[FFmpeg-cvslog] swresample/rematrix: Use error diffusion to avoid error in the DC component of the matrix

[Michael Niedermayer]

ffmpeg | branch: master | Michael Niedermayer <michael at niedermayer.cc> | Sun May 15 18:33:09 2016 +0200| [7fe81bc4f8ba684626fa08f7bef46da3e8abe373] | committer: Michael Niedermayer

swresample/rematrix: Use error diffusion to avoid error in the DC component of the matrix

This fixes the sum of the integer coefficients ending up summing to a value
larger than the value representing unity.

This issue occurs with qN0.dts when converting to stereo

Signed-off-by: Michael Niedermayer <michael at niedermayer.cc>

> http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=7fe81bc4f8ba684626fa08f7bef46da3e8abe373
---

 libswresample/rematrix.c |   12 +++++++++---
 1 file changed, 9 insertions(+), 3 deletions(-)

diff --git a/libswresample/rematrix.c b/libswresample/rematrix.c
index 932088f..f18c9f6 100644
--- a/libswresample/rematrix.c
+++ b/libswresample/rematrix.c
@@ -371,9 +371,15 @@ av_cold int swri_rematrix_init(SwrContext *s){
         s->native_one    = av_mallocz(sizeof(int));
         if (!s->native_matrix || !s->native_one)
             return AVERROR(ENOMEM);
-        for (i = 0; i < nb_out; i++)
-            for (j = 0; j < nb_in; j++)
-                ((int*)s->native_matrix)[i * nb_in + j] = lrintf(s->matrix[i][j] * 32768);
+        for (i = 0; i < nb_out; i++) {
+            double rem = 0;
+
+            for (j = 0; j < nb_in; j++) {
+                double target = s->matrix[i][j] * 32768 + rem;
+                ((int*)s->native_matrix)[i * nb_in + j] = lrintf(target);
+                rem += target - ((int*)s->native_matrix)[i * nb_in + j];
+            }
+        }
         *((int*)s->native_one) = 32768;
         s->mix_1_1_f = (mix_1_1_func_type*)copy_s16;
         s->mix_2_1_f = (mix_2_1_func_type*)sum2_s16;