Merge "vp9: force refresh of long term ref when denoiser reset."

7 files changed, 55 insertions, 13 deletions

diff --git a/vp9/encoder/vp9_denoiser.c b/vp9/encoder/vp9_denoiser.c
index b70890e68..2820b71b4 100644
--- a/vp9/encoder/vp9_denoiser.c
+++ b/vp9/encoder/vp9_denoiser.c
@@ -692,6 +692,7 @@ int vp9_denoiser_alloc(VP9_COMMON *cm, struct SVC *svc, VP9_DENOISER *denoiser,
   denoiser->denoising_level = kDenLow;
   denoiser->prev_denoising_level = kDenLow;
   denoiser->reset = 0;
+  denoiser->current_denoiser_frame = 0;
   return 0;
 }
 
@@ -716,13 +717,29 @@ void vp9_denoiser_free(VP9_DENOISER *denoiser) {
   vpx_free_frame_buffer(&denoiser->last_source);
 }
 
-void vp9_denoiser_set_noise_level(VP9_DENOISER *denoiser, int noise_level) {
+static void force_refresh_longterm_ref(VP9_COMP *const cpi) {
+  SVC *const svc = &cpi->svc;
+  // If long term reference is used, force refresh of that slot, so
+  // denoiser buffer for long term reference stays in sync.
+  if (svc->use_gf_temporal_ref_current_layer) {
+    int index = svc->spatial_layer_id;
+    if (svc->number_spatial_layers == 3) index = svc->spatial_layer_id - 1;
+    assert(index >= 0);
+    cpi->alt_fb_idx = svc->buffer_gf_temporal_ref[index].idx;
+    cpi->refresh_alt_ref_frame = 1;
+  }
+}
+
+void vp9_denoiser_set_noise_level(VP9_COMP *const cpi, int noise_level) {
+  VP9_DENOISER *const denoiser = &cpi->denoiser;
   denoiser->denoising_level = noise_level;
   if (denoiser->denoising_level > kDenLowLow &&
-      denoiser->prev_denoising_level == kDenLowLow)
+      denoiser->prev_denoising_level == kDenLowLow) {
     denoiser->reset = 1;
-  else
+    force_refresh_longterm_ref(cpi);
+  } else {
     denoiser->reset = 0;
+  }
   denoiser->prev_denoising_level = denoiser->denoising_level;
 }
 
@@ -754,14 +771,24 @@ int64_t vp9_scale_acskip_thresh(int64_t threshold,
     return threshold;
 }
 
+void vp9_denoiser_reset_on_first_frame(VP9_COMP *const cpi) {
+  if (vp9_denoise_svc_non_key(cpi) &&
+      cpi->denoiser.current_denoiser_frame == 0) {
+    cpi->denoiser.reset = 1;
+    force_refresh_longterm_ref(cpi);
+  }
+}
+
 void vp9_denoiser_update_ref_frame(VP9_COMP *const cpi) {
   VP9_COMMON *const cm = &cpi->common;
   SVC *const svc = &cpi->svc;
+
   if (cpi->oxcf.noise_sensitivity > 0 && denoise_svc(cpi) &&
       cpi->denoiser.denoising_level > kDenLowLow) {
     int svc_refresh_denoiser_buffers = 0;
     int denoise_svc_second_layer = 0;
     FRAME_TYPE frame_type = cm->intra_only ? KEY_FRAME : cm->frame_type;
+    cpi->denoiser.current_denoiser_frame++;
     if (cpi->use_svc) {
       const int svc_buf_shift =
           svc->number_spatial_layers - svc->spatial_layer_id == 2
diff --git a/vp9/encoder/vp9_denoiser.h b/vp9/encoder/vp9_denoiser.h
index 2362c4f50..1973e9898 100644
--- a/vp9/encoder/vp9_denoiser.h
+++ b/vp9/encoder/vp9_denoiser.h
@@ -50,6 +50,7 @@ typedef struct vp9_denoiser {
   int reset;
   int num_ref_frames;
   int num_layers;
+  unsigned int current_denoiser_frame;
   VP9_DENOISER_LEVEL denoising_level;
   VP9_DENOISER_LEVEL prev_denoising_level;
 } VP9_DENOISER;
@@ -111,7 +112,9 @@ static INLINE int total_adj_strong_thresh(BLOCK_SIZE bs,
 
 void vp9_denoiser_free(VP9_DENOISER *denoiser);
 
-void vp9_denoiser_set_noise_level(VP9_DENOISER *denoiser, int noise_level);
+void vp9_denoiser_set_noise_level(struct VP9_COMP *const cpi, int noise_level);
+
+void vp9_denoiser_reset_on_first_frame(struct VP9_COMP *const cpi);
 
 int64_t vp9_scale_part_thresh(int64_t threshold, VP9_DENOISER_LEVEL noise_level,
                               int content_state, int temporal_layer_id);
diff --git a/vp9/encoder/vp9_encoder.c b/vp9/encoder/vp9_encoder.c
index 746f234e2..a73185623 100644
--- a/vp9/encoder/vp9_encoder.c
+++ b/vp9/encoder/vp9_encoder.c
@@ -3815,6 +3815,10 @@ static int encode_without_recode_loop(VP9_COMP *cpi, size_t *size,
            cm->mi_rows * cm->mi_cols * sizeof(*cpi->consec_zero_mv));
   }
 
+#if CONFIG_VP9_TEMPORAL_DENOISING
+  if (cpi->oxcf.noise_sensitivity > 0 && cpi->use_svc)
+    vp9_denoiser_reset_on_first_frame(cpi);
+#endif
   vp9_update_noise_estimate(cpi);
 
   // Scene detection is always used for VBR mode or screen-content case.
diff --git a/vp9/encoder/vp9_noise_estimate.c b/vp9/encoder/vp9_noise_estimate.c
index 8c9a40f55..fc189dbb1 100644
--- a/vp9/encoder/vp9_noise_estimate.c
+++ b/vp9/encoder/vp9_noise_estimate.c
@@ -159,7 +159,7 @@ void vp9_update_noise_estimate(VP9_COMP *const cpi) {
 #if CONFIG_VP9_TEMPORAL_DENOISING
     if (cpi->oxcf.noise_sensitivity > 0 && noise_est_svc(cpi) &&
         cpi->svc.current_superframe > 1) {
-      vp9_denoiser_set_noise_level(&cpi->denoiser, ne->level);
+      vp9_denoiser_set_noise_level(cpi, ne->level);
       copy_frame(&cpi->denoiser.last_source, cpi->Source);
     }
 #endif
@@ -269,7 +269,7 @@ void vp9_update_noise_estimate(VP9_COMP *const cpi) {
         ne->level = vp9_noise_estimate_extract_level(ne);
 #if CONFIG_VP9_TEMPORAL_DENOISING
         if (cpi->oxcf.noise_sensitivity > 0 && noise_est_svc(cpi))
-          vp9_denoiser_set_noise_level(&cpi->denoiser, ne->level);
+          vp9_denoiser_set_noise_level(cpi, ne->level);
 #endif
       }
     }
diff --git a/vp9/encoder/vp9_pickmode.c b/vp9/encoder/vp9_pickmode.c
index 1324b5bc8..fe8f24444 100644
--- a/vp9/encoder/vp9_pickmode.c
+++ b/vp9/encoder/vp9_pickmode.c
@@ -1815,13 +1815,7 @@ void vp9_pick_inter_mode(VP9_COMP *cpi, MACROBLOCK *x, TileDataEnc *tile_data,
 
 #if CONFIG_VP9_TEMPORAL_DENOISING
   if (cpi->oxcf.noise_sensitivity > 0) {
-    if (cpi->use_svc) {
-      int layer =
-          LAYER_IDS_TO_IDX(svc->spatial_layer_id, svc->temporal_layer_id,
-                           svc->number_temporal_layers);
-      LAYER_CONTEXT *lc = &svc->layer_context[layer];
-      denoise_svc_pickmode = denoise_svc(cpi) && !lc->is_key_frame;
-    }
+    if (cpi->use_svc) denoise_svc_pickmode = vp9_denoise_svc_non_key(cpi);
     if (cpi->denoiser.denoising_level > kDenLowLow && denoise_svc_pickmode)
       vp9_denoiser_reset_frame_stats(ctx);
   }
diff --git a/vp9/encoder/vp9_svc_layercontext.c b/vp9/encoder/vp9_svc_layercontext.c
index 21b920f11..510087580 100644
--- a/vp9/encoder/vp9_svc_layercontext.c
+++ b/vp9/encoder/vp9_svc_layercontext.c
@@ -1106,6 +1106,16 @@ void vp9_svc_assert_constraints_pattern(VP9_COMP *const cpi) {
   }
 }
 
+#if CONFIG_VP9_TEMPORAL_DENOISING
+int vp9_denoise_svc_non_key(VP9_COMP *const cpi) {
+  int layer =
+      LAYER_IDS_TO_IDX(cpi->svc.spatial_layer_id, cpi->svc.temporal_layer_id,
+                       cpi->svc.number_temporal_layers);
+  LAYER_CONTEXT *lc = &cpi->svc.layer_context[layer];
+  return denoise_svc(cpi) && !lc->is_key_frame;
+}
+#endif
+
 void vp9_svc_check_spatial_layer_sync(VP9_COMP *const cpi) {
   SVC *const svc = &cpi->svc;
   // Only for superframes whose base is not key, as those are
diff --git a/vp9/encoder/vp9_svc_layercontext.h b/vp9/encoder/vp9_svc_layercontext.h
index 945312044..f1f2457b2 100644
--- a/vp9/encoder/vp9_svc_layercontext.h
+++ b/vp9/encoder/vp9_svc_layercontext.h
@@ -235,6 +235,10 @@ struct lookahead_entry *vp9_svc_lookahead_pop(struct VP9_COMP *const cpi,
 // Start a frame and initialize svc parameters
 int vp9_svc_start_frame(struct VP9_COMP *const cpi);
 
+#if CONFIG_VP9_TEMPORAL_DENOISING
+int vp9_denoise_svc_non_key(struct VP9_COMP *const cpi);
+#endif
+
 void vp9_copy_flags_ref_update_idx(struct VP9_COMP *const cpi);
 
 int vp9_one_pass_cbr_svc_start_layer(struct VP9_COMP *const cpi);