Move SVC per-frame loop from sample app into libvpx proper
SVC multiple layer per frame encoding is invoked with vpx_svc_init and
vpx_svc_encode. These interfaces are designed to be invoked from ffmpeg.
Additional improvements:
- make dummy frame handling a bit more explicit
- fixed bug with single layer encodes
- track individual frame sizes and psnrs instead of averages
- parameterized quantizer, 16th scalefactors, more logging,
- enabled single layer encodes to generate baseline
- include new mode for 3 layer I frame with 5 total layers
Change-Id: I46cfa600d102e208c6af8acd6132e0cc25cda8d4
diff --git a/vp9_spatial_scalable_encoder.c b/vp9_spatial_scalable_encoder.c
index 8bb582f..9acfa29 100644
--- a/vp9_spatial_scalable_encoder.c
+++ b/vp9_spatial_scalable_encoder.c
@@ -13,61 +13,101 @@
* VP9 encoding scheme based on spatial scalability for video applications
* that benefit from a scalable bitstream.
*/
-#include <stdio.h>
-#include <stdlib.h>
+
#include <stdarg.h>
-#include <time.h>
+#include <stdlib.h>
#include <string.h>
-#include <unistd.h>
-#include <libgen.h>
-#define VPX_CODEC_DISABLE_COMPAT 1
-#include "vpx/vpx_encoder.h"
+#include <time.h>
+#include "./args.h"
+#include "vpx/svc_context.h"
#include "vpx/vp8cx.h"
-#define interface (vpx_codec_vp9_cx())
-#define fourcc 0x30395056
-#define IVF_FILE_HDR_SZ (32)
-#define IVF_FRAME_HDR_SZ (12)
-#define NUM_BUFFERS 8
+#include "vpx/vpx_encoder.h"
-char *input_filename;
-char *output_filename;
-unsigned int number_frames_to_code = 60 * 60;
-unsigned int number_frames_to_skip = 0;
-unsigned int number_spatial_layers = 5;
-unsigned int key_period = 100;
+#define VP90_FOURCC 0x30395056
-typedef enum ENCODING_MODE {
- INTER_LAYER_PREDICTION_I,
- INTER_LAYER_PREDICTION_IP,
- USE_GOLDEN_FRAME
-} ENCODING_MODE;
+static const struct arg_enum_list encoding_mode_enum[] = {
+ {"i", INTER_LAYER_PREDICTION_I},
+ {"alt-ip", ALT_INTER_LAYER_PREDICTION_IP},
+ {"ip", INTER_LAYER_PREDICTION_IP},
+ {"gf", USE_GOLDEN_FRAME},
+ {NULL, 0}
+};
-static void mem_put_le16(char *mem, unsigned int val) {
+static const arg_def_t encoding_mode_arg = ARG_DEF_ENUM(
+ "m", "encoding-mode", 1, "Encoding mode algorithm", encoding_mode_enum);
+static const arg_def_t skip_frames_arg =
+ ARG_DEF("s", "skip-frames", 1, "input frames to skip");
+static const arg_def_t frames_arg =
+ ARG_DEF("f", "frames", 1, "number of frames to encode");
+static const arg_def_t width_arg = ARG_DEF("w", "width", 1, "source width");
+static const arg_def_t height_arg = ARG_DEF("h", "height", 1, "source height");
+static const arg_def_t timebase_arg =
+ ARG_DEF("t", "timebase", 1, "timebase (num/den)");
+static const arg_def_t bitrate_arg = ARG_DEF(
+ "b", "target-bitrate", 1, "encoding bitrate, in kilobits per second");
+static const arg_def_t layers_arg =
+ ARG_DEF("l", "layers", 1, "number of SVC layers");
+static const arg_def_t kf_dist_arg =
+ ARG_DEF("k", "kf-dist", 1, "number of frames between keyframes");
+static const arg_def_t scale_factors_arg =
+ ARG_DEF("r", "scale-factors", 1, "scale factors (lowest to highest layer)");
+static const arg_def_t quantizers_arg =
+ ARG_DEF("q", "quantizers", 1, "quantizers (lowest to highest layer)");
+static const arg_def_t dummy_frame_arg =
+ ARG_DEF("z", "dummy-frame", 1, "make first frame blank and full size");
+
+static const arg_def_t *svc_args[] = {
+ &encoding_mode_arg, &frames_arg, &width_arg, &height_arg,
+ &timebase_arg, &bitrate_arg, &skip_frames_arg, &layers_arg,
+ &kf_dist_arg, &scale_factors_arg, &quantizers_arg, &dummy_frame_arg,
+ NULL
+};
+
+static const SVC_ENCODING_MODE default_encoding_mode =
+ INTER_LAYER_PREDICTION_IP;
+static const uint32_t default_frames_to_skip = 0;
+static const uint32_t default_frames_to_code = 60 * 60;
+static const uint32_t default_width = 1920;
+static const uint32_t default_height = 1080;
+static const uint32_t default_timebase_num = 1;
+static const uint32_t default_timebase_den = 60;
+static const uint32_t default_bitrate = 1000;
+static const uint32_t default_spatial_layers = 5;
+static const uint32_t default_kf_dist = 100;
+static const int default_use_dummy_frame = 1;
+
+typedef struct {
+ char *input_filename;
+ char *output_filename;
+ uint32_t frames_to_code;
+ uint32_t frames_to_skip;
+} AppInput;
+
+static void mem_put_le16(char *mem, uint32_t val) {
mem[0] = val;
mem[1] = val >> 8;
}
-static void mem_put_le32(char *mem, unsigned int val) {
+static void mem_put_le32(char *mem, uint32_t val) {
mem[0] = val;
mem[1] = val >> 8;
mem[2] = val >> 16;
mem[3] = val >> 24;
}
-static void usage(char *program_name) {
- printf(
- "Usage: %s [-f frames] [-s skip_frames] [-w width] [-h height] \n\t"
- "[-n rate_num] [-d rate_den] [-b bitrate] [-l layers] "
- "<input_filename> <output_filename>\n",
- basename(program_name));
+static void usage(const char *exec_name) {
+ fprintf(stderr, "Usage: %s <options> input_filename output_filename\n",
+ exec_name);
+ fprintf(stderr, "Options:\n");
+ arg_show_usage(stderr, svc_args);
exit(EXIT_FAILURE);
}
-static void die(const char *fmt, ...) {
+void die(const char *fmt, ...) {
va_list ap;
va_start(ap, fmt);
- vprintf(fmt, ap);
+ vfprintf(stderr, fmt, ap);
if (fmt[strlen(fmt) - 1] != '\n') printf("\n");
exit(EXIT_FAILURE);
}
@@ -81,407 +121,261 @@
}
static int read_frame(FILE *f, vpx_image_t *img) {
- size_t nbytes, to_read;
+ size_t nbytes;
int res = 1;
+ int plane;
- to_read = img->w * img->h * 3 / 2;
- nbytes = fread(img->planes[0], 1, to_read, f);
- if (nbytes != to_read) {
- res = 0;
- if (nbytes > 0)
- printf("Warning: Read partial frame. Check your width & height!\n");
+ for (plane = 0; plane < 3; ++plane) {
+ uint8_t *ptr;
+ const int w = (plane ? (1 + img->d_w) / 2 : img->d_w);
+ const int h = (plane ? (1 + img->d_h) / 2 : img->d_h);
+ int r;
+
+ switch (plane) {
+ case 1:
+ ptr = img->planes[VPX_PLANE_U];
+ break;
+ case 2:
+ ptr = img->planes[VPX_PLANE_V];
+ break;
+ default:
+ ptr = img->planes[plane];
+ }
+ for (r = 0; r < h; ++r) {
+ const int to_read = w;
+
+ nbytes = fread(ptr, 1, to_read, f);
+ if (nbytes != to_read) {
+ res = 0;
+ if (nbytes > 0)
+ printf("Warning: Read partial frame. Check your width & height!\n");
+ break;
+ }
+ ptr += img->stride[plane];
+ }
+ if (!res) break;
}
return res;
}
-static int read_dummy_frame(vpx_image_t *img) {
- size_t to_read;
-
- to_read = img->w * img->h * 3 / 2;
- memset(img->planes[0], 129, to_read);
+static int create_dummy_frame(vpx_image_t *img) {
+ const size_t buf_size = img->w * img->h * 3 / 2;
+ memset(img->planes[0], 129, buf_size);
return 1;
}
-static void write_ivf_file_header(FILE *outfile, const vpx_codec_enc_cfg_t *cfg,
+static void write_ivf_file_header(FILE *outfile,
+ uint32_t width, uint32_t height,
+ int timebase_num, int timebase_den,
int frame_cnt) {
char header[32];
- if (cfg->g_pass != VPX_RC_ONE_PASS && cfg->g_pass != VPX_RC_LAST_PASS) return;
header[0] = 'D';
header[1] = 'K';
header[2] = 'I';
header[3] = 'F';
- mem_put_le16(header + 4, 0); /* version */
- mem_put_le16(header + 6, 32); /* headersize */
- mem_put_le32(header + 8, fourcc); /* headersize */
- mem_put_le16(header + 12, cfg->g_w); /* width */
- mem_put_le16(header + 14, cfg->g_h); /* height */
- mem_put_le32(header + 16, cfg->g_timebase.den); /* rate */
- mem_put_le32(header + 20, cfg->g_timebase.num); /* scale */
- mem_put_le32(header + 24, frame_cnt); /* length */
- mem_put_le32(header + 28, 0); /* unused */
+ mem_put_le16(header + 4, 0); /* version */
+ mem_put_le16(header + 6, 32); /* headersize */
+ mem_put_le32(header + 8, VP90_FOURCC); /* fourcc */
+ mem_put_le16(header + 12, width); /* width */
+ mem_put_le16(header + 14, height); /* height */
+ mem_put_le32(header + 16, timebase_den); /* rate */
+ mem_put_le32(header + 20, timebase_num); /* scale */
+ mem_put_le32(header + 24, frame_cnt); /* length */
+ mem_put_le32(header + 28, 0); /* unused */
(void)fwrite(header, 1, 32, outfile);
}
-static void write_ivf_frame_header(FILE *outfile,
- const vpx_codec_cx_pkt_t *pkt) {
+static void write_ivf_frame_header(FILE *outfile, vpx_codec_pts_t pts,
+ size_t sz) {
char header[12];
- vpx_codec_pts_t pts;
-
- if (pkt->kind != VPX_CODEC_CX_FRAME_PKT) return;
-
- pts = pkt->data.frame.pts;
- mem_put_le32(header, pkt->data.frame.sz);
+ mem_put_le32(header, (uint32_t)sz);
mem_put_le32(header + 4, pts & 0xFFFFFFFF);
mem_put_le32(header + 8, pts >> 32);
(void)fwrite(header, 1, 12, outfile);
}
-static void check_parameters() {
- if (number_spatial_layers > 5) die("Cannot support more than 5 layers");
-}
-
-static void parse_command_line(int argc, char **argv,
- vpx_codec_enc_cfg_t *cfg) {
- unsigned int width = 1920;
- unsigned int height = 1080;
- unsigned int timebase_num = 1;
- unsigned int timebase_den = 60;
- unsigned int bitrate = 1000;
- int c;
+static void parse_command_line(int argc, const char **argv_,
+ AppInput *app_input, SvcContext *svc_ctx,
+ vpx_codec_enc_cfg_t *enc_cfg) {
+ struct arg arg;
+ char **argv, **argi, **argj;
vpx_codec_err_t res;
- opterr = 0;
- while ((c = getopt(argc, argv, "f:w:h:n:d:b:s:l:p:")) != -1) switch (c) {
- case 'f':
- number_frames_to_code = atoi(optarg);
- break;
- case 'w':
- width = atoi(optarg);
- break;
- case 'h':
- height = atoi(optarg);
- break;
- case 'n':
- timebase_num = atoi(optarg);
- break;
- case 'd':
- timebase_den = atoi(optarg);
- break;
- case 'b':
- bitrate = atoi(optarg);
- break;
- case 's':
- number_frames_to_skip = atoi(optarg);
- break;
- case 'l':
- number_spatial_layers = atoi(optarg);
- break;
- case 'p':
- key_period = atoi(optarg);
- break;
- case '?':
- usage(argv[0]);
- }
+ // initialize SvcContext with parameters that will be passed to vpx_svc_init
+ svc_ctx->log_level = SVC_LOG_DEBUG;
+ svc_ctx->spatial_layers = default_spatial_layers;
+ svc_ctx->encoding_mode = default_encoding_mode;
+ // when using a dummy frame, that frame is only encoded to be full size
+ svc_ctx->first_frame_full_size = default_use_dummy_frame;
- // Parse required parameters
- if (argc - optind != 2) {
- usage(argv[0]);
- }
-
- input_filename = argv[optind];
- output_filename = argv[optind + 1];
-
- if (width < 16 || width % 2 || height < 16 || height % 2)
- die("Invalid resolution: %d x %d", width, height);
-
- /* Populate encoder configuration */
- res = vpx_codec_enc_config_default(interface, cfg, 0);
+ // start with default encoder configuration
+ res = vpx_codec_enc_config_default(vpx_codec_vp9_cx(), enc_cfg, 0);
if (res) {
die("Failed to get config: %s\n", vpx_codec_err_to_string(res));
}
- printf(
- "Codec %s\nframes: %d, skip: %d, layers: %d\n"
- "width %d, height: %d, \n"
- "num: %d, den: %d, bitrate: %d, \n"
- "key period: %d \n",
- vpx_codec_iface_name(interface), number_frames_to_code,
- number_frames_to_skip, number_spatial_layers, width, height, timebase_num,
- timebase_den, bitrate, key_period);
+ // update enc_cfg with app default values
+ enc_cfg->g_w = default_width;
+ enc_cfg->g_h = default_height;
+ enc_cfg->g_timebase.num = default_timebase_num;
+ enc_cfg->g_timebase.den = default_timebase_den;
+ enc_cfg->rc_target_bitrate = default_bitrate;
+ enc_cfg->kf_min_dist = default_kf_dist;
+ enc_cfg->kf_max_dist = default_kf_dist;
- // Do minimal check at the application level. Encoder parameters will be
- // checked internally
- check_parameters();
+ // initialize AppInput with default values
+ app_input->frames_to_code = default_frames_to_code;
+ app_input->frames_to_skip = default_frames_to_skip;
- cfg->rc_target_bitrate = bitrate;
- cfg->g_w = width;
- cfg->g_h = height;
- cfg->g_timebase.num = timebase_num;
- cfg->g_timebase.den = timebase_den;
- cfg->ss_number_layers = number_spatial_layers;
-}
+ // process command line options
+ argv = argv_dup(argc - 1, argv_ + 1);
+ for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
+ arg.argv_step = 1;
-static void set_default_configuration(vpx_codec_enc_cfg_t *cfg) {
- /* Real time parameters */
- cfg->rc_dropframe_thresh = 0;
- cfg->rc_end_usage = VPX_CBR;
- cfg->rc_resize_allowed = 0;
- cfg->rc_min_quantizer = 33;
- cfg->rc_max_quantizer = 33;
- cfg->rc_undershoot_pct = 100;
- cfg->rc_overshoot_pct = 15;
- cfg->rc_buf_initial_sz = 500;
- cfg->rc_buf_optimal_sz = 600;
- cfg->rc_buf_sz = 1000;
-
- /* Enable error resilient mode */
- cfg->g_error_resilient = 1;
- cfg->g_lag_in_frames = 0;
-
- /* Disable automatic keyframe placement */
- cfg->kf_mode = VPX_KF_DISABLED;
- cfg->kf_min_dist = cfg->kf_max_dist = 3000;
-}
-
-static void initialize_codec(vpx_codec_ctx_t *codec, vpx_codec_enc_cfg_t *cfg) {
- int max_intra_size_pct;
-
- /* Initialize codec */
- if (vpx_codec_enc_init(codec, interface, cfg, VPX_CODEC_USE_PSNR))
- die_codec(codec, "Failed to initialize encoder");
-
- vpx_codec_control(codec, VP9E_SET_SVC, 1);
- /* Cap CPU & first I-frame size */
- vpx_codec_control(codec, VP8E_SET_CPUUSED, 1);
- vpx_codec_control(codec, VP8E_SET_STATIC_THRESHOLD, 1);
- vpx_codec_control(codec, VP8E_SET_NOISE_SENSITIVITY, 1);
- vpx_codec_control(codec, VP8E_SET_TOKEN_PARTITIONS, 1);
-
- max_intra_size_pct =
- (int)(((double)cfg->rc_buf_optimal_sz * 0.5) *
- ((double)cfg->g_timebase.den / cfg->g_timebase.num) / 10.0);
- /* printf ("max_intra_size_pct=%d\n", max_intra_size_pct); */
-
- vpx_codec_control(codec, VP8E_SET_MAX_INTRA_BITRATE_PCT, max_intra_size_pct);
-}
-
-static int calculate_layer(int frame_cnt, int number_spatial_layers) {
- if (frame_cnt == 0)
- return 0;
- else
- return (frame_cnt + number_spatial_layers - 1) % number_spatial_layers;
-}
-
-static void switch_to_layer(int layer, unsigned int initial_width,
- unsigned int initial_height,
- vpx_codec_ctx_t *codec) {
- // Set layer size
- int scaling_factor_num[MAX_LAYERS] = {2, 1, 4, 2, 1};
- int scaling_factor_den[MAX_LAYERS] = {9, 3, 9, 3, 1};
-
- int quantizer[MAX_LAYERS] = {60, 53, 39, 33, 27};
-
- unsigned int current_width;
- unsigned int current_height;
-
- current_width = initial_width *
- scaling_factor_num[layer + 5 - number_spatial_layers] /
- scaling_factor_den[layer + 5 - number_spatial_layers];
- current_height = initial_height *
- scaling_factor_num[layer + 5 - number_spatial_layers] /
- scaling_factor_den[layer + 5 - number_spatial_layers];
-
- current_width += current_width % 2;
- current_height += current_height % 2;
-
- vpx_codec_control(codec, VP9E_SET_WIDTH, ¤t_width);
- vpx_codec_control(codec, VP9E_SET_HEIGHT, ¤t_height);
-
- // Set layer context
- vpx_codec_control(codec, VP9E_SET_LAYER, &layer);
- vpx_codec_control(codec, VP9E_SET_MAX_Q,
- quantizer[layer + 5 - number_spatial_layers]);
- vpx_codec_control(codec, VP9E_SET_MIN_Q,
- quantizer[layer + 5 - number_spatial_layers]);
-}
-
-static int get_flag(int is_I_frame_in_layer, int layer, ENCODING_MODE mode) {
- // First layer
- switch (mode) {
- case INTER_LAYER_PREDICTION_I:
- if (is_I_frame_in_layer && layer == 0) return VPX_EFLAG_FORCE_KF;
- if (layer == 0)
- return VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_ARF |
- VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_REF_ARF;
- else if (is_I_frame_in_layer)
- return VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_ARF |
- VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_REF_LAST;
- else
- return VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_ARF |
- VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_REF_ARF;
- break;
-
- case INTER_LAYER_PREDICTION_IP:
- if (is_I_frame_in_layer && layer == 0) return VPX_EFLAG_FORCE_KF;
- if (layer == 0)
- return VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_ARF |
- VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_REF_ARF;
- else if (is_I_frame_in_layer)
- return VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_ARF |
- VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_REF_LAST;
- else
- return VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_ARF | VP8_EFLAG_NO_REF_GF;
- break;
-
- case USE_GOLDEN_FRAME:
- if (is_I_frame_in_layer && layer == 0) return VPX_EFLAG_FORCE_KF;
- if (2 * number_spatial_layers - NUM_BUFFERS <= layer) {
- if (layer == 0)
- return VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_ARF |
- VP8_EFLAG_NO_REF_ARF;
- else if (is_I_frame_in_layer)
- return VP8_EFLAG_NO_UPD_ARF | VP8_EFLAG_NO_REF_GF |
- VP8_EFLAG_NO_REF_LAST;
- else
- return VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_ARF;
- } else {
- if (layer == 0)
- return VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_ARF |
- VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_REF_ARF;
- else if (is_I_frame_in_layer)
- return VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_ARF |
- VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_REF_LAST;
- else
- return VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_ARF |
- VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_REF_ARF;
- }
- break;
- default:
- return VPX_EFLAG_FORCE_KF;
+ if (arg_match(&arg, &encoding_mode_arg, argi)) {
+ svc_ctx->encoding_mode = arg_parse_enum_or_int(&arg);
+ } else if (arg_match(&arg, &frames_arg, argi)) {
+ app_input->frames_to_code = arg_parse_uint(&arg);
+ } else if (arg_match(&arg, &width_arg, argi)) {
+ enc_cfg->g_w = arg_parse_uint(&arg);
+ } else if (arg_match(&arg, &height_arg, argi)) {
+ enc_cfg->g_h = arg_parse_uint(&arg);
+ } else if (arg_match(&arg, &height_arg, argi)) {
+ enc_cfg->g_h = arg_parse_uint(&arg);
+ } else if (arg_match(&arg, &timebase_arg, argi)) {
+ enc_cfg->g_timebase = arg_parse_rational(&arg);
+ } else if (arg_match(&arg, &bitrate_arg, argi)) {
+ enc_cfg->rc_target_bitrate = arg_parse_uint(&arg);
+ } else if (arg_match(&arg, &skip_frames_arg, argi)) {
+ app_input->frames_to_skip = arg_parse_uint(&arg);
+ } else if (arg_match(&arg, &layers_arg, argi)) {
+ svc_ctx->spatial_layers = arg_parse_uint(&arg);
+ } else if (arg_match(&arg, &kf_dist_arg, argi)) {
+ enc_cfg->kf_min_dist = arg_parse_uint(&arg);
+ enc_cfg->kf_max_dist = enc_cfg->kf_min_dist;
+ } else if (arg_match(&arg, &scale_factors_arg, argi)) {
+ vpx_svc_set_scale_factors(svc_ctx, arg.val);
+ } else if (arg_match(&arg, &quantizers_arg, argi)) {
+ vpx_svc_set_quantizers(svc_ctx, arg.val);
+ } else if (arg_match(&arg, &dummy_frame_arg, argi)) {
+ svc_ctx->first_frame_full_size = arg_parse_int(&arg);
+ } else {
+ ++argj;
+ }
}
+
+ // Check for unrecognized options
+ for (argi = argv; *argi; ++argi)
+ if (argi[0][0] == '-' && strlen(argi[0]) > 1)
+ die("Error: Unrecognized option %s\n", *argi);
+
+ if (argv[0] == NULL || argv[1] == 0) {
+ usage(argv_[0]);
+ }
+ app_input->input_filename = argv[0];
+ app_input->output_filename = argv[1];
+ free(argv);
+
+ 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);
+
+ printf(
+ "Codec %s\nframes: %d, skip: %d\n"
+ "mode: %d, layers: %d\n"
+ "width %d, height: %d,\n"
+ "num: %d, den: %d, bitrate: %d,\n"
+ "gop size: %d, use_dummy_frame: %d\n",
+ vpx_codec_iface_name(vpx_codec_vp9_cx()), app_input->frames_to_code,
+ app_input->frames_to_skip, svc_ctx->encoding_mode,
+ svc_ctx->spatial_layers, enc_cfg->g_w, enc_cfg->g_h,
+ enc_cfg->g_timebase.num, enc_cfg->g_timebase.den,
+ enc_cfg->rc_target_bitrate, enc_cfg->kf_max_dist,
+ svc_ctx->first_frame_full_size);
}
-int main(int argc, char **argv) {
- FILE *infile, *outfile[MAX_LAYERS];
+int main(int argc, const char **argv) {
+ AppInput app_input = {0};
+ FILE *infile, *outfile;
vpx_codec_ctx_t codec;
- vpx_codec_enc_cfg_t cfg;
- int frame_cnt = 0;
+ vpx_codec_enc_cfg_t enc_cfg;
+ SvcContext svc_ctx;
+ uint32_t i;
+ uint32_t frame_cnt = 0;
vpx_image_t raw;
- int frame_avail = 1;
- int got_data = 0;
- int i;
- int frames_in_layer[MAX_LAYERS] = {0};
- clock_t before;
- clock_t after;
+ vpx_codec_err_t res;
int pts = 0; /* PTS starts at 0 */
int frame_duration = 1; /* 1 timebase tick per frame */
- parse_command_line(argc, argv, &cfg);
+ memset(&svc_ctx, 0, sizeof(svc_ctx));
+ svc_ctx.log_print = 1;
+ parse_command_line(argc, argv, &app_input, &svc_ctx, &enc_cfg);
// Allocate image buffer
- if (!vpx_img_alloc(&raw, VPX_IMG_FMT_I420, cfg.g_w, cfg.g_h, 32))
- die("Failed to allocate image", cfg.g_w, cfg.g_h);
+ if (!vpx_img_alloc(&raw, VPX_IMG_FMT_I420, enc_cfg.g_w, enc_cfg.g_h, 32))
+ die("Failed to allocate image %dx%d\n", enc_cfg.g_w, enc_cfg.g_h);
- set_default_configuration(&cfg);
+ if (!(infile = fopen(app_input.input_filename, "rb")))
+ die("Failed to open %s for reading\n", app_input.input_filename);
- /* Open input file */
- if (!(infile = fopen(input_filename, "rb")))
- die("Failed to open %s for reading", argv[1]);
+ if (!(outfile = fopen(app_input.output_filename, "wb")))
+ die("Failed to open %s for writing\n", app_input.output_filename);
- /* Open output file */
- for (i = 0; i < number_spatial_layers; i++) {
- char file_name[512];
- snprintf(file_name, sizeof(file_name), "%s_%d.ivf", output_filename, i);
- if (!(outfile[i] = fopen(file_name, "wb")))
- die("Failed to open %s for writing", file_name);
- write_ivf_file_header(outfile[i], &cfg, 0);
- }
+ // Initialize codec
+ if (vpx_svc_init(&svc_ctx, &codec, vpx_codec_vp9_cx(), &enc_cfg) !=
+ VPX_CODEC_OK)
+ die("Failed to initialize encoder\n");
- initialize_codec(&codec, &cfg);
+ write_ivf_file_header(outfile, enc_cfg.g_w, enc_cfg.g_h,
+ enc_cfg.g_timebase.num, enc_cfg.g_timebase.den, 0);
// skip initial frames
- for (i = 0; i < number_frames_to_skip; i++) {
+ for (i = 0; i < app_input.frames_to_skip; ++i) {
read_frame(infile, &raw);
}
- before = clock();
- // Encoding frames
- while ((frame_avail || got_data) &&
- frame_cnt <= number_frames_to_code * number_spatial_layers) {
- int flags = 0;
- vpx_codec_iter_t iter = NULL;
- const vpx_codec_cx_pkt_t *pkt;
-
- int layer = calculate_layer(frame_cnt, number_spatial_layers);
- int is_I_frame_in_layer =
- (((frame_cnt - 1) / number_spatial_layers % key_period) == 0);
- int is_dummy = (frame_cnt == 0);
-
- if (is_dummy) { // Dummy frame
- flags = VPX_EFLAG_FORCE_KF;
- frame_avail = read_dummy_frame(&raw);
-
- } else { // Regular frame
- // Read a new frame only at the base layer
- if (layer == 0) frame_avail = read_frame(infile, &raw);
- switch_to_layer(layer, cfg.g_w, cfg.g_h, &codec);
- flags = get_flag(is_I_frame_in_layer, layer, INTER_LAYER_PREDICTION_I);
+ // Encode frames
+ while (frame_cnt <= app_input.frames_to_code) {
+ if (frame_cnt == 0 && svc_ctx.first_frame_full_size) {
+ create_dummy_frame(&raw);
+ } else {
+ if (!read_frame(infile, &raw)) break;
}
-
- // Actual Encoding
- if (vpx_codec_encode(&codec, frame_avail ? &raw : NULL, pts, 1, flags,
- VPX_DL_REALTIME))
+ res = vpx_svc_encode(&svc_ctx, &codec, &raw, pts, frame_duration,
+ VPX_DL_REALTIME);
+ printf("%s", vpx_svc_get_message(&svc_ctx));
+ if (res != VPX_CODEC_OK) {
die_codec(&codec, "Failed to encode frame");
-
- got_data = 0;
- // Process data / Get PSNR statistics
- while ((pkt = vpx_codec_get_cx_data(&codec, &iter))) {
- got_data = 1;
- switch (pkt->kind) {
- case VPX_CODEC_CX_FRAME_PKT:
- for (i = layer; i < number_spatial_layers; i++) {
- write_ivf_frame_header(outfile[i], pkt);
- (void)fwrite(pkt->data.frame.buf, 1, pkt->data.frame.sz,
- outfile[i]);
- frames_in_layer[i]++;
- }
- break;
- case VPX_CODEC_PSNR_PKT:
- if (frame_cnt != 0)
- printf(
- "Processed Frame %d, layer %d, PSNR(Total/Y/U/V): "
- "%2.3f %2.3f %2.3f %2.3f \n",
- (frame_cnt - 1) / number_spatial_layers + 1, layer,
- pkt->data.psnr.psnr[0], pkt->data.psnr.psnr[1],
- pkt->data.psnr.psnr[2], pkt->data.psnr.psnr[3]);
- break;
- default:
- break;
- }
}
- frame_cnt++;
- // TODO(ivan): Modify ts later if(!layer)
+ if (vpx_svc_get_frame_size(&svc_ctx) > 0) {
+ write_ivf_frame_header(outfile, pts, vpx_svc_get_frame_size(&svc_ctx));
+ (void)fwrite(vpx_svc_get_buffer(&svc_ctx), 1,
+ vpx_svc_get_frame_size(&svc_ctx), outfile);
+ }
+ ++frame_cnt;
pts += frame_duration;
}
- // end while
- after = clock();
- printf("Processed %d frames in different resolutions in %ld ms.\n",
- frame_cnt - 1, (int)(after - before) / (CLOCKS_PER_SEC / 1000));
+ printf("Processed %d frames\n", frame_cnt - svc_ctx.first_frame_full_size);
fclose(infile);
-
if (vpx_codec_destroy(&codec)) die_codec(&codec, "Failed to destroy codec");
- /* Try to rewrite the output file headers with the actual frame count */
- for (i = 0; i < number_spatial_layers; i++) {
- if (!fseek(outfile[i], 0, SEEK_SET)) {
- write_ivf_file_header(outfile[i], &cfg, frames_in_layer[i]);
- }
- fclose(outfile[i]);
+ // rewrite the output file headers with the actual frame count
+ if (!fseek(outfile, 0, SEEK_SET)) {
+ write_ivf_file_header(outfile, enc_cfg.g_w, enc_cfg.g_h,
+ enc_cfg.g_timebase.num, enc_cfg.g_timebase.den,
+ frame_cnt);
}
+ fclose(outfile);
+ vpx_img_free(&raw);
+
+ // display average size, psnr
+ printf("%s", vpx_svc_dump_statistics(&svc_ctx));
+
+ vpx_svc_release(&svc_ctx);
return EXIT_SUCCESS;
}
