Do joint motion search iteratively

Allow motion search multiple times iteratively, and break out
the loop if this search couldn't find better motion vectors.
Limit the maximum number of search to 2.

Tests results:
1. stdhd set: 0.311%(overall psnr); 0.346%(ssim).
positive gain on 10 out of 16 clips(best: 2.746% on sunflower;
worst: -0.434% on old_town_cross).
2. derf set: 0.016%(overall psnr); 0.062%(ssim).
positive gain on half of the clips(best: 0.499% on bowing;
worst: -0.387 on city).

Change-Id: Ibf0a51776d4caf7707be0586346db08128117559

diff --git a/vp9/encoder/vp9_rdopt.c b/vp9/encoder/vp9_rdopt.c
index 24b81c0..ef8cb2b 100644 (file)
--- a/vp9/encoder/vp9_rdopt.c
+++ b/vp9/encoder/vp9_rdopt.c
@@ -1863,6 +1863,7 @@ static int64_t handle_inter_mode(VP9_COMP *cpi, MACROBLOCK *x,
         struct buf_2d backup_yv12[MAX_MB_PLANE] = {{0}};
         struct buf_2d backup_second_yv12[MAX_MB_PLANE] = {{0}};
         struct buf_2d scaled_first_yv12;
+        int last_besterr[2] = {INT_MAX, INT_MAX};
 
         if (scaled_ref_frame[0]) {
           int i;
@@ -1897,11 +1898,9 @@ static int64_t handle_inter_mode(VP9_COMP *cpi, MACROBLOCK *x,
         frame_mv[NEWMV][refs[0]].as_int = single_newmv[refs[0]].as_int;
         frame_mv[NEWMV][refs[1]].as_int = single_newmv[refs[1]].as_int;
 
-        // Iteration: joint search is done once for each ref frame.
-        // Tried allowing search multiple times iteratively, and break out if
-        // it couldn't find better mv. But tests didn't show noticeable
-        // improvement.
-        for (ite = 0; ite < 2; ite++) {
+        // Allow joint search multiple times iteratively for each ref frame, and
+        // break out the search loop if it couldn't find better mv.
+        for (ite = 0; ite < 4; ite++) {
           struct buf_2d ref_yv12[2] = {xd->plane[0].pre[0],
                                        xd->plane[0].pre[1]};
           int bestsme = INT_MAX;
@@ -1952,19 +1951,26 @@ static int64_t handle_inter_mode(VP9_COMP *cpi, MACROBLOCK *x,
             int dis; /* TODO: use dis in distortion calculation later. */
             unsigned int sse;
 
-            vp9_find_best_sub_pixel_comp(x, &tmp_mv,
-                                         &ref_mv[id],
-                                         x->errorperbit,
-                                         &cpi->fn_ptr[block_size],
-                                         x->nmvjointcost, x->mvcost,
-                                         &dis, &sse, second_pred,
-                                         b_sz[bsize][0], b_sz[bsize][1]);
+            bestsme = vp9_find_best_sub_pixel_comp(x, &tmp_mv,
+                                                   &ref_mv[id],
+                                                   x->errorperbit,
+                                                   &cpi->fn_ptr[block_size],
+                                                   x->nmvjointcost, x->mvcost,
+                                                   &dis, &sse, second_pred,
+                                                   b_sz[bsize][0],
+                                                   b_sz[bsize][1]);
           }
 
-          frame_mv[NEWMV][refs[id]].as_int =
-              xd->mode_info_context->bmi[0].as_mv[1].as_int = tmp_mv.as_int;
           if (id)
             xd->plane[0].pre[0] = scaled_first_yv12;
+
+          if (bestsme < last_besterr[id]) {
+            frame_mv[NEWMV][refs[id]].as_int =
+                xd->mode_info_context->bmi[0].as_mv[1].as_int = tmp_mv.as_int;
+            last_besterr[id] = bestsme;
+          } else {
+            break;
+          }
         }
 
         // restore the predictor