Use mesh full pixel motion search to build the source ARF

Append mesh search to the diamond shape search to refine
the full pixel motion estimation for source ARF generation.
It improves the average compression performance.

Speed 0
        avg PSNR     overall PSNR     SSIM
mid      -0.18%        -0.18%        -0.22%
hd       -0.25%        -0.23%        -0.36%
nflx2k   -0.22%        -0.23%        -0.37%

Speed 1
       avg PSNR     overall PSNR      SSIM
mid     -0.10%         -0.08%        -0.11%
hd      -0.25%         -0.27%        -0.38%
nflx2k  -0.20%         -0.20%        -0.34%

The additional encoding time is close to the sample noise
range. For bus_cif at 1000 kbps, the speed 0 encoding time
goes from 83.0 s -> 83.6 s.

Change-Id: I48647f50ec3e8f7ae4550a4bde831f569f46ecf3

4 files changed, 35 insertions, 28 deletions

diff --git a/vp9/encoder/vp9_mcomp.c b/vp9/encoder/vp9_mcomp.c
index ba72c0be5..5c19cc039 100644
--- a/vp9/encoder/vp9_mcomp.c
+++ b/vp9/encoder/vp9_mcomp.c
@@ -2186,6 +2186,8 @@ int vp9_full_pixel_search(VP9_COMP *cpi, MACROBLOCK *x, BLOCK_SIZE bsize,
   const SEARCH_METHODS method = (SEARCH_METHODS)search_method;
   vp9_variance_fn_ptr_t *fn_ptr = &cpi->fn_ptr[bsize];
   int var = 0;
+  int run_mesh_search = (method == MESH);
+
   if (cost_list) {
     cost_list[0] = INT_MAX;
     cost_list[1] = INT_MAX;
@@ -2216,29 +2218,32 @@ int vp9_full_pixel_search(VP9_COMP *cpi, MACROBLOCK *x, BLOCK_SIZE bsize,
                           fn_ptr, 1, ref_mv, tmp_mv);
       break;
     default:
-      assert(method == NSTEP);
+      assert(method == NSTEP || method == MESH);
       var = full_pixel_diamond(cpi, x, mvp_full, step_param, error_per_bit,
                                MAX_MVSEARCH_STEPS - 1 - step_param, 1,
                                cost_list, fn_ptr, ref_mv, tmp_mv);
 
-      // Should we allow a follow on exhaustive search?
-      if ((sf->exhaustive_searches_thresh < INT_MAX) &&
-          !cpi->rc.is_src_frame_alt_ref) {
-        int64_t exhuastive_thr = sf->exhaustive_searches_thresh;
-        exhuastive_thr >>=
-            8 - (b_width_log2_lookup[bsize] + b_height_log2_lookup[bsize]);
-
-        // Threshold variance for an exhaustive full search.
-        if (var > exhuastive_thr) {
-          int var_ex;
-          MV tmp_mv_ex;
-          var_ex = full_pixel_exhaustive(cpi, x, tmp_mv, error_per_bit,
-                                         cost_list, fn_ptr, ref_mv, &tmp_mv_ex);
-
-          if (var_ex < var) {
-            var = var_ex;
-            *tmp_mv = tmp_mv_ex;
-          }
+      // Should we allow a follow on exhaustive search, in regular rd search
+      // mode.
+      if (run_mesh_search == 0) {
+        if ((sf->exhaustive_searches_thresh < INT_MAX) &&
+            !cpi->rc.is_src_frame_alt_ref) {
+          int64_t exhuastive_thr = sf->exhaustive_searches_thresh;
+          exhuastive_thr >>=
+              8 - (b_width_log2_lookup[bsize] + b_height_log2_lookup[bsize]);
+          if (var > exhuastive_thr) run_mesh_search = 1;
+        }
+      }
+
+      if (run_mesh_search) {
+        int var_ex;
+        MV tmp_mv_ex;
+        var_ex = full_pixel_exhaustive(cpi, x, tmp_mv, error_per_bit, cost_list,
+                                       fn_ptr, ref_mv, &tmp_mv_ex);
+
+        if (var_ex < var) {
+          var = var_ex;
+          *tmp_mv = tmp_mv_ex;
         }
       }
       break;
diff --git a/vp9/encoder/vp9_speed_features.c b/vp9/encoder/vp9_speed_features.c
index 6d179f7bd..fb3d231e9 100644
--- a/vp9/encoder/vp9_speed_features.c
+++ b/vp9/encoder/vp9_speed_features.c
@@ -201,17 +201,18 @@ static void set_good_speed_feature_framesize_independent(VP9_COMP *cpi,
 
   if (cpi->twopass.fr_content_type == FC_GRAPHICS_ANIMATION) {
     sf->exhaustive_searches_thresh = (1 << 22);
-    for (i = 0; i < MAX_MESH_STEP; ++i) {
-      int mesh_density_level = 0;
-      sf->mesh_patterns[i].range =
-          good_quality_mesh_patterns[mesh_density_level][i].range;
-      sf->mesh_patterns[i].interval =
-          good_quality_mesh_patterns[mesh_density_level][i].interval;
-    }
   } else {
     sf->exhaustive_searches_thresh = INT_MAX;
   }
 
+  for (i = 0; i < MAX_MESH_STEP; ++i) {
+    const int mesh_density_level = 0;
+    sf->mesh_patterns[i].range =
+        good_quality_mesh_patterns[mesh_density_level][i].range;
+    sf->mesh_patterns[i].interval =
+        good_quality_mesh_patterns[mesh_density_level][i].interval;
+  }
+
   if (speed >= 1) {
     sf->enable_tpl_model = 0;
     sf->prune_ref_frame_for_rect_partitions = 0;
diff --git a/vp9/encoder/vp9_speed_features.h b/vp9/encoder/vp9_speed_features.h
index fd4973fb2..406b40755 100644
--- a/vp9/encoder/vp9_speed_features.h
+++ b/vp9/encoder/vp9_speed_features.h
@@ -57,7 +57,8 @@ typedef enum {
   BIGDIA = 3,
   SQUARE = 4,
   FAST_HEX = 5,
-  FAST_DIAMOND = 6
+  FAST_DIAMOND = 6,
+  MESH = 7
 } SEARCH_METHODS;
 
 typedef enum {
diff --git a/vp9/encoder/vp9_temporal_filter.c b/vp9/encoder/vp9_temporal_filter.c
index 9a357c105..99a03edf3 100644
--- a/vp9/encoder/vp9_temporal_filter.c
+++ b/vp9/encoder/vp9_temporal_filter.c
@@ -230,7 +230,7 @@ static uint32_t temporal_filter_find_matching_mb_c(VP9_COMP *cpi,
   MACROBLOCK *const x = &td->mb;
   MACROBLOCKD *const xd = &x->e_mbd;
   MV_SPEED_FEATURES *const mv_sf = &cpi->sf.mv;
-  const SEARCH_METHODS search_method = NSTEP;
+  const SEARCH_METHODS search_method = MESH;
   int step_param;
   int sadpb = x->sadperbit16;
   uint32_t bestsme = UINT_MAX;