[FFmpeg-devel] [PATCH 8/9] lavfi/nlmeans: move final weighted averaging out of nlmeans_plane

[Clément Bœsch]

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.
---
 libavfilter/vf_nlmeans.c | 33 ++++++++++++++++++++++-----------
 1 file changed, 22 insertions(+), 11 deletions(-)

diff --git a/libavfilter/vf_nlmeans.c b/libavfilter/vf_nlmeans.c
index 201e4feb41..abe708a2fc 100644
--- a/libavfilter/vf_nlmeans.c
+++ b/libavfilter/vf_nlmeans.c
@@ -20,7 +20,6 @@
 
 /**
  * @todo
- * - SIMD for final weighted averaging
  * - better automatic defaults? see "Parameters" @ http://www.ipol.im/pub/art/2011/bcm_nlm/
  * - temporal support (probably doesn't need any displacement according to
  *   "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;
 }
 
+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 += 1.0;
+            wa[x].sum += 1.0 * 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,
                          uint8_t *dst, ptrdiff_t dst_linesize,
                          const uint8_t *src, ptrdiff_t src_linesize)
 {
-    int x, y;
     int offx, offy;
     NLMeansContext *s = ctx->priv;
     /* 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
-            wa->total_weight += 1.0;
-            wa->sum += 1.0 * src[y*src_linesize + x];
+    weight_averages(dst, dst_linesize, src, src_linesize,
+                    s->wa, s->wa_linesize, w, h);
 
-            dst[y*dst_linesize + x] = av_clip_uint8(wa->sum / wa->total_weight);
-        }
-    }
     return 0;
 }
 
-- 
2.17.0