media/filters/h264_parser.h

   1 // Copyright 2014 The Chromium Authors. All rights reserved.
   2 // Use of this source code is governed by a BSD-style license that can be
   3 // found in the LICENSE file.
   4 //
   5 // This file contains an implementation of an H264 Annex-B video stream parser.
   6
   7 #ifndef MEDIA_FILTERS_H264_PARSER_H_
   8 #define MEDIA_FILTERS_H264_PARSER_H_
   9
  10 #include <sys/types.h>
  11
  12 #include <map>
  13
  14 #include "base/basictypes.h"
  15 #include "media/base/media_export.h"
  16 #include "media/filters/h264_bit_reader.h"
  17
  18 namespace media {
  19
  20 // For explanations of each struct and its members, see H.264 specification
  21 // at http://www.itu.int/rec/T-REC-H.264.
  22 struct MEDIA_EXPORT H264NALU {
  23   H264NALU();
  24
  25   enum Type {
  26     kUnspecified = 0,
  27     kNonIDRSlice = 1,
  28     kSliceDataA = 2,
  29     kSliceDataB = 3,
  30     kSliceDataC = 4,
  31     kIDRSlice = 5,
  32     kSEIMessage = 6,
  33     kSPS = 7,
  34     kPPS = 8,
  35     kAUD = 9,
  36     kEOSeq = 10,
  37     kEOStream = 11,
  38     kFiller = 12,
  39     kSPSExt = 13,
  40     kReserved14 = 14,
  41     kReserved15 = 15,
  42     kReserved16 = 16,
  43     kReserved17 = 17,
  44     kReserved18 = 18,
  45     kCodedSliceAux = 19,
  46     kCodedSliceExtension = 20,
  47   };
  48
  49   // After (without) start code; we don't own the underlying memory
  50   // and a shallow copy should be made when copying this struct.
  51   const uint8* data;
  52   off_t size;  // From after start code to start code of next NALU (or EOS).
  53
  54   int nal_ref_idc;
  55   int nal_unit_type;
  56 };
  57
  58 enum {
  59   kH264ScalingList4x4Length = 16,
  60   kH264ScalingList8x8Length = 64,
  61 };
  62
  63 struct MEDIA_EXPORT H264SPS {
  64   H264SPS();
  65
  66   enum H264ProfileIDC {
  67     kProfileIDCBaseline = 66,
  68     kProfileIDCConstrainedBaseline = kProfileIDCBaseline,
  69     kProfileIDCMain = 77,
  70     kProfileIDCHigh = 100,
  71   };
  72
  73   enum AspectRatioIdc {
  74     kExtendedSar = 255,
  75   };
  76
  77   enum {
  78     // Constants for HRD parameters (spec ch. E.2.2).
  79     kBitRateScaleConstantTerm = 6,  // Equation E-37.
  80     kCPBSizeScaleConstantTerm = 4,  // Equation E-38.
  81     kDefaultInitialCPBRemovalDelayLength = 24,
  82     kDefaultDPBOutputDelayLength = 24,
  83     kDefaultTimeOffsetLength = 24,
  84   };
  85
  86   int profile_idc;
  87   bool constraint_set0_flag;
  88   bool constraint_set1_flag;
  89   bool constraint_set2_flag;
  90   bool constraint_set3_flag;
  91   bool constraint_set4_flag;
  92   bool constraint_set5_flag;
  93   int level_idc;
  94   int seq_parameter_set_id;
  95
  96   int chroma_format_idc;
  97   bool separate_colour_plane_flag;
  98   int bit_depth_luma_minus8;
  99   int bit_depth_chroma_minus8;
 100   bool qpprime_y_zero_transform_bypass_flag;
 101
 102   bool seq_scaling_matrix_present_flag;
 103   int scaling_list4x4[6][kH264ScalingList4x4Length];
 104   int scaling_list8x8[6][kH264ScalingList8x8Length];
 105
 106   int log2_max_frame_num_minus4;
 107   int pic_order_cnt_type;
 108   int log2_max_pic_order_cnt_lsb_minus4;
 109   bool delta_pic_order_always_zero_flag;
 110   int offset_for_non_ref_pic;
 111   int offset_for_top_to_bottom_field;
 112   int num_ref_frames_in_pic_order_cnt_cycle;
 113   int expected_delta_per_pic_order_cnt_cycle;  // calculated
 114   int offset_for_ref_frame[255];
 115   int max_num_ref_frames;
 116   bool gaps_in_frame_num_value_allowed_flag;
 117   int pic_width_in_mbs_minus1;
 118   int pic_height_in_map_units_minus1;
 119   bool frame_mbs_only_flag;
 120   bool mb_adaptive_frame_field_flag;
 121   bool direct_8x8_inference_flag;
 122   bool frame_cropping_flag;
 123   int frame_crop_left_offset;
 124   int frame_crop_right_offset;
 125   int frame_crop_top_offset;
 126   int frame_crop_bottom_offset;
 127
 128   bool vui_parameters_present_flag;
 129   int sar_width;    // Set to 0 when not specified.
 130   int sar_height;   // Set to 0 when not specified.
 131   bool bitstream_restriction_flag;
 132   int max_num_reorder_frames;
 133   int max_dec_frame_buffering;
 134   bool timing_info_present_flag;
 135   int num_units_in_tick;
 136   int time_scale;
 137   bool fixed_frame_rate_flag;
 138
 139   // TODO(posciak): actually parse these instead of ParseAndIgnoreHRDParameters.
 140   bool nal_hrd_parameters_present_flag;
 141   int cpb_cnt_minus1;
 142   int bit_rate_scale;
 143   int cpb_size_scale;
 144   int bit_rate_value_minus1[32];
 145   int cpb_size_value_minus1[32];
 146   bool cbr_flag[32];
 147   int initial_cpb_removal_delay_length_minus_1;
 148   int cpb_removal_delay_length_minus1;
 149   int dpb_output_delay_length_minus1;
 150   int time_offset_length;
 151
 152   bool low_delay_hrd_flag;
 153
 154   int chroma_array_type;
 155 };
 156
 157 struct MEDIA_EXPORT H264PPS {
 158   H264PPS();
 159
 160   int pic_parameter_set_id;
 161   int seq_parameter_set_id;
 162   bool entropy_coding_mode_flag;
 163   bool bottom_field_pic_order_in_frame_present_flag;
 164   int num_slice_groups_minus1;
 165   // TODO(posciak): Slice groups not implemented, could be added at some point.
 166   int num_ref_idx_l0_default_active_minus1;
 167   int num_ref_idx_l1_default_active_minus1;
 168   bool weighted_pred_flag;
 169   int weighted_bipred_idc;
 170   int pic_init_qp_minus26;
 171   int pic_init_qs_minus26;
 172   int chroma_qp_index_offset;
 173   bool deblocking_filter_control_present_flag;
 174   bool constrained_intra_pred_flag;
 175   bool redundant_pic_cnt_present_flag;
 176   bool transform_8x8_mode_flag;
 177
 178   bool pic_scaling_matrix_present_flag;
 179   int scaling_list4x4[6][kH264ScalingList4x4Length];
 180   int scaling_list8x8[6][kH264ScalingList8x8Length];
 181
 182   int second_chroma_qp_index_offset;
 183 };
 184
 185 struct MEDIA_EXPORT H264ModificationOfPicNum {
 186   int modification_of_pic_nums_idc;
 187   union {
 188     int abs_diff_pic_num_minus1;
 189     int long_term_pic_num;
 190   };
 191 };
 192
 193 struct MEDIA_EXPORT H264WeightingFactors {
 194   bool luma_weight_flag;
 195   bool chroma_weight_flag;
 196   int luma_weight[32];
 197   int luma_offset[32];
 198   int chroma_weight[32][2];
 199   int chroma_offset[32][2];
 200 };
 201
 202 struct MEDIA_EXPORT H264DecRefPicMarking {
 203   int memory_mgmnt_control_operation;
 204   int difference_of_pic_nums_minus1;
 205   int long_term_pic_num;
 206   int long_term_frame_idx;
 207   int max_long_term_frame_idx_plus1;
 208 };
 209
 210 struct MEDIA_EXPORT H264SliceHeader {
 211   H264SliceHeader();
 212
 213   enum {
 214     kRefListSize = 32,
 215     kRefListModSize = kRefListSize
 216   };
 217
 218   enum Type {
 219     kPSlice = 0,
 220     kBSlice = 1,
 221     kISlice = 2,
 222     kSPSlice = 3,
 223     kSISlice = 4,
 224   };
 225
 226   bool IsPSlice() const;
 227   bool IsBSlice() const;
 228   bool IsISlice() const;
 229   bool IsSPSlice() const;
 230   bool IsSISlice() const;
 231
 232   bool idr_pic_flag;       // from NAL header
 233   int nal_ref_idc;         // from NAL header
 234   const uint8* nalu_data;  // from NAL header
 235   off_t nalu_size;         // from NAL header
 236   off_t header_bit_size;   // calculated
 237
 238   int first_mb_in_slice;
 239   int slice_type;
 240   int pic_parameter_set_id;
 241   int colour_plane_id;  // TODO(posciak): use this!  http://crbug.com/139878
 242   int frame_num;
 243   bool field_pic_flag;
 244   bool bottom_field_flag;
 245   int idr_pic_id;
 246   int pic_order_cnt_lsb;
 247   int delta_pic_order_cnt_bottom;
 248   int delta_pic_order_cnt[2];
 249   int redundant_pic_cnt;
 250   bool direct_spatial_mv_pred_flag;
 251
 252   bool num_ref_idx_active_override_flag;
 253   int num_ref_idx_l0_active_minus1;
 254   int num_ref_idx_l1_active_minus1;
 255   bool ref_pic_list_modification_flag_l0;
 256   bool ref_pic_list_modification_flag_l1;
 257   H264ModificationOfPicNum ref_list_l0_modifications[kRefListModSize];
 258   H264ModificationOfPicNum ref_list_l1_modifications[kRefListModSize];
 259
 260   int luma_log2_weight_denom;
 261   int chroma_log2_weight_denom;
 262
 263   bool luma_weight_l0_flag;
 264   bool chroma_weight_l0_flag;
 265   H264WeightingFactors pred_weight_table_l0;
 266
 267   bool luma_weight_l1_flag;
 268   bool chroma_weight_l1_flag;
 269   H264WeightingFactors pred_weight_table_l1;
 270
 271   bool no_output_of_prior_pics_flag;
 272   bool long_term_reference_flag;
 273
 274   bool adaptive_ref_pic_marking_mode_flag;
 275   H264DecRefPicMarking ref_pic_marking[kRefListSize];
 276
 277   int cabac_init_idc;
 278   int slice_qp_delta;
 279   bool sp_for_switch_flag;
 280   int slice_qs_delta;
 281   int disable_deblocking_filter_idc;
 282   int slice_alpha_c0_offset_div2;
 283   int slice_beta_offset_div2;
 284 };
 285
 286 struct H264SEIRecoveryPoint {
 287   int recovery_frame_cnt;
 288   bool exact_match_flag;
 289   bool broken_link_flag;
 290   int changing_slice_group_idc;
 291 };
 292
 293 struct MEDIA_EXPORT H264SEIMessage {
 294   H264SEIMessage();
 295
 296   enum Type {
 297     kSEIRecoveryPoint = 6,
 298   };
 299
 300   int type;
 301   int payload_size;
 302   union {
 303     // Placeholder; in future more supported types will contribute to more
 304     // union members here.
 305     H264SEIRecoveryPoint recovery_point;
 306   };
 307 };
 308
 309 // Class to parse an Annex-B H.264 stream,
 310 // as specified in chapters 7 and Annex B of the H.264 spec.
 311 class MEDIA_EXPORT H264Parser {
 312  public:
 313   enum Result {
 314     kOk,
 315     kInvalidStream,      // error in stream
 316     kUnsupportedStream,  // stream not supported by the parser
 317     kEOStream,           // end of stream
 318   };
 319
 320   // Find offset from start of data to next NALU start code
 321   // and size of found start code (3 or 4 bytes).
 322   // If no start code is found, offset is pointing to the first unprocessed byte
 323   // (i.e. the first byte that was not considered as a possible start of a start
 324   // code) and |*start_code_size| is set to 0.
 325   // Preconditions:
 326   // - |data_size| >= 0
 327   // Postconditions:
 328   // - |*offset| is between 0 and |data_size| included.
 329   //   It is strictly less than |data_size| if |data_size| > 0.
 330   // - |*start_code_size| is either 0, 3 or 4.
 331   static bool FindStartCode(const uint8* data, off_t data_size,
 332                             off_t* offset, off_t* start_code_size);
 333
 334   H264Parser();
 335   ~H264Parser();
 336
 337   void Reset();
 338   // Set current stream pointer to |stream| of |stream_size| in bytes,
 339   // |stream| owned by caller.
 340   void SetStream(const uint8* stream, off_t stream_size);
 341
 342   // Read the stream to find the next NALU, identify it and return
 343   // that information in |*nalu|. This advances the stream to the beginning
 344   // of this NALU, but not past it, so subsequent calls to NALU-specific
 345   // parsing functions (ParseSPS, etc.)  will parse this NALU.
 346   // If the caller wishes to skip the current NALU, it can call this function
 347   // again, instead of any NALU-type specific parse functions below.
 348   Result AdvanceToNextNALU(H264NALU* nalu);
 349
 350   // NALU-specific parsing functions.
 351   // These should be called after AdvanceToNextNALU().
 352
 353   // SPSes and PPSes are owned by the parser class and the memory for their
 354   // structures is managed here, not by the caller, as they are reused
 355   // across NALUs.
 356   //
 357   // Parse an SPS/PPS NALU and save their data in the parser, returning id
 358   // of the parsed structure in |*pps_id|/|*sps_id|.
 359   // To get a pointer to a given SPS/PPS structure, use GetSPS()/GetPPS(),
 360   // passing the returned |*sps_id|/|*pps_id| as parameter.
 361   // TODO(posciak,fischman): consider replacing returning Result from Parse*()
 362   // methods with a scoped_ptr and adding an AtEOS() function to check for EOS
 363   // if Parse*() return NULL.
 364   Result ParseSPS(int* sps_id);
 365   Result ParsePPS(int* pps_id);
 366
 367   // Return a pointer to SPS/PPS with given |sps_id|/|pps_id| or NULL if not
 368   // present.
 369   const H264SPS* GetSPS(int sps_id);
 370   const H264PPS* GetPPS(int pps_id);
 371
 372   // Slice headers and SEI messages are not used across NALUs by the parser
 373   // and can be discarded after current NALU, so the parser does not store
 374   // them, nor does it manage their memory.
 375   // The caller has to provide and manage it instead.
 376
 377   // Parse a slice header, returning it in |*shdr|. |*nalu| must be set to
 378   // the NALU returned from AdvanceToNextNALU() and corresponding to |*shdr|.
 379   Result ParseSliceHeader(const H264NALU& nalu, H264SliceHeader* shdr);
 380
 381   // Parse a SEI message, returning it in |*sei_msg|, provided and managed
 382   // by the caller.
 383   Result ParseSEI(H264SEIMessage* sei_msg);
 384
 385  private:
 386   // Move the stream pointer to the beginning of the next NALU,
 387   // i.e. pointing at the next start code.
 388   // Return true if a NALU has been found.
 389   // If a NALU is found:
 390   // - its size in bytes is returned in |*nalu_size| and includes
 391   //   the start code as well as the trailing zero bits.
 392   // - the size in bytes of the start code is returned in |*start_code_size|.
 393   bool LocateNALU(off_t* nalu_size, off_t* start_code_size);
 394
 395   // Exp-Golomb code parsing as specified in chapter 9.1 of the spec.
 396   // Read one unsigned exp-Golomb code from the stream and return in |*val|.
 397   Result ReadUE(int* val);
 398
 399   // Read one signed exp-Golomb code from the stream and return in |*val|.
 400   Result ReadSE(int* val);
 401
 402   // Parse scaling lists (see spec).
 403   Result ParseScalingList(int size, int* scaling_list, bool* use_default);
 404   Result ParseSPSScalingLists(H264SPS* sps);
 405   Result ParsePPSScalingLists(const H264SPS& sps, H264PPS* pps);
 406
 407   // Parse optional VUI parameters in SPS (see spec).
 408   Result ParseVUIParameters(H264SPS* sps);
 409   // Set |hrd_parameters_present| to true only if they are present.
 410   Result ParseAndIgnoreHRDParameters(bool* hrd_parameters_present);
 411
 412   // Parse reference picture lists' modifications (see spec).
 413   Result ParseRefPicListModifications(H264SliceHeader* shdr);
 414   Result ParseRefPicListModification(int num_ref_idx_active_minus1,
 415                                      H264ModificationOfPicNum* ref_list_mods);
 416
 417   // Parse prediction weight table (see spec).
 418   Result ParsePredWeightTable(const H264SPS& sps, H264SliceHeader* shdr);
 419
 420   // Parse weighting factors (see spec).
 421   Result ParseWeightingFactors(int num_ref_idx_active_minus1,
 422                                int chroma_array_type,
 423                                int luma_log2_weight_denom,
 424                                int chroma_log2_weight_denom,
 425                                H264WeightingFactors* w_facts);
 426
 427   // Parse decoded reference picture marking information (see spec).
 428   Result ParseDecRefPicMarking(H264SliceHeader* shdr);
 429
 430   // Pointer to the current NALU in the stream.
 431   const uint8* stream_;
 432
 433   // Bytes left in the stream after the current NALU.
 434   off_t bytes_left_;
 435
 436   H264BitReader br_;
 437
 438   // PPSes and SPSes stored for future reference.
 439   typedef std::map<int, H264SPS*> SPSById;
 440   typedef std::map<int, H264PPS*> PPSById;
 441   SPSById active_SPSes_;
 442   PPSById active_PPSes_;
 443
 444   DISALLOW_COPY_AND_ASSIGN(H264Parser);
 445 };
 446
 447 }  // namespace media
 448
 449 #endif  // MEDIA_FILTERS_H264_PARSER_H_