7#include <packager/media/codecs/vp9_parser.h>
13#include <absl/log/check.h>
14#include <absl/log/log.h>
16#include <packager/media/base/bit_reader.h>
17#include <packager/media/base/rcheck.h>
18#include <packager/media/codecs/vp_codec_configuration_record.h>
19#include <packager/media/codecs/vpx_parser.h>
25const uint32_t VP9_FRAME_MARKER = 2;
26const uint32_t VP9_SYNC_CODE = 0x498342;
27const uint32_t REFS_PER_FRAME = 3;
28const uint32_t REF_FRAMES_LOG2 = 3;
29const uint32_t REF_FRAMES = (1 << REF_FRAMES_LOG2);
30const uint32_t FRAME_CONTEXTS_LOG2 = 2;
31const uint32_t MAX_REF_LF_DELTAS = 4;
32const uint32_t MAX_MODE_LF_DELTAS = 2;
33const uint32_t QINDEX_BITS = 8;
34const uint32_t MAX_SEGMENTS = 8;
35const uint32_t SEG_TREE_PROBS = (MAX_SEGMENTS - 1);
36const uint32_t PREDICTION_PROBS = 3;
37const uint32_t SEG_LVL_MAX = 4;
38const uint32_t MI_SIZE_LOG2 = 3;
39const uint32_t MI_BLOCK_SIZE_LOG2 = (6 - MI_SIZE_LOG2);
40const uint32_t MIN_TILE_WIDTH_B64 = 4;
41const uint32_t MAX_TILE_WIDTH_B64 = 64;
43const bool SEG_FEATURE_DATA_SIGNED[SEG_LVL_MAX] = {
true,
true,
false,
false};
44const uint32_t SEG_FEATURE_DATA_MAX_BITS[SEG_LVL_MAX] = {8, 6, 2, 0};
47 VPX_COLOR_SPACE_UNKNOWN = 0,
48 VPX_COLOR_SPACE_BT_601 = 1,
49 VPX_COLOR_SPACE_BT_709 = 2,
50 VPX_COLOR_SPACE_SMPTE_170 = 3,
51 VPX_COLOR_SPACE_SMPTE_240 = 4,
52 VPX_COLOR_SPACE_BT_2020 = 5,
53 VPX_COLOR_SPACE_RESERVED = 6,
54 VPX_COLOR_SPACE_SRGB = 7,
57uint32_t RoundupShift(uint32_t value, uint32_t n) {
58 return (value + (1 << n) - 1) >> n;
62uint32_t GetNumMiUnits(uint32_t pixels) {
63 return RoundupShift(pixels, MI_SIZE_LOG2);
67uint32_t GetNumBlocks(uint32_t mi_units) {
68 return RoundupShift(mi_units, MI_BLOCK_SIZE_LOG2);
71uint32_t GetMinLog2TileCols(uint32_t sb64_cols) {
72 uint32_t min_log2 = 0;
73 while ((MAX_TILE_WIDTH_B64 << min_log2) < sb64_cols)
78uint32_t GetMaxLog2TileCols(uint32_t sb64_cols) {
79 uint32_t max_log2 = 1;
80 while ((sb64_cols >> max_log2) >= MIN_TILE_WIDTH_B64)
85void GetTileNBits(uint32_t mi_cols,
86 uint32_t* min_log2_tile_cols,
87 uint32_t* max_log2_tile_cols) {
88 const uint32_t sb64_cols = GetNumBlocks(mi_cols);
89 *min_log2_tile_cols = GetMinLog2TileCols(sb64_cols);
90 *max_log2_tile_cols = GetMaxLog2TileCols(sb64_cols);
91 CHECK_LE(*min_log2_tile_cols, *max_log2_tile_cols);
98bool ParseIfSuperframeIndex(
const uint8_t* data,
100 std::vector<VPxFrameInfo>* vpx_frames) {
102 uint8_t superframe_marker = data[data_size - 1];
103 VPxFrameInfo vpx_frame;
104 if ((superframe_marker & 0xe0) != 0xc0) {
106 vpx_frame.frame_size = data_size;
107 vpx_frames->push_back(vpx_frame);
111 const size_t num_frames = (superframe_marker & 0x07) + 1;
112 const size_t frame_size_length = ((superframe_marker >> 3) & 0x03) + 1;
114 const size_t index_size = 2 + num_frames * frame_size_length;
116 if (data_size < index_size) {
117 LOG(ERROR) <<
"This chunk is marked as having a superframe index but "
118 "doesn't have enough data for it.";
121 const uint8_t superframe_marker2 = data[data_size - index_size];
122 if (superframe_marker2 != superframe_marker) {
123 LOG(ERROR) <<
"This chunk is marked as having a superframe index but "
124 "doesn't have the matching marker byte at the front of the "
128 VLOG(3) <<
"Superframe num_frames=" << num_frames
129 <<
" frame_size_length=" << frame_size_length;
131 data += data_size - index_size + 1;
132 size_t total_frame_sizes = 0;
133 for (
size_t frame = 0; frame < num_frames; ++frame) {
134 vpx_frame.frame_size = 0;
135 for (
size_t i = 0; i < frame_size_length; ++i) {
136 vpx_frame.frame_size |= *data << (i * 8);
139 total_frame_sizes += vpx_frame.frame_size;
140 vpx_frames->push_back(vpx_frame);
142 if (total_frame_sizes + index_size != data_size) {
143 LOG(ERROR) <<
"Data size (" << data_size
144 <<
") does not match with sum of frame sizes ("
145 << total_frame_sizes <<
") + index_size (" << index_size <<
")";
151bool ReadProfile(BitReader* reader, VPCodecConfigurationRecord* codec_config) {
153 RCHECK(reader->ReadBits(1, &bit[0]));
154 RCHECK(reader->ReadBits(1, &bit[1]));
155 uint8_t profile = bit[0] | (bit[1] << 1);
158 RCHECK(reader->ReadBits(1, &reserved));
161 codec_config->set_profile(profile);
165bool ReadSyncCode(BitReader* reader) {
167 RCHECK(reader->ReadBits(24, &sync_code));
168 return sync_code == VP9_SYNC_CODE;
171void SetColorAttributes(uint8_t bit_depth,
173 VPCodecConfigurationRecord* codec_config) {
174 switch (color_space) {
175 case VPX_COLOR_SPACE_UNKNOWN:
176 codec_config->set_color_primaries(AVCOL_PRI_UNSPECIFIED);
177 codec_config->set_matrix_coefficients(AVCOL_SPC_UNSPECIFIED);
178 codec_config->set_transfer_characteristics(AVCOL_TRC_UNSPECIFIED);
180 case VPX_COLOR_SPACE_BT_601:
182 codec_config->set_color_primaries(AVCOL_PRI_UNSPECIFIED);
183 codec_config->set_matrix_coefficients(AVCOL_SPC_UNSPECIFIED);
184 codec_config->set_transfer_characteristics(AVCOL_TRC_SMPTE170M);
186 case VPX_COLOR_SPACE_BT_709:
187 codec_config->set_color_primaries(AVCOL_PRI_BT709);
188 codec_config->set_matrix_coefficients(AVCOL_SPC_BT709);
189 codec_config->set_transfer_characteristics(AVCOL_TRC_BT709);
191 case VPX_COLOR_SPACE_SMPTE_170:
192 codec_config->set_color_primaries(AVCOL_PRI_SMPTE170M);
193 codec_config->set_matrix_coefficients(AVCOL_SPC_SMPTE170M);
194 codec_config->set_transfer_characteristics(AVCOL_TRC_SMPTE170M);
196 case VPX_COLOR_SPACE_SMPTE_240:
197 codec_config->set_color_primaries(AVCOL_PRI_SMPTE240M);
198 codec_config->set_matrix_coefficients(AVCOL_SPC_SMPTE240M);
199 codec_config->set_transfer_characteristics(AVCOL_TRC_SMPTE240M);
201 case VPX_COLOR_SPACE_BT_2020:
202 codec_config->set_color_primaries(AVCOL_PRI_BT2020);
207 codec_config->set_matrix_coefficients(AVCOL_SPC_BT2020_NCL);
210 codec_config->set_transfer_characteristics(AVCOL_TRC_BT2020_10);
213 codec_config->set_transfer_characteristics(AVCOL_TRC_BT2020_12);
216 codec_config->set_transfer_characteristics(AVCOL_TRC_UNSPECIFIED);
220 case VPX_COLOR_SPACE_SRGB:
221 codec_config->set_color_primaries(AVCOL_PRI_UNSPECIFIED);
222 codec_config->set_matrix_coefficients(AVCOL_SPC_RGB);
223 codec_config->set_transfer_characteristics(AVCOL_TRC_UNSPECIFIED);
226 LOG(WARNING) <<
"Unknown color space: " <<
static_cast<int>(color_space);
227 codec_config->set_color_primaries(AVCOL_PRI_UNSPECIFIED);
228 codec_config->set_matrix_coefficients(AVCOL_SPC_UNSPECIFIED);
229 codec_config->set_transfer_characteristics(AVCOL_TRC_UNSPECIFIED);
234VPCodecConfigurationRecord::ChromaSubsampling GetChromaSubsampling(
235 uint8_t subsampling) {
236 switch (subsampling) {
238 return VPCodecConfigurationRecord::CHROMA_444;
240 return VPCodecConfigurationRecord::CHROMA_440;
242 return VPCodecConfigurationRecord::CHROMA_422;
246 return VPCodecConfigurationRecord::CHROMA_420_COLLOCATED_WITH_LUMA;
248 LOG(WARNING) <<
"Unexpected chroma subsampling value: "
249 <<
static_cast<int>(subsampling);
250 return VPCodecConfigurationRecord::CHROMA_420_COLLOCATED_WITH_LUMA;
254bool ReadBitDepthAndColorSpace(BitReader* reader,
255 VPCodecConfigurationRecord* codec_config) {
256 uint8_t bit_depth = 8;
257 if (codec_config->profile() >= 2) {
258 bool use_vpx_bits_12;
259 RCHECK(reader->ReadBits(1, &use_vpx_bits_12));
260 bit_depth = use_vpx_bits_12 ? 12 : 10;
262 codec_config->set_bit_depth(bit_depth);
265 RCHECK(reader->ReadBits(3, &color_space));
266 SetColorAttributes(bit_depth, color_space, codec_config);
268 bool yuv_full_range =
false;
269 auto chroma_subsampling = VPCodecConfigurationRecord::CHROMA_444;
270 if (color_space != VPX_COLOR_SPACE_SRGB) {
271 RCHECK(reader->ReadBits(1, &yuv_full_range));
273 if (codec_config->profile() & 1) {
275 RCHECK(reader->ReadBits(2, &subsampling));
276 chroma_subsampling = GetChromaSubsampling(subsampling);
277 if (chroma_subsampling ==
278 VPCodecConfigurationRecord::CHROMA_420_COLLOCATED_WITH_LUMA) {
279 LOG(ERROR) <<
"4:2:0 color not supported in profile "
280 <<
static_cast<int>(codec_config->profile());
285 RCHECK(reader->ReadBits(1, &reserved));
289 VPCodecConfigurationRecord::CHROMA_420_COLLOCATED_WITH_LUMA;
293 yuv_full_range =
true;
294 chroma_subsampling = VPCodecConfigurationRecord::CHROMA_444;
295 if (codec_config->profile() & 1) {
297 RCHECK(reader->ReadBits(1, &reserved));
300 LOG(ERROR) <<
"4:4:4 color not supported in profile 0 or 2.";
304 codec_config->set_video_full_range_flag(yuv_full_range);
305 codec_config->SetChromaSubsampling(chroma_subsampling);
307 VLOG(3) <<
"\n profile " <<
static_cast<int>(codec_config->profile())
308 <<
"\n bit depth " <<
static_cast<int>(codec_config->bit_depth())
309 <<
"\n matrix coefficients "
310 <<
static_cast<int>(codec_config->matrix_coefficients())
312 <<
static_cast<int>(codec_config->video_full_range_flag())
313 <<
"\n chroma subsampling "
314 <<
static_cast<int>(codec_config->chroma_subsampling());
318bool ReadFrameSize(BitReader* reader, uint32_t* width, uint32_t* height) {
319 RCHECK(reader->ReadBits(16, width));
321 RCHECK(reader->ReadBits(16, height));
326bool ReadDisplayFrameSize(BitReader* reader,
327 uint32_t* display_width,
328 uint32_t* display_height) {
329 bool has_display_size;
330 RCHECK(reader->ReadBits(1, &has_display_size));
331 if (has_display_size)
332 RCHECK(ReadFrameSize(reader, display_width, display_height));
336bool ReadFrameSizes(BitReader* reader, uint32_t* width, uint32_t* height) {
339 RCHECK(ReadFrameSize(reader, &new_width, &new_height));
340 if (new_width != *width) {
341 VLOG(1) <<
"Width updates from " << *width <<
" to " << new_width;
344 if (new_height != *height) {
345 VLOG(1) <<
"Height updates from " << *height <<
" to " << new_height;
346 *height = new_height;
349 uint32_t display_width = *width;
350 uint32_t display_height = *height;
351 RCHECK(ReadDisplayFrameSize(reader, &display_width, &display_height));
355bool ReadFrameSizesWithRefs(BitReader* reader,
359 for (uint32_t i = 0; i < REFS_PER_FRAME; ++i) {
360 RCHECK(reader->ReadBits(1, &found));
365 RCHECK(ReadFrameSizes(reader, width, height));
367 uint32_t display_width;
368 uint32_t display_height;
369 RCHECK(ReadDisplayFrameSize(reader, &display_width, &display_height));
374bool ReadLoopFilter(BitReader* reader) {
375 RCHECK(reader->SkipBits(9));
376 bool mode_ref_delta_enabled;
377 RCHECK(reader->ReadBits(1, &mode_ref_delta_enabled));
378 if (!mode_ref_delta_enabled)
380 bool mode_ref_delta_update;
381 RCHECK(reader->ReadBits(1, &mode_ref_delta_update));
382 if (!mode_ref_delta_update)
385 for (uint32_t i = 0; i < MAX_REF_LF_DELTAS + MAX_MODE_LF_DELTAS; ++i)
386 RCHECK(reader->SkipBitsConditional(
true, 6 + 1));
390bool ReadQuantization(BitReader* reader) {
391 RCHECK(reader->SkipBits(QINDEX_BITS));
393 for (uint32_t i = 0; i < 3; ++i)
394 RCHECK(reader->SkipBitsConditional(
true, 4 + 1));
398bool ReadSegmentation(BitReader* reader) {
400 RCHECK(reader->ReadBits(1, &enabled));
405 RCHECK(reader->ReadBits(1, &update_map));
407 for (uint32_t i = 0; i < SEG_TREE_PROBS; ++i)
408 RCHECK(reader->SkipBitsConditional(
true, 8));
410 bool temporal_update;
411 RCHECK(reader->ReadBits(1, &temporal_update));
412 if (temporal_update) {
413 for (uint32_t j = 0; j < PREDICTION_PROBS; ++j)
414 RCHECK(reader->SkipBitsConditional(
true, 8));
419 RCHECK(reader->ReadBits(1, &update_data));
421 RCHECK(reader->SkipBits(1));
422 for (uint32_t i = 0; i < MAX_SEGMENTS; ++i) {
423 for (uint32_t j = 0; j < SEG_LVL_MAX; ++j) {
424 bool feature_enabled;
425 RCHECK(reader->ReadBits(1, &feature_enabled));
426 if (feature_enabled) {
427 RCHECK(reader->SkipBits(SEG_FEATURE_DATA_MAX_BITS[j]));
428 if (SEG_FEATURE_DATA_SIGNED[j])
429 RCHECK(reader->SkipBits(1));
437bool ReadTileInfo(uint32_t width, BitReader* reader) {
438 uint32_t mi_cols = GetNumMiUnits(width);
440 uint32_t min_log2_tile_cols;
441 uint32_t max_log2_tile_cols;
442 GetTileNBits(mi_cols, &min_log2_tile_cols, &max_log2_tile_cols);
443 uint32_t max_ones = max_log2_tile_cols - min_log2_tile_cols;
445 uint32_t log2_tile_cols = min_log2_tile_cols;
448 RCHECK(reader->ReadBits(1, &has_more));
453 RCHECK(log2_tile_cols <= 6);
455 RCHECK(reader->SkipBitsConditional(
true, 1));
461VP9Parser::VP9Parser() : width_(0), height_(0) {}
462VP9Parser::~VP9Parser() {}
464bool VP9Parser::Parse(
const uint8_t* data,
466 std::vector<VPxFrameInfo>* vpx_frames) {
469 RCHECK(ParseIfSuperframeIndex(data, data_size, vpx_frames));
471 for (
auto& vpx_frame : *vpx_frames) {
472 VLOG(4) <<
"process frame with size " << vpx_frame.frame_size;
473 BitReader reader(data, vpx_frame.frame_size);
474 uint8_t frame_marker;
475 RCHECK(reader.
ReadBits(2, &frame_marker));
476 RCHECK(frame_marker == VP9_FRAME_MARKER);
478 RCHECK(ReadProfile(&reader, writable_codec_config()));
480 bool show_existing_frame;
481 RCHECK(reader.
ReadBits(1, &show_existing_frame));
482 if (show_existing_frame) {
487 vpx_frame.is_keyframe =
false;
488 vpx_frame.uncompressed_header_size = vpx_frame.frame_size;
489 vpx_frame.width = width_;
490 vpx_frame.height = height_;
495 RCHECK(reader.
ReadBits(1, &is_interframe));
496 vpx_frame.is_keyframe = !is_interframe;
499 RCHECK(reader.
ReadBits(1, &show_frame));
500 bool error_resilient_mode;
501 RCHECK(reader.
ReadBits(1, &error_resilient_mode));
503 if (vpx_frame.is_keyframe) {
504 RCHECK(ReadSyncCode(&reader));
505 RCHECK(ReadBitDepthAndColorSpace(&reader, writable_codec_config()));
506 RCHECK(ReadFrameSizes(&reader, &width_, &height_));
508 bool intra_only =
false;
510 RCHECK(reader.
ReadBits(1, &intra_only));
511 if (!error_resilient_mode)
515 RCHECK(ReadSyncCode(&reader));
516 if (codec_config().profile() > 0) {
517 RCHECK(ReadBitDepthAndColorSpace(&reader, writable_codec_config()));
523 writable_codec_config()->SetChromaSubsampling(
524 VPCodecConfigurationRecord::CHROMA_420_COLLOCATED_WITH_LUMA);
525 writable_codec_config()->set_bit_depth(8);
528 RCHECK(reader.
SkipBits(REF_FRAMES));
529 RCHECK(ReadFrameSizes(&reader, &width_, &height_));
531 RCHECK(reader.
SkipBits(REF_FRAMES));
532 RCHECK(reader.
SkipBits(REFS_PER_FRAME * (REF_FRAMES_LOG2 + 1)));
537 RCHECK(ReadFrameSizesWithRefs(&reader, &width_, &height_));
542 RCHECK(reader.
ReadBits(1, &interp_filter));
548 if (!error_resilient_mode) {
552 RCHECK(reader.
SkipBits(FRAME_CONTEXTS_LOG2));
554 VLOG(4) <<
"bits read before ReadLoopFilter: " << reader.
bit_position();
555 RCHECK(ReadLoopFilter(&reader));
556 RCHECK(ReadQuantization(&reader));
557 RCHECK(ReadSegmentation(&reader));
558 RCHECK(ReadTileInfo(width_, &reader));
560 uint16_t header_size;
561 RCHECK(reader.
ReadBits(16, &header_size));
562 vpx_frame.uncompressed_header_size =
564 vpx_frame.width = width_;
565 vpx_frame.height = height_;
567 VLOG(3) <<
"\n frame_size: " << vpx_frame.frame_size
568 <<
"\n uncompressed_header_size: "
569 << vpx_frame.uncompressed_header_size
571 <<
"\n header_size: " << header_size;
573 RCHECK(header_size > 0);
576 data += vpx_frame.frame_size;
581bool VP9Parser::IsKeyframe(
const uint8_t* data,
size_t data_size) {
583 uint8_t frame_marker;
584 RCHECK(reader.
ReadBits(2, &frame_marker));
585 RCHECK(frame_marker == VP9_FRAME_MARKER);
588 RCHECK(ReadProfile(&reader, &codec_config));
590 bool show_existing_frame;
591 RCHECK(reader.
ReadBits(1, &show_existing_frame));
592 if (show_existing_frame)
596 RCHECK(reader.
ReadBits(1, &is_interframe));
602 RCHECK(ReadSyncCode(&reader));
All the methods that are virtual are virtual for mocking.