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

10 files changed, 210 insertions, 13 deletions

diff --git a/test/svc_datarate_test.cc b/test/svc_datarate_test.cc
index 0c35163bd..0342f92da 100644
--- a/test/svc_datarate_test.cc
+++ b/test/svc_datarate_test.cc
@@ -61,6 +61,8 @@ class DatarateOnePassCbrSvc : public ::svc_test::OnePassCbrSvc {
     force_intra_only_frame_ = 0;
     superframe_has_intra_only_ = 0;
     use_post_encode_drop_ = 0;
+    denoiser_off_on_ = false;
+    denoiser_enable_layers_ = false;
   }
   virtual void BeginPassHook(unsigned int /*pass*/) {}
 
@@ -181,6 +183,46 @@ class DatarateOnePassCbrSvc : public ::svc_test::OnePassCbrSvc {
       }
     }
 
+    if (denoiser_off_on_) {
+      encoder->Control(VP9E_SET_AQ_MODE, 3);
+      // Set inter_layer_pred to INTER_LAYER_PRED_OFF_NONKEY (K-SVC).
+      encoder->Control(VP9E_SET_SVC_INTER_LAYER_PRED, 2);
+      if (!denoiser_enable_layers_) {
+        if (video->frame() == 0)
+          encoder->Control(VP9E_SET_NOISE_SENSITIVITY, 0);
+        else if (video->frame() == 100)
+          encoder->Control(VP9E_SET_NOISE_SENSITIVITY, 1);
+      } else {
+        // Cumulative bitrates for top spatial layers, for
+        // 3 temporal layers.
+        if (video->frame() == 0) {
+          encoder->Control(VP9E_SET_NOISE_SENSITIVITY, 0);
+          // Change layer bitrates to set top spatial layer to 0.
+          // This is for 3 spatial 3 temporal layers.
+          // This will trigger skip encoding/dropping of top spatial layer.
+          cfg_.rc_target_bitrate -= cfg_.layer_target_bitrate[8];
+          for (int i = 0; i < 3; i++)
+            bitrate_sl3_[i] = cfg_.layer_target_bitrate[i + 6];
+          cfg_.layer_target_bitrate[6] = 0;
+          cfg_.layer_target_bitrate[7] = 0;
+          cfg_.layer_target_bitrate[8] = 0;
+          encoder->Config(&cfg_);
+        } else if (video->frame() == 100) {
+          // Change layer bitrates to non-zero on top spatial layer.
+          // This will trigger skip encoding of top spatial layer
+          // on key frame (period = 100).
+          for (int i = 0; i < 3; i++)
+            cfg_.layer_target_bitrate[i + 6] = bitrate_sl3_[i];
+          cfg_.rc_target_bitrate += cfg_.layer_target_bitrate[8];
+          encoder->Config(&cfg_);
+        } else if (video->frame() == 120) {
+          // Enable denoiser and top spatial layer after key frame (period is
+          // 100).
+          encoder->Control(VP9E_SET_NOISE_SENSITIVITY, 1);
+        }
+      }
+    }
+
     if (update_pattern_ && video->frame() >= 100) {
       vpx_svc_layer_id_t layer_id;
       if (video->frame() == 100) {
@@ -488,6 +530,11 @@ class DatarateOnePassCbrSvc : public ::svc_test::OnePassCbrSvc {
   int force_intra_only_frame_;
   int superframe_has_intra_only_;
   int use_post_encode_drop_;
+  int bitrate_sl3_[3];
+  // Denoiser switched on the fly.
+  bool denoiser_off_on_;
+  // Top layer enabled on the fly.
+  bool denoiser_enable_layers_;
 };
 
 // Params: speed setting.
@@ -644,6 +691,112 @@ TEST_P(DatarateOnePassCbrSvcSingleBR, OnePassCbrSvc3SL2TLDynamicPatternChange) {
 #endif
 }
 
+// Check basic rate targeting for 1 pass CBR SVC with 3 spatial and 3 temporal
+// layers, for inter_layer_pred=OffKey (K-SVC) and on the fly switching
+// of denoiser from off to on (on at frame = 100). Key frame period is set to
+// 1000 so denoise is enabled on non-key.
+TEST_P(DatarateOnePassCbrSvcSingleBR,
+       OnePassCbrSvc3SL3TL_DenoiserOffOnFixedLayers) {
+  cfg_.rc_buf_initial_sz = 500;
+  cfg_.rc_buf_optimal_sz = 500;
+  cfg_.rc_buf_sz = 1000;
+  cfg_.rc_min_quantizer = 0;
+  cfg_.rc_max_quantizer = 63;
+  cfg_.rc_end_usage = VPX_CBR;
+  cfg_.g_lag_in_frames = 0;
+  cfg_.ss_number_layers = 3;
+  cfg_.ts_number_layers = 3;
+  cfg_.ts_rate_decimator[0] = 4;
+  cfg_.ts_rate_decimator[1] = 2;
+  cfg_.ts_rate_decimator[2] = 1;
+  cfg_.temporal_layering_mode = 3;
+  cfg_.g_error_resilient = 1;
+  cfg_.g_threads = 1;
+  svc_params_.scaling_factor_num[0] = 72;
+  svc_params_.scaling_factor_den[0] = 288;
+  svc_params_.scaling_factor_num[1] = 144;
+  svc_params_.scaling_factor_den[1] = 288;
+  svc_params_.scaling_factor_num[2] = 288;
+  svc_params_.scaling_factor_den[2] = 288;
+  cfg_.rc_dropframe_thresh = 30;
+  cfg_.kf_max_dist = 1000;
+  number_spatial_layers_ = cfg_.ss_number_layers;
+  number_temporal_layers_ = cfg_.ts_number_layers;
+  ::libvpx_test::I420VideoSource video("desktop_office1.1280_720-020.yuv", 1280,
+                                       720, 30, 1, 0, 300);
+  top_sl_width_ = 1280;
+  top_sl_height_ = 720;
+  cfg_.rc_target_bitrate = 1000;
+  ResetModel();
+  denoiser_off_on_ = true;
+  denoiser_enable_layers_ = false;
+  AssignLayerBitrates();
+  ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
+  // Don't check rate targeting on two top spatial layer since they will be
+  // skipped for part of the sequence.
+  CheckLayerRateTargeting(number_spatial_layers_ - 2, number_temporal_layers_,
+                          0.78, 1.15);
+#if CONFIG_VP9_DECODER
+  // The non-reference frames are expected to be mismatched frames as the
+  // encoder will avoid loopfilter on these frames.
+  EXPECT_EQ(num_nonref_frames_, GetMismatchFrames());
+#endif
+}
+
+// Check basic rate targeting for 1 pass CBR SVC with 3 spatial and 3 temporal
+// layers, for inter_layer_pred=OffKey (K-SVC) and on the fly switching
+// of denoiser from off to on, for dynamic layers. Start at 2 spatial layers
+// and enable 3rd spatial layer at frame = 100. Use periodic key frame with
+// period 100 so enabling of spatial layer occurs at key frame. Enable denoiser
+// at frame > 100, after the key frame sync.
+TEST_P(DatarateOnePassCbrSvcSingleBR,
+       OnePassCbrSvc3SL3TL_DenoiserOffOnEnableLayers) {
+  cfg_.rc_buf_initial_sz = 500;
+  cfg_.rc_buf_optimal_sz = 500;
+  cfg_.rc_buf_sz = 1000;
+  cfg_.rc_min_quantizer = 0;
+  cfg_.rc_max_quantizer = 63;
+  cfg_.rc_end_usage = VPX_CBR;
+  cfg_.g_lag_in_frames = 0;
+  cfg_.ss_number_layers = 3;
+  cfg_.ts_number_layers = 3;
+  cfg_.ts_rate_decimator[0] = 4;
+  cfg_.ts_rate_decimator[1] = 2;
+  cfg_.ts_rate_decimator[2] = 1;
+  cfg_.temporal_layering_mode = 3;
+  cfg_.g_error_resilient = 1;
+  cfg_.g_threads = 1;
+  svc_params_.scaling_factor_num[0] = 72;
+  svc_params_.scaling_factor_den[0] = 288;
+  svc_params_.scaling_factor_num[1] = 144;
+  svc_params_.scaling_factor_den[1] = 288;
+  svc_params_.scaling_factor_num[2] = 288;
+  svc_params_.scaling_factor_den[2] = 288;
+  cfg_.rc_dropframe_thresh = 0;
+  cfg_.kf_max_dist = 100;
+  number_spatial_layers_ = cfg_.ss_number_layers;
+  number_temporal_layers_ = cfg_.ts_number_layers;
+  ::libvpx_test::I420VideoSource video("desktop_office1.1280_720-020.yuv", 1280,
+                                       720, 30, 1, 0, 300);
+  top_sl_width_ = 1280;
+  top_sl_height_ = 720;
+  cfg_.rc_target_bitrate = 1000;
+  ResetModel();
+  denoiser_off_on_ = true;
+  denoiser_enable_layers_ = true;
+  AssignLayerBitrates();
+  ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
+  // Don't check rate targeting on two top spatial layer since they will be
+  // skipped for part of the sequence.
+  CheckLayerRateTargeting(number_spatial_layers_ - 2, number_temporal_layers_,
+                          0.78, 1.15);
+#if CONFIG_VP9_DECODER
+  // The non-reference frames are expected to be mismatched frames as the
+  // encoder will avoid loopfilter on these frames.
+  EXPECT_EQ(num_nonref_frames_, GetMismatchFrames());
+#endif
+}
+
 // Check basic rate targeting for 1 pass CBR SVC with 3 spatial layers and on
 // the fly switching to 1 and then 2 and back to 3 spatial layers. This switch
 // is done by setting spatial layer bitrates to 0, and then back to non-zero,
diff --git a/test/test-data.mk b/test/test-data.mk
index 1d42b94e0..9f8b7e600 100644
--- a/test/test-data.mk
+++ b/test/test-data.mk
@@ -3,6 +3,7 @@ LIBVPX_TEST_SRCS-yes += test-data.mk
 # Encoder test source
 LIBVPX_TEST_DATA-$(CONFIG_ENCODERS) += hantro_collage_w352h288.yuv
 LIBVPX_TEST_DATA-$(CONFIG_ENCODERS) += hantro_odd.yuv
+LIBVPX_TEST_DATA-$(CONFIG_ENCODERS) += desktop_office1.1280_720-020.yuv
 
 LIBVPX_TEST_DATA-$(CONFIG_ENCODERS) += park_joy_90p_10_420.y4m
 LIBVPX_TEST_DATA-$(CONFIG_ENCODERS) += park_joy_90p_10_422.y4m
diff --git a/test/test-data.sha1 b/test/test-data.sha1
index 771aeca01..75b136c06 100644
--- a/test/test-data.sha1
+++ b/test/test-data.sha1
@@ -863,3 +863,4 @@ f1026c03efd5da21b381c8eb21f0d64e6d7e4ba3 *invalid-crbug-1558.ivf
 eb198c25f861c3fe2cbd310de11eb96843019345 *invalid-crbug-1558.ivf.res
 c62b005a9fd32c36a1b3f67de6840330f9915e34 *invalid-crbug-1562.ivf
 f0cd8389948ad16085714d96567612136f6a46c5 *invalid-crbug-1562.ivf.res
+bac455906360b45338a16dd626ac5f19bc36a307 *desktop_office1.1280_720-020.yuv
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);