vp9: Updates to SVC sample encoder.

Allow for passing in the layer bitrates at command line.
Fix to allow passing in bitrate for each spatial-temporal layer.

Change to some default values for 1 pass cbr mode:
spatial scale and qp-max/min.

Small fixes to some build warnings.

Change-Id: I3f9a776262712480a6570bb863a835b2fc49935a

diff --git a/examples/vp9_spatial_svc_encoder.c b/examples/vp9_spatial_svc_encoder.c
index cecdce0804ceb9571f19e7972e63b92c4670fe14..e00fcfbeac3c8b75797fdcfec2f7752888705dd4 100644 (file)
--- a/examples/vp9_spatial_svc_encoder.c
+++ b/examples/vp9_spatial_svc_encoder.c
@@ -84,6 +84,8 @@ static const arg_def_t speed_arg =
     ARG_DEF("sp", "speed", 1, "speed configuration");
 static const arg_def_t aqmode_arg =
     ARG_DEF("aq", "aqmode", 1, "aq-mode off/on");
+static const arg_def_t bitrates_arg =
+    ARG_DEF("bl", "bitrates", 1, "bitrates[sl * num_tl + tl]");
 
 #if CONFIG_VP9_HIGHBITDEPTH
 static const struct arg_enum_list bitdepth_enum[] = {
@@ -124,6 +126,7 @@ static const arg_def_t *svc_args[] = { &frames_arg,
 #endif
                                        &speed_arg,
                                        &rc_end_usage_arg,
+                                       &bitrates_arg,
                                        NULL };
 
 static const uint32_t default_frames_to_skip = 0;
@@ -250,6 +253,9 @@ static void parse_command_line(int argc, const char **argv_,
     } else if (arg_match(&arg, &scale_factors_arg, argi)) {
       snprintf(string_options, sizeof(string_options), "%s scale-factors=%s",
                string_options, arg.val);
+    } else if (arg_match(&arg, &bitrates_arg, argi)) {
+      snprintf(string_options, sizeof(string_options), "%s bitrates=%s",
+               string_options, arg.val);
     } else if (arg_match(&arg, &passes_arg, argi)) {
       passes = arg_parse_uint(&arg);
       if (passes < 1 || passes > 2) {
@@ -620,7 +626,7 @@ int main(int argc, const char **argv) {
   struct RateControlStats rc;
   vpx_svc_layer_id_t layer_id;
   vpx_svc_ref_frame_config_t ref_frame_config;
-  int sl, tl;
+  unsigned int sl, tl;
   double sum_bitrate = 0.0;
   double sum_bitrate2 = 0.0;
   double framerate = 30.0;
@@ -695,6 +701,8 @@ int main(int argc, const char **argv) {
     vpx_codec_control(&codec, VP9E_SET_TILE_COLUMNS, (svc_ctx.threads >> 1));
   if (svc_ctx.speed >= 5 && svc_ctx.aqmode == 1)
     vpx_codec_control(&codec, VP9E_SET_AQ_MODE, 3);
+  if (svc_ctx.speed >= 5)
+    vpx_codec_control(&codec, VP8E_SET_STATIC_THRESHOLD, 1);
 
   // Encode frames
   while (!end_of_stream) {
@@ -730,7 +738,7 @@ int main(int argc, const char **argv) {
                         &ref_frame_config);
       // Keep track of input frames, to account for frame drops in rate control
       // stats/metrics.
-      for (sl = 0; sl < enc_cfg.ss_number_layers; ++sl) {
+      for (sl = 0; sl < (unsigned int)enc_cfg.ss_number_layers; ++sl) {
         ++rc.layer_input_frames[sl * enc_cfg.ts_number_layers +
                                 layer_id.temporal_layer_id];
       }
@@ -793,7 +801,7 @@ int main(int argc, const char **argv) {
                   rc.layer_encoding_bitrate[layer] += 8.0 * sizes[sl];
                   // Keep count of rate control stats per layer, for non-key
                   // frames.
-                  if (tl == layer_id.temporal_layer_id &&
+                  if (tl == (unsigned int)layer_id.temporal_layer_id &&
                       !(cx_pkt->data.frame.flags & VPX_FRAME_IS_KEY)) {
                     rc.layer_avg_frame_size[layer] += 8.0 * sizes[sl];
                     rc.layer_avg_rate_mismatch[layer] +=
@@ -807,7 +815,7 @@ int main(int argc, const char **argv) {
               // Update for short-time encoding bitrate states, for moving
               // window of size rc->window, shifted by rc->window / 2.
               // Ignore first window segment, due to key frame.
-              if (frame_cnt > rc.window_size) {
+              if (frame_cnt > (unsigned int)rc.window_size) {
                 tl = layer_id.temporal_layer_id;
                 for (sl = 0; sl < enc_cfg.ss_number_layers; ++sl) {
                   sum_bitrate += 0.001 * 8.0 * sizes[sl] * framerate;
@@ -823,13 +831,14 @@ int main(int argc, const char **argv) {
               }
 
               // Second shifted window.
-              if (frame_cnt > rc.window_size + rc.window_size / 2) {
+              if (frame_cnt >
+                  (unsigned int)(rc.window_size + rc.window_size / 2)) {
                 tl = layer_id.temporal_layer_id;
                 for (sl = 0; sl < enc_cfg.ss_number_layers; ++sl) {
                   sum_bitrate2 += 0.001 * 8.0 * sizes[sl] * framerate;
                 }
 
-                if (frame_cnt > 2 * rc.window_size &&
+                if (frame_cnt > (unsigned int)(2 * rc.window_size) &&
                     frame_cnt % rc.window_size == 0) {
                   rc.window_count += 1;
                   rc.avg_st_encoding_bitrate += sum_bitrate2 / rc.window_size;
@@ -842,10 +851,11 @@ int main(int argc, const char **argv) {
             }
 #endif
           }
-
+          /*
           printf("SVC frame: %d, kf: %d, size: %d, pts: %d\n", frames_received,
                  !!(cx_pkt->data.frame.flags & VPX_FRAME_IS_KEY),
                  (int)cx_pkt->data.frame.sz, (int)cx_pkt->data.frame.pts);
+          */
           if (enc_cfg.ss_number_layers == 1 && enc_cfg.ts_number_layers == 1)
             si->bytes_sum[0] += (int)cx_pkt->data.frame.sz;
           ++frames_received;

diff --git a/vpx/src/svc_encodeframe.c b/vpx/src/svc_encodeframe.c
index 5aa0b8ddb84122e5a3be2c2230de4ccb21ec86b2..fc0e0f65ebf4a1757b54c44df8df1da16e8d28f4 100644 (file)
--- a/vpx/src/svc_encodeframe.c
+++ b/vpx/src/svc_encodeframe.c
@@ -53,6 +53,10 @@ static const int DEFAULT_SCALE_FACTORS_NUM[VPX_SS_MAX_LAYERS] = { 4, 5, 7, 11,
 static const int DEFAULT_SCALE_FACTORS_DEN[VPX_SS_MAX_LAYERS] = { 16, 16, 16,
                                                                   16, 16 };
 
+static const int DEFAULT_SCALE_FACTORS_NUM_2x[VPX_SS_MAX_LAYERS] = { 1, 2, 4 };
+
+static const int DEFAULT_SCALE_FACTORS_DEN_2x[VPX_SS_MAX_LAYERS] = { 4, 4, 4 };
+
 typedef enum {
   QUANTIZER = 0,
   BITRATE,
@@ -156,6 +160,9 @@ static vpx_codec_err_t parse_layer_options_from_string(SvcContext *svc_ctx,
   char *token;
   const char *delim = ",";
   char *save_ptr;
+  int num_layers = svc_ctx->spatial_layers;
+  if (type == BITRATE)
+    num_layers = svc_ctx->spatial_layers * svc_ctx->temporal_layers;
 
   if (input == NULL || option0 == NULL ||
       (option1 == NULL && type == SCALE_FACTOR))
@@ -163,7 +170,7 @@ static vpx_codec_err_t parse_layer_options_from_string(SvcContext *svc_ctx,
 
   input_string = strdup(input);
   token = strtok_r(input_string, delim, &save_ptr);
-  for (i = 0; i < svc_ctx->spatial_layers; ++i) {
+  for (i = 0; i < num_layers; ++i) {
     if (token != NULL) {
       res = extract_option(type, token, option0 + i, option1 + i);
       if (res != VPX_CODEC_OK) break;
@@ -172,7 +179,7 @@ static vpx_codec_err_t parse_layer_options_from_string(SvcContext *svc_ctx,
       break;
     }
   }
-  if (res == VPX_CODEC_OK && i != svc_ctx->spatial_layers) {
+  if (res == VPX_CODEC_OK && i != num_layers) {
     svc_log(svc_ctx, SVC_LOG_ERROR,
             "svc: layer params type: %d    %d values required, "
             "but only %d specified\n",
@@ -297,7 +304,9 @@ void assign_layer_bitrates(const SvcContext *svc_ctx,
     if (si->bitrates[0] != 0) {
       enc_cfg->rc_target_bitrate = 0;
       for (sl = 0; sl < svc_ctx->spatial_layers; ++sl) {
-        enc_cfg->ss_target_bitrate[sl * svc_ctx->temporal_layers] = 0;
+        enc_cfg->rc_target_bitrate +=
+            si->bitrates[sl * svc_ctx->temporal_layers +
+                         svc_ctx->temporal_layers - 1];
         for (tl = 0; tl < svc_ctx->temporal_layers; ++tl) {
           enc_cfg->ss_target_bitrate[sl * svc_ctx->temporal_layers] +=
               (unsigned int)si->bitrates[sl * svc_ctx->temporal_layers + tl];
@@ -344,6 +353,7 @@ void assign_layer_bitrates(const SvcContext *svc_ctx,
       enc_cfg->rc_target_bitrate = 0;
       for (i = 0; i < svc_ctx->spatial_layers; ++i) {
         enc_cfg->ss_target_bitrate[i] = (unsigned int)si->bitrates[i];
+        enc_cfg->layer_target_bitrate[i] = (unsigned int)si->bitrates[i];
         enc_cfg->rc_target_bitrate += si->bitrates[i];
       }
     } else {
@@ -412,12 +422,24 @@ vpx_codec_err_t vpx_svc_init(SvcContext *svc_ctx, vpx_codec_ctx_t *codec_ctx,
     si->svc_params.scaling_factor_den[sl] = DEFAULT_SCALE_FACTORS_DEN[sl];
     si->svc_params.speed_per_layer[sl] = svc_ctx->speed;
   }
-
+  if (enc_cfg->rc_end_usage == VPX_CBR && enc_cfg->g_pass == VPX_RC_ONE_PASS &&
+      svc_ctx->spatial_layers <= 3) {
+    for (sl = 0; sl < svc_ctx->spatial_layers; ++sl) {
+      int sl2 = (svc_ctx->spatial_layers == 2) ? sl + 1 : sl;
+      si->svc_params.scaling_factor_num[sl] = DEFAULT_SCALE_FACTORS_NUM_2x[sl2];
+      si->svc_params.scaling_factor_den[sl] = DEFAULT_SCALE_FACTORS_DEN_2x[sl2];
+    }
+  }
   for (tl = 0; tl < svc_ctx->temporal_layers; ++tl) {
     for (sl = 0; sl < svc_ctx->spatial_layers; ++sl) {
       i = sl * svc_ctx->temporal_layers + tl;
       si->svc_params.max_quantizers[i] = MAX_QUANTIZER;
       si->svc_params.min_quantizers[i] = 0;
+      if (enc_cfg->rc_end_usage == VPX_CBR &&
+          enc_cfg->g_pass == VPX_RC_ONE_PASS) {
+        si->svc_params.max_quantizers[i] = 56;
+        si->svc_params.min_quantizers[i] = 2;
+      }
     }
   }
 

diff --git a/vpx/svc_context.h b/vpx/svc_context.h
index c8bde5832a5b17d0b6406d4a9ffc99b51f04eb95..462785075cbb9f8c703842b28ad2ba4b89a4aa51 100644 (file)
--- a/vpx/svc_context.h
+++ b/vpx/svc_context.h
@@ -54,7 +54,7 @@ typedef struct SvcInternal {
   // values extracted from option, quantizers
   vpx_svc_extra_cfg_t svc_params;
   int enable_auto_alt_ref[VPX_SS_MAX_LAYERS];
-  int bitrates[VPX_SS_MAX_LAYERS];
+  int bitrates[VPX_MAX_LAYERS];
 
   // accumulated statistics
   double psnr_sum[VPX_SS_MAX_LAYERS][COMPONENTS];  // total/Y/U/V