highperfocused/ffmpeg
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

lavfi/nlmeans: move final weighted averaging out of nlmeans_plane

Browse Source 
This helps figuring out where the filter is slow:

  70.53%  ffmpeg_g  ffmpeg_g          [.] nlmeans_slice
  25.73%  ffmpeg_g  ffmpeg_g          [.] compute_safe_ssd_integral_image_c
   1.74%  ffmpeg_g  ffmpeg_g          [.] compute_unsafe_ssd_integral_image
   0.82%  ffmpeg_g  ffmpeg_g          [.] ff_mjpeg_decode_sos
   0.51%  ffmpeg_g  [unknown]         [k] 0xffffffff91800a80
   0.24%  ffmpeg_g  ffmpeg_g          [.] weight_averages

(Tested with a large image that takes several seconds to process)

Since this function is irrelevant speed wise, the file's TODO is
updated.
This commit is contained in:
Clément Bœsch
2018-05-06 13:26:45 +02:00
parent 667503ef68
commit 34e1e53e28
1 changed files with 22 additions and 11 deletions
Download Patch File Download Diff File Expand all files Collapse all files

33
libavfilter/vf_nlmeans.c
View File

@@ -20,7 +20,6 @@
/** /**
* @todo * @todo
* - SIMD for final weighted averaging
* - better automatic defaults? see "Parameters" @ http://www.ipol.im/pub/art/2011/bcm_nlm/ * - better automatic defaults? see "Parameters" @ http://www.ipol.im/pub/art/2011/bcm_nlm/
* - temporal support (probably doesn't need any displacement according to * - temporal support (probably doesn't need any displacement according to
* "Denoising image sequences does not require motion estimation") * "Denoising image sequences does not require motion estimation")
@@ -411,11 +410,30 @@ static int nlmeans_slice(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs
return 0; return 0;
} }
static void weight_averages(uint8_t *dst, ptrdiff_t dst_linesize,
const uint8_t *src, ptrdiff_t src_linesize,
struct weighted_avg *wa, ptrdiff_t wa_linesize,
int w, int h)
{
int x, y;
for (y = 0; y < h; y++) {
for (x = 0; x < w; x++) {
// Also weight the centered pixel
wa[x].total_weight += 1.f;
wa[x].sum += 1.f * src[x];
dst[x] = av_clip_uint8(wa[x].sum / wa[x].total_weight);
}
dst += dst_linesize;
src += src_linesize;
wa += wa_linesize;
}
}
static int nlmeans_plane(AVFilterContext *ctx, int w, int h, int p, int r, static int nlmeans_plane(AVFilterContext *ctx, int w, int h, int p, int r,
uint8_t *dst, ptrdiff_t dst_linesize, uint8_t *dst, ptrdiff_t dst_linesize,
const uint8_t *src, ptrdiff_t src_linesize) const uint8_t *src, ptrdiff_t src_linesize)
{ {
int x, y;
int offx, offy; int offx, offy;
NLMeansContext *s = ctx->priv; NLMeansContext *s = ctx->priv;
/* patches center points cover the whole research window so the patches /* patches center points cover the whole research window so the patches
@@ -448,17 +466,10 @@ static int nlmeans_plane(AVFilterContext *ctx, int w, int h, int p, int r,
} }
} }
} }
for (y = 0; y < h; y++) {
for (x = 0; x < w; x++) {
struct weighted_avg *wa = &s->wa[y*s->wa_linesize + x];
// Also weight the centered pixel weight_averages(dst, dst_linesize, src, src_linesize,
wa->total_weight += 1.f; s->wa, s->wa_linesize, w, h);
wa->sum += 1.f * src[y*src_linesize + x];
dst[y*dst_linesize + x] = av_clip_uint8(wa->sum / wa->total_weight);
}
}
return 0; return 0;
} }