shithub: libvpx

Info  •  Files  •  Log  •  Branches

adjusting the calculation of errorperbit

RDMULT/RDDIV defines a bit worth of distortion in term of sum squared
difference. This has also been used as errorperbit in subpixel motion
search, where the distortions computed as variance of the difference.
The variance of differences is different from sum squared differences
by amount of DC squared. Typically, for inter predicted MBs, this
difference averages around 10% between the two distortion, so this patch
introduces a 110% constant in deriving errorperbit from RDMULT/RDDIV.

Test on CIF set shows small but positive gain on overall PSNR (.03%)
and SSIM (.07%), overall impact on average PSNR is 0.

Change-Id: I95425f922d037b4d96083064a10c7cdd4948ee62

--- a/vp8/encoder/encodeframe.c

+++ b/vp8/encoder/encodeframe.c

@@ -335,7 +335,8 @@

 {

 #if USE_ACT_INDEX

     x->rdmult += *(x->mb_activity_ptr) * (x->rdmult >> 2);

-    x->errorperbit = x->rdmult/x->rddiv;

+    x->errorperbit = x->rdmult * 100 /(110 * x->rddiv);

+    x->errorperbit += (x->errorperbit==0);

 #else

     INT64 a;

     INT64 b;

@@ -346,8 +347,8 @@

     b = (2*act) + cpi->activity_avg;

     x->rdmult = (unsigned int)(((INT64)x->rdmult*b + (a>>1))/a);

-    x->errorperbit = x->rdmult/x->rddiv;

-

+    x->errorperbit = x->rdmult * 100 /(110 * x->rddiv);

+    x->errorperbit += (x->errorperbit==0);

 #endif

     // Activity based Zbin adjustment

--- a/vp8/encoder/rdopt.c

+++ b/vp8/encoder/rdopt.c

@@ -238,7 +238,7 @@

                 (cpi->RDMULT * rd_iifactor[cpi->twopass.next_iiratio]) >> 4;

     }

-    cpi->mb.errorperbit = (cpi->RDMULT / 100);

+    cpi->mb.errorperbit = (cpi->RDMULT / 110);

     cpi->mb.errorperbit += (cpi->mb.errorperbit==0);

     vp8_set_speed_features(cpi);

-- 

⑨