lavfi/delogo: don't recompute the same difference again and again
[ffmpeg.git] / libavfilter / vf_delogo.c
index 10dbc14..3a9bdd4 100644 (file)
@@ -1,6 +1,7 @@
 /*
  * Copyright (c) 2002 Jindrich Makovicka <makovick@gmail.com>
  * Copyright (c) 2011 Stefano Sabatini
+ * Copyright (c) 2013 Jean Delvare <khali@linux-fr.org>
  *
  * This file is part of FFmpeg.
  *
@@ -22,7 +23,8 @@
 /**
  * @file
  * A very simple tv station logo remover
- * Ported from MPlayer libmpcodecs/vf_delogo.c.
+ * Originally imported from MPlayer libmpcodecs/vf_delogo.c,
+ * the algorithm was later improved.
  */
 
 #include "libavutil/common.h"
@@ -35,8 +37,8 @@
 #include "video.h"
 
 /**
- * Apply a simple delogo algorithm to the image in dst and put the
- * result in src.
+ * Apply a simple delogo algorithm to the image in src and put the
+ * result in dst.
  *
  * The algorithm is only applied to the region specified by the logo
  * parameters.
  */
 static void apply_delogo(uint8_t *dst, int dst_linesize,
                          uint8_t *src, int src_linesize,
-                         int w, int h,
+                         int w, int h, AVRational sar,
                          int logo_x, int logo_y, int logo_w, int logo_h,
                          int band, int show, int direct)
 {
-    int x, y;
-    int interp, dist;
+    int x, y, dist;
+    uint64_t interp, weightl, weightr, weightt, weightb;
     uint8_t *xdst, *xsrc;
 
     uint8_t *topleft, *botleft, *topright;
@@ -90,23 +92,30 @@ static void apply_delogo(uint8_t *dst, int dst_linesize,
         for (x = logo_x1+1,
              xdst = dst+logo_x1+1,
              xsrc = src+logo_x1+1; x < logo_x2-1; x++, xdst++, xsrc++) {
+
+            /* Weighted interpolation based on relative distances, taking SAR into account */
+            weightl = (uint64_t)              (logo_x2-1-x) * (y-logo_y1) * (logo_y2-1-y) * sar.den;
+            weightr = (uint64_t)(x-logo_x1)                 * (y-logo_y1) * (logo_y2-1-y) * sar.den;
+            weightt = (uint64_t)(x-logo_x1) * (logo_x2-1-x)               * (logo_y2-1-y) * sar.num;
+            weightb = (uint64_t)(x-logo_x1) * (logo_x2-1-x) * (y-logo_y1)                 * sar.num;
+
             interp =
-                (topleft[src_linesize*(y-logo_y  -yclipt)]   +
-                 topleft[src_linesize*(y-logo_y-1-yclipt)]   +
-                 topleft[src_linesize*(y-logo_y+1-yclipt)])  * (logo_w-(x-logo_x))/logo_w
+                (topleft[src_linesize*(y-logo_y1)]    +
+                 topleft[src_linesize*(y-logo_y1-1)]  +
+                 topleft[src_linesize*(y-logo_y1+1)]) * weightl
                 +
-                (topright[src_linesize*(y-logo_y-yclipt)]    +
-                 topright[src_linesize*(y-logo_y-1-yclipt)]  +
-                 topright[src_linesize*(y-logo_y+1-yclipt)]) * (x-logo_x)/logo_w
+                (topright[src_linesize*(y-logo_y1)]    +
+                 topright[src_linesize*(y-logo_y1-1)]  +
+                 topright[src_linesize*(y-logo_y1+1)]) * weightr
                 +
-                (topleft[x-logo_x-xclipl]    +
-                 topleft[x-logo_x-1-xclipl]  +
-                 topleft[x-logo_x+1-xclipl]) * (logo_h-(y-logo_y))/logo_h
+                (topleft[x-logo_x1]    +
+                 topleft[x-logo_x1-1]  +
+                 topleft[x-logo_x1+1]) * weightt
                 +
-                (botleft[x-logo_x-xclipl]    +
-                 botleft[x-logo_x-1-xclipl]  +
-                 botleft[x-logo_x+1-xclipl]) * (y-logo_y)/logo_h;
-            interp /= 6;
+                (botleft[x-logo_x1]    +
+                 botleft[x-logo_x1-1]  +
+                 botleft[x-logo_x1+1]) * weightb;
+            interp /= (weightl + weightr + weightt + weightb) * 3U;
 
             if (y >= logo_y+band && y < logo_y+logo_h-band &&
                 x >= logo_x+band && x < logo_x+logo_w-band) {
@@ -143,14 +152,14 @@ typedef struct {
 #define FLAGS AV_OPT_FLAG_FILTERING_PARAM|AV_OPT_FLAG_VIDEO_PARAM
 
 static const AVOption delogo_options[]= {
-    {"x",    "set logo x position",       OFFSET(x),    AV_OPT_TYPE_INT, {.i64 = -1}, -1, INT_MAX, FLAGS},
-    {"y",    "set logo y position",       OFFSET(y),    AV_OPT_TYPE_INT, {.i64 = -1}, -1, INT_MAX, FLAGS},
-    {"w",    "set logo width",            OFFSET(w),    AV_OPT_TYPE_INT, {.i64 = -1}, -1, INT_MAX, FLAGS},
-    {"h",    "set logo height",           OFFSET(h),    AV_OPT_TYPE_INT, {.i64 = -1}, -1, INT_MAX, FLAGS},
-    {"band", "set delogo area band size", OFFSET(band), AV_OPT_TYPE_INT, {.i64 =  4}, -1, INT_MAX, FLAGS},
-    {"t",    "set delogo area band size", OFFSET(band), AV_OPT_TYPE_INT, {.i64 =  4}, -1, INT_MAX, FLAGS},
-    {"show", "show delogo area",          OFFSET(show), AV_OPT_TYPE_INT, {.i64 =  0},  0,       1, FLAGS},
-    {NULL},
+    { "x",    "set logo x position",       OFFSET(x),    AV_OPT_TYPE_INT, { .i64 = -1 }, -1, INT_MAX, FLAGS },
+    { "y",    "set logo y position",       OFFSET(y),    AV_OPT_TYPE_INT, { .i64 = -1 }, -1, INT_MAX, FLAGS },
+    { "w",    "set logo width",            OFFSET(w),    AV_OPT_TYPE_INT, { .i64 = -1 }, -1, INT_MAX, FLAGS },
+    { "h",    "set logo height",           OFFSET(h),    AV_OPT_TYPE_INT, { .i64 = -1 }, -1, INT_MAX, FLAGS },
+    { "band", "set delogo area band size", OFFSET(band), AV_OPT_TYPE_INT, { .i64 =  4 }, -1, INT_MAX, FLAGS },
+    { "t",    "set delogo area band size", OFFSET(band), AV_OPT_TYPE_INT, { .i64 =  4 }, -1, INT_MAX, FLAGS },
+    { "show", "show delogo area",          OFFSET(show), AV_OPT_TYPE_INT, { .i64 =  0 },  0, 1,       FLAGS },
+    { NULL },
 };
 
 AVFILTER_DEFINE_CLASS(delogo);
@@ -168,26 +177,13 @@ static int query_formats(AVFilterContext *ctx)
     return 0;
 }
 
-static av_cold int init(AVFilterContext *ctx, const char *args)
+static av_cold int init(AVFilterContext *ctx)
 {
-    DelogoContext *delogo = ctx->priv;
-    int ret = 0;
-
-    delogo->class = &delogo_class;
-    av_opt_set_defaults(delogo);
-
-    if (args)
-        ret = sscanf(args, "%d:%d:%d:%d:%d",
-                     &delogo->x, &delogo->y, &delogo->w, &delogo->h, &delogo->band);
-    if (ret == 5) {
-        if (delogo->band < 0)
-            delogo->show = 1;
-    } else if ((ret = (av_set_options_string(delogo, args, "=", ":"))) < 0)
-        return ret;
+    DelogoContext *s = ctx->priv;
 
 #define CHECK_UNSET_OPT(opt)                                            \
-    if (delogo->opt == -1) {                                            \
-        av_log(delogo, AV_LOG_ERROR, "Option %s was not set.\n", #opt); \
+    if (s->opt == -1) {                                            \
+        av_log(s, AV_LOG_ERROR, "Option %s was not set.\n", #opt); \
         return AVERROR(EINVAL);                                         \
     }
     CHECK_UNSET_OPT(x);
@@ -195,23 +191,25 @@ static av_cold int init(AVFilterContext *ctx, const char *args)
     CHECK_UNSET_OPT(w);
     CHECK_UNSET_OPT(h);
 
-    if (delogo->show)
-        delogo->band = 4;
+    if (s->band < 0 || s->show) {
+        s->show = 1;
+        s->band = 4;
+    }
 
     av_log(ctx, AV_LOG_VERBOSE, "x:%d y:%d, w:%d h:%d band:%d show:%d\n",
-           delogo->x, delogo->y, delogo->w, delogo->h, delogo->band, delogo->show);
+           s->x, s->y, s->w, s->h, s->band, s->show);
 
-    delogo->w += delogo->band*2;
-    delogo->h += delogo->band*2;
-    delogo->x -= delogo->band;
-    delogo->y -= delogo->band;
+    s->w += s->band*2;
+    s->h += s->band*2;
+    s->x -= s->band;
+    s->y -= s->band;
 
     return 0;
 }
 
 static int filter_frame(AVFilterLink *inlink, AVFrame *in)
 {
-    DelogoContext *delogo = inlink->dst->priv;
+    DelogoContext *s = inlink->dst->priv;
     AVFilterLink *outlink = inlink->dst->outputs[0];
     const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(inlink->format);
     AVFrame *out;
@@ -219,6 +217,7 @@ static int filter_frame(AVFilterLink *inlink, AVFrame *in)
     int vsub0 = desc->log2_chroma_h;
     int direct = 0;
     int plane;
+    AVRational sar;
 
     if (av_frame_is_writable(in)) {
         direct = 1;
@@ -233,17 +232,26 @@ static int filter_frame(AVFilterLink *inlink, AVFrame *in)
         av_frame_copy_props(out, in);
     }
 
+    sar = in->sample_aspect_ratio;
+    /* Assume square pixels if SAR is unknown */
+    if (!sar.num)
+        sar.num = sar.den = 1;
+
     for (plane = 0; plane < 4 && in->data[plane]; plane++) {
         int hsub = plane == 1 || plane == 2 ? hsub0 : 0;
         int vsub = plane == 1 || plane == 2 ? vsub0 : 0;
 
         apply_delogo(out->data[plane], out->linesize[plane],
                      in ->data[plane], in ->linesize[plane],
-                     inlink->w>>hsub, inlink->h>>vsub,
-                     delogo->x>>hsub, delogo->y>>vsub,
-                     delogo->w>>hsub, delogo->h>>vsub,
-                     delogo->band>>FFMIN(hsub, vsub),
-                     delogo->show, direct);
+                     FF_CEIL_RSHIFT(inlink->w, hsub),
+                     FF_CEIL_RSHIFT(inlink->h, vsub),
+                     sar, s->x>>hsub, s->y>>vsub,
+                     /* Up and left borders were rounded down, inject lost bits
+                      * into width and height to avoid error accumulation */
+                     FF_CEIL_RSHIFT(s->w + (s->x & ((1<<hsub)-1)), hsub),
+                     FF_CEIL_RSHIFT(s->h + (s->y & ((1<<vsub)-1)), vsub),
+                     s->band>>FFMIN(hsub, vsub),
+                     s->show, direct);
     }
 
     if (!direct)
@@ -274,10 +282,11 @@ AVFilter avfilter_vf_delogo = {
     .name          = "delogo",
     .description   = NULL_IF_CONFIG_SMALL("Remove logo from input video."),
     .priv_size     = sizeof(DelogoContext),
+    .priv_class    = &delogo_class,
     .init          = init,
     .query_formats = query_formats,
 
     .inputs    = avfilter_vf_delogo_inputs,
     .outputs   = avfilter_vf_delogo_outputs,
-    .priv_class = &delogo_class,
+    .flags     = AVFILTER_FLAG_SUPPORT_TIMELINE_GENERIC,
 };