Add experimental unsharp-mask OpenGL scaler. Certainly not yet perfect.

git-svn-id: svn://svn.mplayerhq.hu/mplayer/trunk@25757 b3059339-0415-0410-9bf9-f77b7e298cf2

DOCS/man/en/mplayer.1

@@ -3654,6 +3654,8 @@ Older cards will not be able to handle this for chroma at least in fullscreen mo
 Works on a few more cards than method 1.
 .br
 3: Same as 1 but does not use a lookup texture. Might be faster on some cards.
+.br
+4: Use unsharp masking with 3x3 support and a strength of 0.5. Experimental.
 .RE
 .IPs cscale=<n>
 Select the scaling function to use for chrominance scaling.

libvo/gl_common.c

@@ -756,6 +756,26 @@ static const char *bicub_x_filt_template_RECT =
   "MUL cdelta.xz, parmx.rrgg, {-1, 0, 1, 0};"
   BICUB_X_FILT_MAIN("RECT");
 
+#define UNSHARP_FILT_MAIN(textype) \
+  "TEX a.r, fragment.texcoord[%c], texture[%c], "textype";" \
+  "TEX b.r, coord.xyxy, texture[%c], "textype";" \
+  "TEX b.g, coord.zwzw, texture[%c], "textype";" \
+  "TEX b.b, coord2.xyxy, texture[%c], "textype";" \
+  "TEX b.a, coord2.zwzw, texture[%c], "textype";" \
+  "DP4 b, b, {0.25, 0.25, 0.25, 0.25};" \
+  "SUB b.r, a.r, b.r;" \
+  "MAD yuv.%c, b.r, %s, a.r;"
+
+static const char *unsharp_filt_template_2D =
+  "ADD coord, fragment.texcoord[%c].xyxy, {%f, %f, -%f, -%f};"
+  "ADD coord2, fragment.texcoord[%c].xyxy, {%f, -%f, -%f, %f};"
+  UNSHARP_FILT_MAIN("2D");
+
+static const char *unsharp_filt_template_RECT =
+  "ADD coord, fragment.texcoord[%c].xyxy, {0.5, 0.5, -0.5, -0.5};"
+  "ADD coord2, fragment.texcoord[%c].xyxy, {0.5, -0.5, -0.5, 0.5};" 
+  UNSHARP_FILT_MAIN("RECT");
+
 static const char *yuv_prog_template =
   "PARAM ycoef = {%.4f, %.4f, %.4f};"
   "PARAM ucoef = {%.4f, %.4f, %.4f};"
@@ -812,6 +832,7 @@ static void create_scaler_textures(int scaler, int *texu, char *texs) {
   switch (scaler) {
     case YUV_SCALER_BILIN:
     case YUV_SCALER_BICUB_NOTEX:
+    case YUV_SCALER_UNSHARP:
       break;
     case YUV_SCALER_BICUB:
     case YUV_SCALER_BICUB_X:
@@ -1033,6 +1054,20 @@ static void add_scaler(int scaler, char **prog_pos, int *remain, char *texs,
                  (float)1.0 / texh, (float)1.0 / texh,
                  in_tex, in_tex, in_tex, in_tex, in_tex, in_tex, out_comp);
       break;
+    case YUV_SCALER_UNSHARP:
+      if (rect)
+        snprintf(*prog_pos, *remain, unsharp_filt_template_RECT,
+                 in_tex,
+                 in_tex,
+                 in_tex, in_tex, in_tex, in_tex, in_tex, in_tex, out_comp,
+                 "{0.5}");
+      else
+        snprintf(*prog_pos, *remain, unsharp_filt_template_2D,
+                 in_tex, (float)0.5 / texw, (float)0.5 / texh, (float)0.5 / texw, (float)0.5 / texh,
+                 in_tex, (float)0.5 / texw, (float)0.5 / texh, (float)0.5 / texw, (float)0.5 / texh,
+                 in_tex, in_tex, in_tex, in_tex, in_tex, in_tex, out_comp,
+                 "{0.5}");
+      break;
   }
   *remain -= strlen(*prog_pos);
   *prog_pos += strlen(*prog_pos);

libvo/gl_common.h

@@ -243,6 +243,7 @@ int loadGPUProgram(GLenum target, char *prog);
 #define YUV_SCALER_BICUB_X 2
 //! use cubic scaling without additional lookup texture
 #define YUV_SCALER_BICUB_NOTEX 3
+#define YUV_SCALER_UNSHARP 4
 //! mask for conversion type
 #define YUV_CONVERSION_MASK 0xF
 //! mask for scaler type

libvo/vo_gl.c

@@ -856,6 +856,7 @@ static int preinit(const char *arg)
               "    1: use improved bicubic scaling for luma.\n"
               "    2: use cubic in X, linear in Y direction scaling for luma.\n"
               "    3: as 1 but without using a lookup texture.\n"
+              "    4: experimental unsharp masking.\n"
               "  cscale=<n>\n"
               "    as lscale but for chroma (2x slower with little visible effect).\n"
               "  customprog=<filename>\n"