add y4m support to vp9 example encoders.

vp9_spatial_svc_encoder and vpx_temporal_svc_encoder.

Change-Id: I8dfa1dfad83c83a26ddac4e7c57b5f1ff161e588

3 files changed, 50 insertions, 19 deletions

diff --git a/examples/vp8_multi_resolution_encoder.c b/examples/vp8_multi_resolution_encoder.c
index b14b1ff39..e72f8a019 100644
--- a/examples/vp8_multi_resolution_encoder.c
+++ b/examples/vp8_multi_resolution_encoder.c
@@ -61,7 +61,7 @@ void usage_exit(void) { exit(EXIT_FAILURE); }
 
 int (*read_frame_p)(FILE *f, vpx_image_t *img);
 
-static int read_frame(FILE *f, vpx_image_t *img) {
+static int mulres_read_frame(FILE *f, vpx_image_t *img) {
   size_t nbytes, to_read;
   int res = 1;
 
@@ -75,7 +75,7 @@ static int read_frame(FILE *f, vpx_image_t *img) {
   return res;
 }
 
-static int read_frame_by_row(FILE *f, vpx_image_t *img) {
+static int mulres_read_frame_by_row(FILE *f, vpx_image_t *img) {
   size_t nbytes, to_read;
   int res = 1;
   int plane;
@@ -471,9 +471,9 @@ int main(int argc, char **argv) {
       die("Failed to allocate image", cfg[i].g_w, cfg[i].g_h);
 
   if (raw[0].stride[VPX_PLANE_Y] == (int)raw[0].d_w)
-    read_frame_p = read_frame;
+    read_frame_p = mulres_read_frame;
   else
-    read_frame_p = read_frame_by_row;
+    read_frame_p = mulres_read_frame_by_row;
 
   for (i = 0; i < NUM_ENCODERS; i++)
     if (outfile[i]) write_ivf_file_header(outfile[i], &cfg[i], 0);
diff --git a/examples/vp9_spatial_svc_encoder.c b/examples/vp9_spatial_svc_encoder.c
index f8093e1bf..92b310684 100644
--- a/examples/vp9_spatial_svc_encoder.c
+++ b/examples/vp9_spatial_svc_encoder.c
@@ -30,6 +30,8 @@
 #include "vpx/vpx_encoder.h"
 #include "../vpxstats.h"
 #include "vp9/encoder/vp9_encoder.h"
+#include "./y4minput.h"
+
 #define OUTPUT_RC_STATS 1
 
 static const arg_def_t skip_frames_arg =
@@ -161,7 +163,6 @@ static const int32_t default_speed = -1;    // -1 means use library default.
 static const uint32_t default_threads = 0;  // zero means use library default.
 
 typedef struct {
-  const char *input_filename;
   const char *output_filename;
   uint32_t frames_to_code;
   uint32_t frames_to_skip;
@@ -393,10 +394,16 @@ static void parse_command_line(int argc, const char **argv_,
   if (argv[0] == NULL || argv[1] == 0) {
     usage_exit();
   }
-  app_input->input_filename = argv[0];
+  app_input->input_ctx.filename = argv[0];
   app_input->output_filename = argv[1];
   free(argv);
 
+  open_input_file(&app_input->input_ctx);
+  if (app_input->input_ctx.file_type == FILE_TYPE_Y4M) {
+    enc_cfg->g_w = app_input->input_ctx.width;
+    enc_cfg->g_h = app_input->input_ctx.height;
+  }
+
   if (enc_cfg->g_w < 16 || enc_cfg->g_w % 2 || enc_cfg->g_h < 16 ||
       enc_cfg->g_h % 2)
     die("Invalid resolution: %d x %d\n", enc_cfg->g_w, enc_cfg->g_h);
@@ -752,7 +759,6 @@ int main(int argc, const char **argv) {
   vpx_codec_err_t res;
   int pts = 0;            /* PTS starts at 0 */
   int frame_duration = 1; /* 1 timebase tick per frame */
-  FILE *infile = NULL;
   int end_of_stream = 0;
   int frames_received = 0;
 #if OUTPUT_RC_STATS
@@ -773,6 +779,13 @@ int main(int argc, const char **argv) {
   memset(&layer_id, 0, sizeof(vpx_svc_layer_id_t));
   memset(&rc, 0, sizeof(struct RateControlStats));
   exec_name = argv[0];
+
+  /* Setup default input stream settings */
+  app_input.input_ctx.framerate.numerator = 30;
+  app_input.input_ctx.framerate.denominator = 1;
+  app_input.input_ctx.only_i420 = 1;
+  app_input.input_ctx.bit_depth = 0;
+
   parse_command_line(argc, argv, &app_input, &svc_ctx, &enc_cfg);
 
 // Allocate image buffer
@@ -789,9 +802,6 @@ int main(int argc, const char **argv) {
   }
 #endif  // CONFIG_VP9_HIGHBITDEPTH
 
-  if (!(infile = fopen(app_input.input_filename, "rb")))
-    die("Failed to open %s for reading\n", app_input.input_filename);
-
   // Initialize codec
   if (vpx_svc_init(&svc_ctx, &codec, vpx_codec_vp9_cx(), &enc_cfg) !=
       VPX_CODEC_OK)
@@ -835,7 +845,8 @@ int main(int argc, const char **argv) {
 #endif
 
   // skip initial frames
-  for (i = 0; i < app_input.frames_to_skip; ++i) vpx_img_read(&raw, infile);
+  for (i = 0; i < app_input.frames_to_skip; ++i)
+    read_frame(&app_input.input_ctx, &raw);
 
   if (svc_ctx.speed != -1)
     vpx_codec_control(&codec, VP8E_SET_CPUUSED, svc_ctx.speed);
@@ -875,7 +886,8 @@ int main(int argc, const char **argv) {
     // layers, with SL0 only has TL0, and SL1 has both TL0 and TL1. This example
     // uses the extended API.
     int example_pattern = 0;
-    if (frame_cnt >= app_input.frames_to_code || !vpx_img_read(&raw, infile)) {
+    if (frame_cnt >= app_input.frames_to_code ||
+        !read_frame(&app_input.input_ctx, &raw)) {
       // We need one extra vpx_svc_encode call at end of stream to flush
       // encoder and get remaining data
       end_of_stream = 1;
@@ -1091,7 +1103,9 @@ int main(int argc, const char **argv) {
   }
 
   printf("Processed %d frames\n", frame_cnt);
-  fclose(infile);
+
+  close_input_file(&app_input.input_ctx);
+
 #if OUTPUT_RC_STATS
   if (svc_ctx.output_rc_stat) {
     printout_rate_control_summary(&rc, &enc_cfg, frame_cnt);
diff --git a/examples/vpx_temporal_svc_encoder.c b/examples/vpx_temporal_svc_encoder.c
index f49ef7b1d..aa2213a5b 100644
--- a/examples/vpx_temporal_svc_encoder.c
+++ b/examples/vpx_temporal_svc_encoder.c
@@ -19,6 +19,7 @@
 #include <string.h>
 
 #include "./vpx_config.h"
+#include "./y4minput.h"
 #include "../vpx_ports/vpx_timer.h"
 #include "vpx/vp8cx.h"
 #include "vpx/vpx_encoder.h"
@@ -594,7 +595,7 @@ int main(int argc, char **argv) {
 #endif
   vpx_svc_layer_id_t layer_id;
   const VpxInterface *encoder = NULL;
-  FILE *infile = NULL;
+  struct VpxInputContext input_ctx;
   struct RateControlMetrics rc;
   int64_t cx_time = 0;
   const int min_args_base = 13;
@@ -611,6 +612,13 @@ int main(int argc, char **argv) {
 
   zero(rc.layer_target_bitrate);
   memset(&layer_id, 0, sizeof(vpx_svc_layer_id_t));
+  memset(&input_ctx, 0, sizeof(input_ctx));
+  /* Setup default input stream settings */
+  input_ctx.framerate.numerator = 30;
+  input_ctx.framerate.denominator = 1;
+  input_ctx.only_i420 = 1;
+  input_ctx.bit_depth = 0;
+
   exec_name = argv[0];
   // Check usage and arguments.
   if (argc < min_args) {
@@ -754,9 +762,18 @@ int main(int argc, char **argv) {
   // Set to layer_target_bitrate for highest layer (total bitrate).
   cfg.rc_target_bitrate = rc.layer_target_bitrate[cfg.ts_number_layers - 1];
 
-  // Open input file.
-  if (!(infile = fopen(argv[1], "rb"))) {
-    die("Failed to open %s for reading", argv[1]);
+  input_ctx.filename = argv[1];
+  open_input_file(&input_ctx);
+
+  if (input_ctx.file_type == FILE_TYPE_Y4M) {
+    if (input_ctx.width != cfg.g_w || input_ctx.height != cfg.g_h) {
+      die("Incorrect width or height: %d x %d", cfg.g_w, cfg.g_h);
+    }
+    if (input_ctx.framerate.numerator != cfg.g_timebase.den ||
+        input_ctx.framerate.denominator != cfg.g_timebase.num) {
+      die("Incorrect framerate: numerator %d denominator %d",
+          cfg.g_timebase.num, cfg.g_timebase.den);
+    }
   }
 
   framerate = cfg.g_timebase.den / cfg.g_timebase.num;
@@ -865,7 +882,7 @@ int main(int argc, char **argv) {
     }
     flags = layer_flags[frame_cnt % flag_periodicity];
     if (layering_mode == 0) flags = 0;
-    frame_avail = vpx_img_read(&raw, infile);
+    frame_avail = read_frame(&input_ctx, &raw);
     if (frame_avail) ++rc.layer_input_frames[layer_id.temporal_layer_id];
     vpx_usec_timer_start(&timer);
     if (vpx_codec_encode(&codec, frame_avail ? &raw : NULL, pts, 1, flags,
@@ -933,7 +950,7 @@ int main(int argc, char **argv) {
     ++frame_cnt;
     pts += frame_duration;
   }
-  fclose(infile);
+  close_input_file(&input_ctx);
   printout_rate_control_summary(&rc, &cfg, frame_cnt);
   printf("\n");
   printf("Frame cnt and encoding time/FPS stats for encoding: %d %f %f \n",