Entropy coding for dynamic ref mv modes

This commit enables entropy coding for dynamic reference motion
vector modes. The probability model is contexted on the ranking
categories of the reference motion vector candidates.

Change-Id: I09b58d98a409d63ec1a407331e29f8945b7ef17d

diff --git a/vp10/common/entropymode.c b/vp10/common/entropymode.c
index 5c2d51b9a78ee0fdf3217a01e4cfcf9f56419cec..4d9f773d6d485b890a932b0daddf011b5ca438bc 100644 (file)
--- a/vp10/common/entropymode.c
+++ b/vp10/common/entropymode.c
@@ -184,6 +184,10 @@ static const vpx_prob default_refmv_prob[REFMV_MODE_CONTEXTS] = {
     220, 220, 200, 200, 180, 128, 30, 220, 30,
 };
 
+static const vpx_prob default_drl_prob[DRL_MODE_CONTEXTS] = {
+    128, 128, 128,
+};
+
 #if CONFIG_EXT_INTER
 static const vpx_prob default_new2mv_prob = 180;
 #endif
@@ -1292,6 +1296,8 @@ static void init_mode_probs(FRAME_CONTEXT *fc) {
   vp10_copy(fc->newmv_prob, default_newmv_prob);
   vp10_copy(fc->zeromv_prob, default_zeromv_prob);
   vp10_copy(fc->refmv_prob, default_refmv_prob);
+  vp10_copy(fc->drl_prob0, default_drl_prob);
+  vp10_copy(fc->drl_prob1, default_drl_prob);
 #if CONFIG_EXT_INTER
   fc->new2mv_prob = default_new2mv_prob;
 #endif  // CONFIG_EXT_INTER
@@ -1360,6 +1366,13 @@ void vp10_adapt_inter_frame_probs(VP10_COMMON *cm) {
     fc->refmv_prob[i] = mode_mv_merge_probs(pre_fc->refmv_prob[i],
                                             counts->refmv_mode[i]);
 
+  for (i = 0; i < DRL_MODE_CONTEXTS; ++i)
+    fc->drl_prob0[i] = mode_mv_merge_probs(pre_fc->drl_prob0[i],
+                                           counts->drl_mode0[i]);
+  for (i = 0; i < DRL_MODE_CONTEXTS; ++i)
+    fc->drl_prob1[i] = mode_mv_merge_probs(pre_fc->drl_prob1[i],
+                                           counts->drl_mode1[i]);
+
 #if CONFIG_EXT_INTER
   fc->new2mv_prob = mode_mv_merge_probs(pre_fc->new2mv_prob,
                                         counts->new2mv_mode);

diff --git a/vp10/common/entropymode.h b/vp10/common/entropymode.h
index fbadc4e0712790f21db96f05bb56ced58862f622..4b4bdf12f6742e8e1a3a92894103358573034de3 100644 (file)
--- a/vp10/common/entropymode.h
+++ b/vp10/common/entropymode.h
@@ -68,6 +68,9 @@ typedef struct frame_contexts {
   vpx_prob newmv_prob[NEWMV_MODE_CONTEXTS];
   vpx_prob zeromv_prob[ZEROMV_MODE_CONTEXTS];
   vpx_prob refmv_prob[REFMV_MODE_CONTEXTS];
+  vpx_prob drl_prob0[DRL_MODE_CONTEXTS];
+  vpx_prob drl_prob1[DRL_MODE_CONTEXTS];
+
 #if CONFIG_EXT_INTER
   vpx_prob new2mv_prob;
 #endif  // CONFIG_EXT_INTER
@@ -121,6 +124,8 @@ typedef struct FRAME_COUNTS {
   unsigned int newmv_mode[NEWMV_MODE_CONTEXTS][2];
   unsigned int zeromv_mode[ZEROMV_MODE_CONTEXTS][2];
   unsigned int refmv_mode[REFMV_MODE_CONTEXTS][2];
+  unsigned int drl_mode0[DRL_MODE_CONTEXTS][2];
+  unsigned int drl_mode1[DRL_MODE_CONTEXTS][2];
 #if CONFIG_EXT_INTER
   unsigned int new2mv_mode[2];
 #endif  // CONFIG_EXT_INTER

diff --git a/vp10/common/enums.h b/vp10/common/enums.h
index fdcde99e6b07dfc702650d0a518c2d92f74fbcb3..f41b8d9450954872438c660006589c519ed6982c 100644 (file)
--- a/vp10/common/enums.h
+++ b/vp10/common/enums.h
@@ -231,6 +231,7 @@ typedef enum {
 #define NEWMV_MODE_CONTEXTS  7
 #define ZEROMV_MODE_CONTEXTS 2
 #define REFMV_MODE_CONTEXTS  9
+#define DRL_MODE_CONTEXTS    3
 
 #define ZEROMV_OFFSET 3
 #define REFMV_OFFSET  4
@@ -251,6 +252,7 @@ typedef enum {
 #define MAX_MV_REF_CANDIDATES 2
 #if CONFIG_REF_MV
 #define MAX_REF_MV_STACK_SIZE 16
+#define REF_CAT_LEVEL  160
 #endif
 
 #define INTRA_INTER_CONTEXTS 4

diff --git a/vp10/common/mvref_common.c b/vp10/common/mvref_common.c
index 53394e94d0f054e0f34b40a2e5a4aae31878b8b2..1b7fb7d9fbd535ef547c15f0d950bfb007138a6b 100644 (file)
--- a/vp10/common/mvref_common.c
+++ b/vp10/common/mvref_common.c
@@ -329,6 +329,9 @@ static void setup_ref_mv_list(const VP10_COMMON *cm, const MACROBLOCKD *xd,
 
   nearest_refmv_count = *refmv_count;
 
+  for (idx = 0; idx < nearest_refmv_count; ++idx)
+    ref_mv_stack[idx].weight += REF_CAT_LEVEL;
+
   if (prev_frame_mvs_base && cm->show_frame && cm->last_show_frame
       && rf[1] == NONE) {
     int ref;

diff --git a/vp10/common/mvref_common.h b/vp10/common/mvref_common.h
index 3a77147e955e6b33eba93fca3561f275ceda70e1..224c5edfae870a8f84d4035fcc220d4992e7d5f5 100644 (file)
--- a/vp10/common/mvref_common.h
+++ b/vp10/common/mvref_common.h
@@ -257,6 +257,24 @@ static INLINE int16_t vp10_mode_context_analyzer(
   else
     return mode_context[rf[0]];
 }
+
+static INLINE uint8_t vp10_drl_ctx(const CANDIDATE_MV *ref_mv_stack,
+                                   int ref_idx) {
+  if (ref_mv_stack[ref_idx + 1].weight > REF_CAT_LEVEL &&
+      ref_mv_stack[ref_idx + 2].weight > REF_CAT_LEVEL)
+    return 0;
+
+  if (ref_mv_stack[ref_idx + 1].weight > REF_CAT_LEVEL &&
+      ref_mv_stack[ref_idx + 2].weight < REF_CAT_LEVEL)
+    return 1;
+
+  if (ref_mv_stack[ref_idx + 1].weight < REF_CAT_LEVEL &&
+      ref_mv_stack[ref_idx + 2].weight < REF_CAT_LEVEL)
+    return 2;
+
+  assert(0);
+  return 0;
+}
 #endif
 
 typedef void (*find_mv_refs_sync)(void *const data, int mi_row);

diff --git a/vp10/common/thread_common.c b/vp10/common/thread_common.c
index bacc534591e96df8735c10ff4fdb329252e3bd33..6e959edbd351df7988c7d7051c3a780d52334436 100644 (file)
--- a/vp10/common/thread_common.c
+++ b/vp10/common/thread_common.c
@@ -379,6 +379,15 @@ void vp10_accumulate_frame_counts(VP10_COMMON *cm, FRAME_COUNTS *counts,
     for (j = 0; j < 2; ++j)
       cm->counts.refmv_mode[i][j] += counts->refmv_mode[i][j];
 
+
+  for (i = 0; i < DRL_MODE_CONTEXTS; ++i)
+    for (j = 0; j < 2; ++j)
+      cm->counts.drl_mode0[i][j] += counts->drl_mode0[i][j];
+
+  for (i = 0; i < DRL_MODE_CONTEXTS; ++i)
+    for (j = 0; j < 2; ++j)
+      cm->counts.drl_mode1[i][j] += counts->drl_mode1[i][j];
+
 #if CONFIG_EXT_INTER
   for (j = 0; j < 2; ++j)
     cm->counts.new2mv_mode[j] += counts->new2mv_mode[j];

diff --git a/vp10/decoder/decodeframe.c b/vp10/decoder/decodeframe.c
index a26f969ccb2db1556d5122151a9f7ce4594745c8..ec1a5fb783491ab56adb2d5529d1d2bfffde249d 100644 (file)
--- a/vp10/decoder/decodeframe.c
+++ b/vp10/decoder/decodeframe.c
@@ -132,6 +132,10 @@ static void read_inter_mode_probs(FRAME_CONTEXT *fc, vpx_reader *r) {
     vp10_diff_update_prob(r, &fc->zeromv_prob[i]);
   for (i = 0; i < REFMV_MODE_CONTEXTS; ++i)
     vp10_diff_update_prob(r, &fc->refmv_prob[i]);
+  for (i = 0; i < DRL_MODE_CONTEXTS; ++i)
+    vp10_diff_update_prob(r, &fc->drl_prob0[i]);
+  for (i = 0; i < DRL_MODE_CONTEXTS; ++i)
+    vp10_diff_update_prob(r, &fc->drl_prob1[i]);
 #if CONFIG_EXT_INTER
   vp10_diff_update_prob(r, &fc->new2mv_prob);
 #endif  // CONFIG_EXT_INTER

diff --git a/vp10/decoder/decodemv.c b/vp10/decoder/decodemv.c
index 8d26c62de697c5704f4c9afc4d6afe68082a1dc5..51bf2b390f0b043923ffca1ae39fc546a186a471 100644 (file)
--- a/vp10/decoder/decodemv.c
+++ b/vp10/decoder/decodemv.c
@@ -147,6 +147,45 @@ static PREDICTION_MODE read_inter_mode(VP10_COMMON *cm, MACROBLOCKD *xd,
 #endif
 }
 
+#if CONFIG_REF_MV
+static void read_drl_idx(const VP10_COMMON *cm,
+                         MACROBLOCKD *xd,
+                         MB_MODE_INFO *mbmi,
+                         vpx_reader *r) {
+  uint8_t ref_frame_type = vp10_ref_frame_type(mbmi->ref_frame);
+  mbmi->ref_mv_idx = 0;
+
+  if (xd->ref_mv_count[ref_frame_type] > 2) {
+    uint8_t drl0_ctx = vp10_drl_ctx(xd->ref_mv_stack[ref_frame_type], 0);
+    vpx_prob drl0_prob = cm->fc->drl_prob0[drl0_ctx];
+    if (vpx_read(r, drl0_prob)) {
+      mbmi->ref_mv_idx = 1;
+      if (xd->counts)
+        ++xd->counts->drl_mode0[drl0_ctx][1];
+      if (xd->ref_mv_count[ref_frame_type] > 3) {
+        uint8_t drl1_ctx =
+            vp10_drl_ctx(xd->ref_mv_stack[ref_frame_type], 1);
+        vpx_prob drl1_prob = cm->fc->drl_prob1[drl1_ctx];
+        if (vpx_read(r, drl1_prob)) {
+          mbmi->ref_mv_idx = 2;
+          if (xd->counts)
+            ++xd->counts->drl_mode1[drl1_ctx][1];
+
+          return;
+        }
+
+        if (xd->counts)
+          ++xd->counts->drl_mode1[drl1_ctx][0];
+      }
+      return;
+    }
+
+    if (xd->counts)
+      ++xd->counts->drl_mode0[drl0_ctx][0];
+  }
+}
+#endif
+
 #if CONFIG_EXT_INTER
 static PREDICTION_MODE read_inter_compound_mode(VP10_COMMON *cm,
                                                 MACROBLOCKD *xd,
@@ -1056,17 +1095,8 @@ static void read_inter_block_mode_info(VP10Decoder *const pbi,
 #endif  // CONFIG_REF_MV && CONFIG_EXT_INTER
                                    r, mode_ctx);
 #if CONFIG_REF_MV
-      if (mbmi->mode == NEARMV) {
-        uint8_t ref_frame_type = vp10_ref_frame_type(mbmi->ref_frame);
-        if (xd->ref_mv_count[ref_frame_type] > 2) {
-          if (vpx_read_bit(r)) {
-            mbmi->ref_mv_idx = 1;
-            if (xd->ref_mv_count[ref_frame_type] > 3)
-              if (vpx_read_bit(r))
-                mbmi->ref_mv_idx = 2;
-          }
-        }
-      }
+      if (mbmi->mode == NEARMV)
+        read_drl_idx(cm, xd, mbmi, r);
 #endif
     }
   }

diff --git a/vp10/encoder/bitstream.c b/vp10/encoder/bitstream.c
index d4bd12269c9e3d27e2c365649e93e4e7ab0ebcec..a2edade8249b22d576019b9c0e6692493362937b 100644 (file)
--- a/vp10/encoder/bitstream.c
+++ b/vp10/encoder/bitstream.c
@@ -177,6 +177,28 @@ static void write_inter_mode(VP10_COMMON *cm,
 #endif
 }
 
+#if CONFIG_REF_MV
+static void write_drl_idx(const VP10_COMMON *cm,
+                          const MB_MODE_INFO *mbmi,
+                          const MB_MODE_INFO_EXT *mbmi_ext,
+                          vpx_writer *w) {
+  uint8_t ref_frame_type = vp10_ref_frame_type(mbmi->ref_frame);
+  if (mbmi_ext->ref_mv_count[ref_frame_type] > 2) {
+    uint8_t drl0_ctx =
+        vp10_drl_ctx(mbmi_ext->ref_mv_stack[ref_frame_type], 0);
+    vpx_prob drl0_prob = cm->fc->drl_prob0[drl0_ctx];
+    vpx_write(w, mbmi->ref_mv_idx != 0, drl0_prob);
+    if (mbmi_ext->ref_mv_count[ref_frame_type] > 3 &&
+        mbmi->ref_mv_idx > 0) {
+      uint8_t drl1_ctx =
+          vp10_drl_ctx(mbmi_ext->ref_mv_stack[ref_frame_type], 1);
+      vpx_prob drl1_prob = cm->fc->drl_prob1[drl1_ctx];
+      vpx_write(w, mbmi->ref_mv_idx != 1, drl1_prob);
+    }
+  }
+}
+#endif
+
 #if CONFIG_EXT_INTER
 static void write_inter_compound_mode(VP10_COMMON *cm, vpx_writer *w,
                                       PREDICTION_MODE mode,
@@ -312,7 +334,12 @@ static void update_inter_mode_probs(VP10_COMMON *cm, vpx_writer *w,
   for (i = 0; i < REFMV_MODE_CONTEXTS; ++i)
     vp10_cond_prob_diff_update(w, &cm->fc->refmv_prob[i],
                                counts->refmv_mode[i]);
-
+  for (i = 0; i < DRL_MODE_CONTEXTS; ++i)
+    vp10_cond_prob_diff_update(w, &cm->fc->drl_prob0[i],
+                               counts->drl_mode0[i]);
+  for (i = 0; i < DRL_MODE_CONTEXTS; ++i)
+    vp10_cond_prob_diff_update(w, &cm->fc->drl_prob1[i],
+                               counts->drl_mode1[i]);
 #if CONFIG_EXT_INTER
   vp10_cond_prob_diff_update(w, &cm->fc->new2mv_prob, counts->new2mv_mode);
 #endif  // CONFIG_EXT_INTER
@@ -994,15 +1021,8 @@ static void pack_inter_mode_mvs(VP10_COMP *cpi, const MODE_INFO *mi,
                          mode_ctx);
 
 #if CONFIG_REF_MV
-        if (mode == NEARMV) {
-          uint8_t ref_frame_type = vp10_ref_frame_type(mbmi->ref_frame);
-          if (mbmi_ext->ref_mv_count[ref_frame_type] > 2) {
-            vpx_write_bit(w, mbmi->ref_mv_idx != 0);
-            if (mbmi_ext->ref_mv_count[ref_frame_type] > 3 &&
-                mbmi->ref_mv_idx > 0)
-              vpx_write_bit(w, mbmi->ref_mv_idx != 1);
-          }
-        }
+        if (mode == NEARMV)
+          write_drl_idx(cm, mbmi, mbmi_ext, w);
 #endif
       }
     }

diff --git a/vp10/encoder/encodeframe.c b/vp10/encoder/encodeframe.c
index 859ec49c40ffd49998fbfa4a6e5949cd4457dcd9..b758b3bd87b98ed16c805ff72e2f79304b24df7c 100644 (file)
--- a/vp10/encoder/encodeframe.c
+++ b/vp10/encoder/encodeframe.c
@@ -1822,6 +1822,28 @@ static void update_stats(VP10_COMMON *cm, ThreadData *td
                                 has_second_ref(mbmi),
 #endif  // CONFIG_EXT_INTER
                                 mode_ctx);
+
+        if (mode == NEARMV) {
+          uint8_t ref_frame_type = vp10_ref_frame_type(mbmi->ref_frame);
+          if (mbmi_ext->ref_mv_count[ref_frame_type] > 2) {
+            uint8_t drl0_ctx =
+                vp10_drl_ctx(mbmi_ext->ref_mv_stack[ref_frame_type], 0);
+            if (mbmi->ref_mv_idx == 0)
+              ++counts->drl_mode0[drl0_ctx][0];
+            else
+              ++counts->drl_mode0[drl0_ctx][1];
+
+            if (mbmi_ext->ref_mv_count[ref_frame_type] > 3 &&
+                mbmi->ref_mv_idx > 0) {
+              uint8_t drl1_ctx =
+                  vp10_drl_ctx(mbmi_ext->ref_mv_stack[ref_frame_type], 1);
+              if (mbmi->ref_mv_idx == 1)
+                ++counts->drl_mode1[drl1_ctx][0];
+              else
+                ++counts->drl_mode1[drl1_ctx][1];
+            }
+          }
+        }
 #if CONFIG_EXT_INTER
         }
 #endif  // CONFIG_EXT_INTER

diff --git a/vp10/encoder/encoder.h b/vp10/encoder/encoder.h
index 940c9a51acd737ea96924f7b70eb0aa5956af100..f7c60133973d7074e2c67c53f82b92622d7cafeb 100644 (file)
--- a/vp10/encoder/encoder.h
+++ b/vp10/encoder/encoder.h
@@ -474,6 +474,8 @@ typedef struct VP10_COMP {
   int newmv_mode_cost[NEWMV_MODE_CONTEXTS][2];
   int zeromv_mode_cost[ZEROMV_MODE_CONTEXTS][2];
   int refmv_mode_cost[REFMV_MODE_CONTEXTS][2];
+  int drl_mode_cost0[DRL_MODE_CONTEXTS][2];
+  int drl_mode_cost1[DRL_MODE_CONTEXTS][2];
 #if CONFIG_EXT_INTER
   int new2mv_mode_cost[2];
 #endif  // CONFIG_EXT_INTER

diff --git a/vp10/encoder/rd.c b/vp10/encoder/rd.c
index 1aac71908bc7a27bbc4b843f05c5d73d73de47a5..8498ce9cc70179333ec86b99d0358ace4092d129 100644 (file)
--- a/vp10/encoder/rd.c
+++ b/vp10/encoder/rd.c
@@ -382,6 +382,16 @@ void vp10_initialize_rd_consts(VP10_COMP *cpi) {
       cpi->refmv_mode_cost[i][0] = vp10_cost_bit(cm->fc->refmv_prob[i], 0);
       cpi->refmv_mode_cost[i][1] = vp10_cost_bit(cm->fc->refmv_prob[i], 1);
     }
+
+    for (i = 0; i < DRL_MODE_CONTEXTS; ++i) {
+      cpi->drl_mode_cost0[i][0] = vp10_cost_bit(cm->fc->drl_prob0[i], 0);
+      cpi->drl_mode_cost0[i][1] = vp10_cost_bit(cm->fc->drl_prob0[i], 1);
+    }
+
+    for (i = 0; i < DRL_MODE_CONTEXTS; ++i) {
+      cpi->drl_mode_cost1[i][0] = vp10_cost_bit(cm->fc->drl_prob1[i], 0);
+      cpi->drl_mode_cost1[i][1] = vp10_cost_bit(cm->fc->drl_prob1[i], 1);
+    }
 #if CONFIG_EXT_INTER
     cpi->new2mv_mode_cost[0] = vp10_cost_bit(cm->fc->new2mv_prob, 0);
     cpi->new2mv_mode_cost[1] = vp10_cost_bit(cm->fc->new2mv_prob, 1);

diff --git a/vp10/encoder/rdopt.c b/vp10/encoder/rdopt.c
index af360242f3f2bdce0b9b65ef6e34839934e6cb9c..3f617f81ad6fe60b0a6992e7e4404858c2097dbd 100644 (file)
--- a/vp10/encoder/rdopt.c
+++ b/vp10/encoder/rdopt.c
@@ -6380,7 +6380,9 @@ void vp10_rd_pick_inter_mode_sb(VP10_COMP *cpi,
         int ref_idx;
         int ref_set = VPXMIN(2, mbmi_ext->ref_mv_count[ref_frame_type] - 2);
 
-        rate2 += vp10_cost_bit(128, 0);
+        uint8_t drl0_ctx =
+            vp10_drl_ctx(mbmi_ext->ref_mv_stack[ref_frame_type], 0);
+        rate2 += cpi->drl_mode_cost0[drl0_ctx][0];
 
         if (this_rd < INT64_MAX) {
           if (RDCOST(x->rdmult, x->rddiv, rate_y + rate_uv, distortion2) <
@@ -6440,9 +6442,13 @@ void vp10_rd_pick_inter_mode_sb(VP10_COMP *cpi,
                                            dummy_filter_cache);
           }
 
-          tmp_rate += vp10_cost_bit(128, 1);
-          if (mbmi_ext->ref_mv_count[ref_frame_type] > 3)
-            tmp_rate += vp10_cost_bit(128, ref_idx);
+          tmp_rate += cpi->drl_mode_cost0[drl0_ctx][1];
+
+          if (mbmi_ext->ref_mv_count[ref_frame_type] > 3) {
+            uint8_t drl1_ctx =
+                vp10_drl_ctx(mbmi_ext->ref_mv_stack[ref_frame_type], 1);
+            tmp_rate += cpi->drl_mode_cost1[drl1_ctx][ref_idx];
+          }
 
           if (tmp_alt_rd < INT64_MAX) {
             if (RDCOST(x->rdmult, x->rddiv,