat.f) * 65536;
-	fa = invmat.a * 65536;
-	fb = invmat.b * 65536;
-	fc = invmat.c * 65536;
-	fd = invmat.d * 65536;
-
 	mask = fz_newpixmapwithrect(nil, bbox);
 	dest = fz_newpixmapwithrect(dev->model, bbox);
 
 	memset(mask->samples, 0, mask->w * mask->h * mask->n);
 	memset(dest->samples, 0, dest->w * dest->h * dest->n);
 
-	fz_img_1o1(image->samples, image->w, image->h, PDST(mask),
-			u0, v0, fa, fb, fc, fd, w, h);
+	fz_scanconvert(dev->gel, dev->ael, 0, bbox, mask, nil, image, &invmat);
 
 	dev->clipstack[dev->cliptop].scissor = dev->scissor;
 	dev->clipstack[dev->cliptop].mask = mask;
diff --git a/fitz/fitz_draw.h b/fitz/fitz_draw.h
index e26a4386..a4aa8677 100644
--- a/fitz/fitz_draw.h
+++ b/fitz/fitz_draw.h
@@ -57,7 +57,7 @@ fz_ael * fz_newael(void);
 void fz_freeael(fz_ael *ael);
 
 fz_error fz_scanconvert(fz_gel *gel, fz_ael *ael, int eofill,
-	fz_bbox clip, fz_pixmap *pix, unsigned char *argb, int over);
+	fz_bbox clip, fz_pixmap *pix, unsigned char *argb, fz_pixmap *image, fz_matrix *invmat);
 
 void fz_fillpath(fz_gel *gel, fz_path *path, fz_matrix ctm, float flatness);
 void fz_strokepath(fz_gel *gel, fz_path *path, fz_strokestate *stroke, fz_matrix ctm, float flatness, float linewidth);
@@ -67,39 +67,25 @@ void fz_dashpath(fz_gel *gel, fz_path *path, fz_strokestate *stroke, fz_matrix c
  * Function pointers -- they can be replaced by cpu-optimized versions
  */
 
-#define FZ_PSRC \
-	unsigned char *src, int srcw, int srch
-#define FZ_PDST \
-	unsigned char *dst0, int dstw
-#define FZ_PCTM \
-	int u0, int v0, int fa, int fb, int fc, int fd, int w0, int h
-
 extern void fz_accelerate(void);
 
 extern void (*fz_duff_non)(unsigned char*,int,int,unsigned char*,int,int,int);
-extern void (*fz_duff_nimcn)(unsigned char*,int,int,unsigned char*,int,int,unsigned char*,int,int,int);
 extern void (*fz_duff_nimon)(unsigned char*,int,int,unsigned char*,int,int,unsigned char*,int,int,int);
 extern void (*fz_duff_1o1)(unsigned char*,int,unsigned char*,int,int,int);
 extern void (*fz_duff_4o4)(unsigned char*,int,unsigned char*,int,int,int);
-extern void (*fz_duff_1i1c1)(unsigned char*,int,unsigned char*,int,unsigned char*,int,int,int);
-extern void (*fz_duff_4i1c4)(unsigned char*,int,unsigned char*,int,unsigned char*,int,int,int);
 extern void (*fz_duff_1i1o1)(unsigned char*,int,unsigned char*,int,unsigned char*,int,int,int);
 extern void (*fz_duff_4i1o4)(unsigned char*,int,unsigned char*,int,unsigned char*,int,int,int);
 
-extern void (*fz_path_1c1)(unsigned char*,unsigned char,int,unsigned char*);
 extern void (*fz_path_1o1)(unsigned char*,unsigned char,int,unsigned char*);
 extern void (*fz_path_w4i1o4)(unsigned char*,unsigned char*,unsigned char,int,unsigned char*);
 
-extern void (*fz_text_1c1)(unsigned char*,int,unsigned char*,int,int,int);
 extern void (*fz_text_1o1)(unsigned char*,int,unsigned char*,int,int,int);
 extern void (*fz_text_w4i1o4)(unsigned char*,unsigned char*,int,unsigned char*,int,int,int);
 
-extern void (*fz_img_ncn)(FZ_PSRC, int sn, FZ_PDST, FZ_PCTM);
-extern void (*fz_img_1c1)(FZ_PSRC, FZ_PDST, FZ_PCTM);
-extern void (*fz_img_4c4)(FZ_PSRC, FZ_PDST, FZ_PCTM);
-extern void (*fz_img_1o1)(FZ_PSRC, FZ_PDST, FZ_PCTM);
-extern void (*fz_img_4o4)(FZ_PSRC, FZ_PDST, FZ_PCTM);
-extern void (*fz_img_w4i1o4)(unsigned char*,FZ_PSRC,FZ_PDST,FZ_PCTM);
+extern void (*fz_img_non)(unsigned char*,unsigned char,int,unsigned char*,fz_pixmap*,fz_matrix*);
+extern void (*fz_img_1o1)(unsigned char*,unsigned char,int,unsigned char*,fz_pixmap*,int u, int v, int fa, int fb);
+extern void (*fz_img_4o4)(unsigned char*,unsigned char,int,unsigned char*,fz_pixmap*,int u, int v, int fa, int fb);
+extern void (*fz_img_w4i1o4)(unsigned char*,unsigned char*,unsigned char,int,unsigned char*,fz_pixmap*,int u, int v, int fa, int fb);
 
 extern void (*fz_decodetile)(fz_pixmap *pix, int skip, float *decode);
 extern void (*fz_loadtile1)(unsigned char*, int sw, unsigned char*, int dw, int w, int h, int pad);
diff --git a/mupdf/pdf_build.c b/mupdf/pdf_build.c
index 9129d660..996e1212 100644
--- a/mupdf/pdf_build.c
+++ b/mupdf/pdf_build.c
@@ -261,7 +261,6 @@ pdf_showshade(pdf_csi *csi, fz_shade *shd)
 void
 pdf_showimage(pdf_csi *csi, pdf_image *image)
 {
-	fz_path *path = nil;
 	pdf_gstate *gstate = csi->gstate + csi->gtop;
 	fz_pixmap *tile = fz_newpixmap(image->cs, 0, 0, image->w, image->h);
 	fz_error error = pdf_loadtile(image, tile);
@@ -272,15 +271,6 @@ pdf_showimage(pdf_csi *csi, pdf_image *image)
 		return;
 	}
 
-	/* TODO: clip against a unit rectangle in the image blit instead */
-	path = fz_newpath();
-	fz_moveto(path, 0.0, 0.0);
-	fz_lineto(path, 1.0, 0.0);
-	fz_lineto(path, 1.0, 1.0);
-	fz_lineto(path, 0.0, 1.0);
-	fz_closepath(path);
-	csi->dev->clippath(csi->dev->user, path, 0, gstate->ctm);
-
 	if (image->mask)
 	{
 		fz_pixmap *mask = fz_newpixmap(NULL, 0, 0, image->mask->w, image->mask->h);
@@ -330,9 +320,6 @@ pdf_showimage(pdf_csi *csi, pdf_image *image)
 	if (image->mask)
 		csi->dev->popclip(csi->dev->user);
 
-	fz_freepath(path);
-	csi->dev->popclip(csi->dev->user);
-
 	fz_droppixmap(tile);
 }