Modified spatial scalable encoder & unit tests
Modifications to the spatial scalable encoder to match
changes made to the scaling code in the decoder.
In particular, the use of a dummy first frame was removed
now that the decoder is able to handle a smaller first
frame.
SvcTest.FirstFrameHasLayers unit test re-enabled.
Change-Id: Ic2e91fbe4eadf95895569947670d36d68abaf458
diff --git a/test/svc_test.cc b/test/svc_test.cc
index 98a5d94..3ddd9c1 100644
--- a/test/svc_test.cc
+++ b/test/svc_test.cc
@@ -41,7 +41,6 @@
virtual ~SvcTest() {}
virtual void SetUp() {
- svc_.first_frame_full_size = 1;
svc_.encoding_mode = INTER_LAYER_PREDICTION_IP;
svc_.log_level = SVC_LOG_DEBUG;
svc_.log_print = 0;
@@ -185,7 +184,6 @@
res = vpx_svc_set_quantizers(&svc_, NULL);
EXPECT_EQ(VPX_CODEC_INVALID_PARAM, res);
- svc_.first_frame_full_size = 0;
svc_.spatial_layers = 2;
res = vpx_svc_set_quantizers(&svc_, "40");
EXPECT_EQ(VPX_CODEC_OK, res);
@@ -206,7 +204,6 @@
res = vpx_svc_set_scale_factors(&svc_, NULL);
EXPECT_EQ(VPX_CODEC_INVALID_PARAM, res);
- svc_.first_frame_full_size = 0;
svc_.spatial_layers = 2;
res = vpx_svc_set_scale_factors(&svc_, "4/16");
EXPECT_EQ(VPX_CODEC_OK, res);
@@ -220,11 +217,8 @@
codec_initialized_ = true;
}
-// test that decoder can handle an SVC frame as the first frame in a sequence
-// this test is disabled since it always fails because of a decoder issue
-// https://code.google.com/p/webm/issues/detail?id=654
-TEST_F(SvcTest, DISABLED_FirstFrameHasLayers) {
- svc_.first_frame_full_size = 0;
+// Test that decoder can handle an SVC frame as the first frame in a sequence.
+TEST_F(SvcTest, FirstFrameHasLayers) {
svc_.spatial_layers = 2;
vpx_svc_set_scale_factors(&svc_, "4/16,16/16");
vpx_svc_set_quantizers(&svc_, "40,30");
@@ -252,7 +246,6 @@
}
TEST_F(SvcTest, EncodeThreeFrames) {
- svc_.first_frame_full_size = 1;
svc_.spatial_layers = 2;
vpx_svc_set_scale_factors(&svc_, "4/16,16/16");
vpx_svc_set_quantizers(&svc_, "40,30");
@@ -265,9 +258,9 @@
libvpx_test::I420VideoSource video(test_file_name_, kWidth, kHeight,
codec_enc_.g_timebase.den,
codec_enc_.g_timebase.num, 0, 30);
- // FRAME 1
+ // FRAME 0
video.Begin();
- // this frame is full size, with only one layer
+ // This frame is a keyframe.
res = vpx_svc_encode(&svc_, &codec_, video.img(), video.pts(),
video.duration(), VPX_DL_REALTIME);
ASSERT_EQ(VPX_CODEC_OK, res);
@@ -278,13 +271,13 @@
vpx_svc_get_frame_size(&svc_));
ASSERT_EQ(VPX_CODEC_OK, res_dec) << decoder_->DecodeError();
- // FRAME 2
+ // FRAME 1
video.Next();
- // this is an I-frame
+ // This is a P-frame.
res = vpx_svc_encode(&svc_, &codec_, video.img(), video.pts(),
video.duration(), VPX_DL_REALTIME);
ASSERT_EQ(VPX_CODEC_OK, res);
- EXPECT_EQ(1, vpx_svc_is_keyframe(&svc_));
+ EXPECT_EQ(0, vpx_svc_is_keyframe(&svc_));
res_dec = decoder_->DecodeFrame(
static_cast<const uint8_t *>(vpx_svc_get_buffer(&svc_)),
@@ -293,7 +286,7 @@
// FRAME 2
video.Next();
- // this is a P-frame
+ // This is a P-frame.
res = vpx_svc_encode(&svc_, &codec_, video.img(), video.pts(),
video.duration(), VPX_DL_REALTIME);
ASSERT_EQ(VPX_CODEC_OK, res);
@@ -306,7 +299,6 @@
}
TEST_F(SvcTest, GetLayerResolution) {
- svc_.first_frame_full_size = 0;
svc_.spatial_layers = 2;
vpx_svc_set_scale_factors(&svc_, "4/16,8/16");
vpx_svc_set_quantizers(&svc_, "40,30");
diff --git a/vp9_spatial_scalable_encoder.c b/vp9_spatial_scalable_encoder.c
index 9aaec82..a727f50 100644
--- a/vp9_spatial_scalable_encoder.c
+++ b/vp9_spatial_scalable_encoder.c
@@ -53,14 +53,11 @@
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
+ &kf_dist_arg, &scale_factors_arg, &quantizers_arg, NULL
};
static const SVC_ENCODING_MODE default_encoding_mode =
@@ -74,7 +71,6 @@
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 *output_filename;
@@ -116,8 +112,6 @@
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;
// start with default encoder configuration
res = vpx_codec_enc_config_default(vpx_codec_vp9_cx(), enc_cfg, 0);
@@ -150,8 +144,6 @@
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)) {
@@ -167,8 +159,6 @@
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;
}
@@ -195,13 +185,12 @@
"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",
+ "gop size: %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);
+ enc_cfg->rc_target_bitrate, enc_cfg->kf_max_dist);
}
int main(int argc, const char **argv) {
@@ -246,12 +235,9 @@
}
// 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_yuv_frame(&app_input.input_ctx, &raw)) break;
- }
+ while (frame_cnt < app_input.frames_to_code) {
+ if (read_yuv_frame(&app_input.input_ctx, &raw)) break;
+
res = vpx_svc_encode(&svc_ctx, &codec, &raw, pts, frame_duration,
VPX_DL_REALTIME);
printf("%s", vpx_svc_get_message(&svc_ctx));
@@ -269,7 +255,7 @@
pts += frame_duration;
}
- printf("Processed %d frames\n", frame_cnt - svc_ctx.first_frame_full_size);
+ printf("Processed %d frames\n", frame_cnt);
fclose(app_input.input_ctx.file);
if (vpx_codec_destroy(&codec)) die_codec(&codec, "Failed to destroy codec");
diff --git a/vpx/src/svc_encodeframe.c b/vpx/src/svc_encodeframe.c
index 57d21dc..810e881 100644
--- a/vpx/src/svc_encodeframe.c
+++ b/vpx/src/svc_encodeframe.c
@@ -584,21 +584,13 @@
return flags;
}
-/**
- * Helper to check if the current frame is the first, full resolution dummy.
- */
-static int vpx_svc_dummy_frame(SvcContext *svc_ctx) {
- SvcInternal *const si = get_svc_internal(svc_ctx);
- return svc_ctx->first_frame_full_size == 1 && si->encode_frame_count == 0;
-}
-
static void calculate_enc_frame_flags(SvcContext *svc_ctx) {
vpx_enc_frame_flags_t flags = VPX_EFLAG_FORCE_KF;
SvcInternal *const si = get_svc_internal(svc_ctx);
const int is_keyframe = (si->frame_within_gop == 0);
// keyframe layer zero is identical for all modes
- if ((is_keyframe && si->layer == 0) || vpx_svc_dummy_frame(svc_ctx)) {
+ if (is_keyframe && si->layer == 0) {
si->enc_frame_flags = VPX_EFLAG_FORCE_KF;
return;
}
@@ -783,10 +775,9 @@
memset(&superframe, 0, sizeof(superframe));
svc_log_reset(svc_ctx);
- si->layers = vpx_svc_dummy_frame(svc_ctx) ? 1 : svc_ctx->spatial_layers;
+ si->layers = svc_ctx->spatial_layers;
if (si->frame_within_gop >= si->kf_dist ||
- si->encode_frame_count == 0 ||
- (si->encode_frame_count == 1 && svc_ctx->first_frame_full_size == 1)) {
+ si->encode_frame_count == 0) {
si->frame_within_gop = 0;
}
si->is_keyframe = (si->frame_within_gop == 0);
@@ -805,12 +796,8 @@
}
calculate_enc_frame_flags(svc_ctx);
- if (vpx_svc_dummy_frame(svc_ctx)) {
- // do not set svc parameters, use normal encode
- svc_log(svc_ctx, SVC_LOG_DEBUG, "encoding full size first frame\n");
- } else {
- set_svc_parameters(svc_ctx, codec_ctx);
- }
+ set_svc_parameters(svc_ctx, codec_ctx);
+
res = vpx_codec_encode(codec_ctx, rawimg, pts, (uint32_t)duration,
si->enc_frame_flags, deadline);
if (res != VPX_CODEC_OK) {
@@ -822,12 +809,10 @@
switch (cx_pkt->kind) {
case VPX_CODEC_CX_FRAME_PKT: {
const uint32_t frame_pkt_size = (uint32_t)(cx_pkt->data.frame.sz);
- if (!vpx_svc_dummy_frame(svc_ctx)) {
- si->bytes_in_layer[si->layer] += frame_pkt_size;
- svc_log(svc_ctx, SVC_LOG_DEBUG,
- "SVC frame: %d, layer: %d, size: %u\n",
- si->encode_frame_count, si->layer, frame_pkt_size);
- }
+ si->bytes_in_layer[si->layer] += frame_pkt_size;
+ svc_log(svc_ctx, SVC_LOG_DEBUG,
+ "SVC frame: %d, layer: %d, size: %u\n",
+ si->encode_frame_count, si->layer, frame_pkt_size);
layer_data =
ld_create(cx_pkt->data.frame.buf, (size_t)frame_pkt_size);
if (layer_data == NULL) {
@@ -842,15 +827,13 @@
break;
}
case VPX_CODEC_PSNR_PKT: {
- if (!vpx_svc_dummy_frame(svc_ctx)) {
- svc_log(svc_ctx, SVC_LOG_DEBUG,
- "SVC frame: %d, layer: %d, PSNR(Total/Y/U/V): "
- "%2.3f %2.3f %2.3f %2.3f \n",
- si->encode_frame_count, si->layer,
- cx_pkt->data.psnr.psnr[0], cx_pkt->data.psnr.psnr[1],
- cx_pkt->data.psnr.psnr[2], cx_pkt->data.psnr.psnr[3]);
- si->psnr_in_layer[si->layer] += cx_pkt->data.psnr.psnr[0];
- }
+ svc_log(svc_ctx, SVC_LOG_DEBUG,
+ "SVC frame: %d, layer: %d, PSNR(Total/Y/U/V): "
+ "%2.3f %2.3f %2.3f %2.3f \n",
+ si->encode_frame_count, si->layer,
+ cx_pkt->data.psnr.psnr[0], cx_pkt->data.psnr.psnr[1],
+ cx_pkt->data.psnr.psnr[2], cx_pkt->data.psnr.psnr[3]);
+ si->psnr_in_layer[si->layer] += cx_pkt->data.psnr.psnr[0];
break;
}
default: {
@@ -860,11 +843,10 @@
}
}
// add superframe index to layer data list
- if (!vpx_svc_dummy_frame(svc_ctx)) {
- sf_create_index(&superframe);
- layer_data = ld_create(superframe.buffer, superframe.index_size);
- ld_list_add(&cx_layer_list, layer_data);
- }
+ sf_create_index(&superframe);
+ layer_data = ld_create(superframe.buffer, superframe.index_size);
+ ld_list_add(&cx_layer_list, layer_data);
+
// get accumulated size of layer data
si->frame_size = ld_list_get_buffer_size(cx_layer_list);
if (si->frame_size == 0) return VPX_CODEC_ERROR;
@@ -940,7 +922,6 @@
svc_log_reset(svc_ctx);
encode_frame_count = si->encode_frame_count;
- if (svc_ctx->first_frame_full_size) encode_frame_count--;
if (si->encode_frame_count <= 0) return vpx_svc_get_message(svc_ctx);
svc_log(svc_ctx, SVC_LOG_INFO, "\n");
diff --git a/vpx/svc_context.h b/vpx/svc_context.h
index f4933f8..8204f9c 100644
--- a/vpx/svc_context.h
+++ b/vpx/svc_context.h
@@ -39,7 +39,6 @@
typedef struct {
// public interface to svc_command options
int spatial_layers; // number of layers
- int first_frame_full_size; // set to one to force first frame full size
SVC_ENCODING_MODE encoding_mode; // svc encoding strategy
SVC_LOG_LEVEL log_level; // amount of information to display
int log_print; // when set, printf log messages instead of returning the
diff --git a/vpxdec.c b/vpxdec.c
index 622a461..1b9bfd3 100644
--- a/vpxdec.c
+++ b/vpxdec.c
@@ -697,7 +697,7 @@
if (arg_skip)
- fprintf(stderr, "Skiping first %d frames.\n", arg_skip);
+ fprintf(stderr, "Skipping first %d frames.\n", arg_skip);
while (arg_skip) {
if (read_frame(&input, &buf, &bytes_in_buffer, &buffer_size))
break;
