From: Yaowu Xu <yaowu@google.com>
Date: Fri, 5 Feb 2016 00:15:42 +0000 (-0800)
Subject: Normalize fdct8x8 in psnrhvs computation
X-Git-Url: https://granicus.if.org/sourcecode?a=commitdiff_plain;h=ac898d221fd86b5b34c5f979b26893cb9e898655;p=libvpx

Normalize fdct8x8 in psnrhvs computation

This is to match the scale to the fdct8x8 used in original daala
psnrhvs computation.

Change-Id: Ic30b50747ba9c340bcb679f7439640046c69f90a
---

diff --git a/vpx_dsp/psnrhvs.c b/vpx_dsp/psnrhvs.c
index 300170579..9d1f6ee75 100644
--- a/vpx_dsp/psnrhvs.c
+++ b/vpx_dsp/psnrhvs.c
@@ -26,8 +26,12 @@
 
 static void od_bin_fdct8x8(tran_low_t *y, int ystride, const int16_t *x,
                            int xstride) {
+  int i, j;
   (void) xstride;
   vpx_fdct8x8(x, y, ystride);
+  for (i = 0; i < 8; i++)
+    for (j = 0; j< 8; j++)
+      *(y + ystride*i + j) = (*(y + ystride*i + j) + 4) >> 3;
 }
 
 /* Normalized inverse quantization matrix for 8x8 DCT at the point of