Home | History | Annotate | Download | only in vda 
      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 H264_PARSER_H_
      8 #define H264_PARSER_H_
      9 
     10 #include <stddef.h>
     11 #include <stdint.h>
     12 #include <sys/types.h>
     13 
     14 #include <map>
     15 #include <memory>
     16 #include <vector>
     17 
     18 #include "base/macros.h"
     19 #include "base/optional.h"
     20 #include "h264_bit_reader.h"
     21 #include "ranges.h"
     22 #include "size.h"
     23 #include "subsample_entry.h"
     24 
     25 namespace media {
     26 
     27 struct SubsampleEntry;
     28 
     29 // For explanations of each struct and its members, see H.264 specification
     30 // at http://www.itu.int/rec/T-REC-H.264.
     31 struct H264NALU {
     32   H264NALU();
     33 
     34   enum Type {
     35     kUnspecified = 0,
     36     kNonIDRSlice = 1,
     37     kSliceDataA = 2,
     38     kSliceDataB = 3,
     39     kSliceDataC = 4,
     40     kIDRSlice = 5,
     41     kSEIMessage = 6,
     42     kSPS = 7,
     43     kPPS = 8,
     44     kAUD = 9,
     45     kEOSeq = 10,
     46     kEOStream = 11,
     47     kFiller = 12,
     48     kSPSExt = 13,
     49     kReserved14 = 14,
     50     kReserved15 = 15,
     51     kReserved16 = 16,
     52     kReserved17 = 17,
     53     kReserved18 = 18,
     54     kCodedSliceAux = 19,
     55     kCodedSliceExtension = 20,
     56   };
     57 
     58   // After (without) start code; we don't own the underlying memory
     59   // and a shallow copy should be made when copying this struct.
     60   const uint8_t* data;
     61   off_t size;  // From after start code to start code of next NALU (or EOS).
     62 
     63   int nal_ref_idc;
     64   int nal_unit_type;
     65 };
     66 
     67 enum {
     68   kH264ScalingList4x4Length = 16,
     69   kH264ScalingList8x8Length = 64,
     70 };
     71 
     72 struct H264SPS {
     73   H264SPS();
     74 
     75   enum H264ProfileIDC {
     76     kProfileIDCBaseline = 66,
     77     kProfileIDCConstrainedBaseline = kProfileIDCBaseline,
     78     kProfileIDCMain = 77,
     79     kProfileIDScalableBaseline = 83,
     80     kProfileIDScalableHigh = 86,
     81     kProfileIDCHigh = 100,
     82     kProfileIDHigh10 = 110,
     83     kProfileIDSMultiviewHigh = 118,
     84     kProfileIDHigh422 = 122,
     85     kProfileIDStereoHigh = 128,
     86     kProfileIDHigh444Predictive = 244,
     87   };
     88 
     89   enum AspectRatioIdc {
     90     kExtendedSar = 255,
     91   };
     92 
     93   enum {
     94     // Constants for HRD parameters (spec ch. E.2.2).
     95     kBitRateScaleConstantTerm = 6,  // Equation E-37.
     96     kCPBSizeScaleConstantTerm = 4,  // Equation E-38.
     97     kDefaultInitialCPBRemovalDelayLength = 24,
     98     kDefaultDPBOutputDelayLength = 24,
     99     kDefaultTimeOffsetLength = 24,
    100   };
    101 
    102   int profile_idc;
    103   bool constraint_set0_flag;
    104   bool constraint_set1_flag;
    105   bool constraint_set2_flag;
    106   bool constraint_set3_flag;
    107   bool constraint_set4_flag;
    108   bool constraint_set5_flag;
    109   int level_idc;
    110   int seq_parameter_set_id;
    111 
    112   int chroma_format_idc;
    113   bool separate_colour_plane_flag;
    114   int bit_depth_luma_minus8;
    115   int bit_depth_chroma_minus8;
    116   bool qpprime_y_zero_transform_bypass_flag;
    117 
    118   bool seq_scaling_matrix_present_flag;
    119   int scaling_list4x4[6][kH264ScalingList4x4Length];
    120   int scaling_list8x8[6][kH264ScalingList8x8Length];
    121 
    122   int log2_max_frame_num_minus4;
    123   int pic_order_cnt_type;
    124   int log2_max_pic_order_cnt_lsb_minus4;
    125   bool delta_pic_order_always_zero_flag;
    126   int offset_for_non_ref_pic;
    127   int offset_for_top_to_bottom_field;
    128   int num_ref_frames_in_pic_order_cnt_cycle;
    129   int expected_delta_per_pic_order_cnt_cycle;  // calculated
    130   int offset_for_ref_frame[255];
    131   int max_num_ref_frames;
    132   bool gaps_in_frame_num_value_allowed_flag;
    133   int pic_width_in_mbs_minus1;
    134   int pic_height_in_map_units_minus1;
    135   bool frame_mbs_only_flag;
    136   bool mb_adaptive_frame_field_flag;
    137   bool direct_8x8_inference_flag;
    138   bool frame_cropping_flag;
    139   int frame_crop_left_offset;
    140   int frame_crop_right_offset;
    141   int frame_crop_top_offset;
    142   int frame_crop_bottom_offset;
    143 
    144   bool vui_parameters_present_flag;
    145   int sar_width;    // Set to 0 when not specified.
    146   int sar_height;   // Set to 0 when not specified.
    147   bool bitstream_restriction_flag;
    148   int max_num_reorder_frames;
    149   int max_dec_frame_buffering;
    150   bool timing_info_present_flag;
    151   int num_units_in_tick;
    152   int time_scale;
    153   bool fixed_frame_rate_flag;
    154 
    155   bool video_signal_type_present_flag;
    156   int video_format;
    157   bool video_full_range_flag;
    158   bool colour_description_present_flag;
    159   int colour_primaries;
    160   int transfer_characteristics;
    161   int matrix_coefficients;
    162 
    163   // TODO(posciak): actually parse these instead of ParseAndIgnoreHRDParameters.
    164   bool nal_hrd_parameters_present_flag;
    165   int cpb_cnt_minus1;
    166   int bit_rate_scale;
    167   int cpb_size_scale;
    168   int bit_rate_value_minus1[32];
    169   int cpb_size_value_minus1[32];
    170   bool cbr_flag[32];
    171   int initial_cpb_removal_delay_length_minus_1;
    172   int cpb_removal_delay_length_minus1;
    173   int dpb_output_delay_length_minus1;
    174   int time_offset_length;
    175 
    176   bool low_delay_hrd_flag;
    177 
    178   int chroma_array_type;
    179 
    180   // Helpers to compute frequently-used values. These methods return
    181   // base::nullopt if they encounter integer overflow. They do not verify that
    182   // the results are in-spec for the given profile or level.
    183   base::Optional<Size> GetCodedSize() const;
    184 };
    185 
    186 struct H264PPS {
    187   H264PPS();
    188 
    189   int pic_parameter_set_id;
    190   int seq_parameter_set_id;
    191   bool entropy_coding_mode_flag;
    192   bool bottom_field_pic_order_in_frame_present_flag;
    193   int num_slice_groups_minus1;
    194   // TODO(posciak): Slice groups not implemented, could be added at some point.
    195   int num_ref_idx_l0_default_active_minus1;
    196   int num_ref_idx_l1_default_active_minus1;
    197   bool weighted_pred_flag;
    198   int weighted_bipred_idc;
    199   int pic_init_qp_minus26;
    200   int pic_init_qs_minus26;
    201   int chroma_qp_index_offset;
    202   bool deblocking_filter_control_present_flag;
    203   bool constrained_intra_pred_flag;
    204   bool redundant_pic_cnt_present_flag;
    205   bool transform_8x8_mode_flag;
    206 
    207   bool pic_scaling_matrix_present_flag;
    208   int scaling_list4x4[6][kH264ScalingList4x4Length];
    209   int scaling_list8x8[6][kH264ScalingList8x8Length];
    210 
    211   int second_chroma_qp_index_offset;
    212 };
    213 
    214 struct H264ModificationOfPicNum {
    215   int modification_of_pic_nums_idc;
    216   union {
    217     int abs_diff_pic_num_minus1;
    218     int long_term_pic_num;
    219   };
    220 };
    221 
    222 struct H264WeightingFactors {
    223   bool luma_weight_flag;
    224   bool chroma_weight_flag;
    225   int luma_weight[32];
    226   int luma_offset[32];
    227   int chroma_weight[32][2];
    228   int chroma_offset[32][2];
    229 };
    230 
    231 struct H264DecRefPicMarking {
    232   int memory_mgmnt_control_operation;
    233   int difference_of_pic_nums_minus1;
    234   int long_term_pic_num;
    235   int long_term_frame_idx;
    236   int max_long_term_frame_idx_plus1;
    237 };
    238 
    239 struct H264SliceHeader {
    240   H264SliceHeader();
    241 
    242   enum {
    243     kRefListSize = 32,
    244     kRefListModSize = kRefListSize
    245   };
    246 
    247   enum Type {
    248     kPSlice = 0,
    249     kBSlice = 1,
    250     kISlice = 2,
    251     kSPSlice = 3,
    252     kSISlice = 4,
    253   };
    254 
    255   bool IsPSlice() const;
    256   bool IsBSlice() const;
    257   bool IsISlice() const;
    258   bool IsSPSlice() const;
    259   bool IsSISlice() const;
    260 
    261   bool idr_pic_flag;       // from NAL header
    262   int nal_ref_idc;         // from NAL header
    263   const uint8_t* nalu_data;  // from NAL header
    264   off_t nalu_size;         // from NAL header
    265   off_t header_bit_size;   // calculated
    266 
    267   int first_mb_in_slice;
    268   int slice_type;
    269   int pic_parameter_set_id;
    270   int colour_plane_id;  // TODO(posciak): use this!  http://crbug.com/139878
    271   int frame_num;
    272   bool field_pic_flag;
    273   bool bottom_field_flag;
    274   int idr_pic_id;
    275   int pic_order_cnt_lsb;
    276   int delta_pic_order_cnt_bottom;
    277   int delta_pic_order_cnt0;
    278   int delta_pic_order_cnt1;
    279   int redundant_pic_cnt;
    280   bool direct_spatial_mv_pred_flag;
    281 
    282   bool num_ref_idx_active_override_flag;
    283   int num_ref_idx_l0_active_minus1;
    284   int num_ref_idx_l1_active_minus1;
    285   bool ref_pic_list_modification_flag_l0;
    286   bool ref_pic_list_modification_flag_l1;
    287   H264ModificationOfPicNum ref_list_l0_modifications[kRefListModSize];
    288   H264ModificationOfPicNum ref_list_l1_modifications[kRefListModSize];
    289 
    290   int luma_log2_weight_denom;
    291   int chroma_log2_weight_denom;
    292 
    293   bool luma_weight_l0_flag;
    294   bool chroma_weight_l0_flag;
    295   H264WeightingFactors pred_weight_table_l0;
    296 
    297   bool luma_weight_l1_flag;
    298   bool chroma_weight_l1_flag;
    299   H264WeightingFactors pred_weight_table_l1;
    300 
    301   bool no_output_of_prior_pics_flag;
    302   bool long_term_reference_flag;
    303 
    304   bool adaptive_ref_pic_marking_mode_flag;
    305   H264DecRefPicMarking ref_pic_marking[kRefListSize];
    306 
    307   int cabac_init_idc;
    308   int slice_qp_delta;
    309   bool sp_for_switch_flag;
    310   int slice_qs_delta;
    311   int disable_deblocking_filter_idc;
    312   int slice_alpha_c0_offset_div2;
    313   int slice_beta_offset_div2;
    314 
    315   // Calculated.
    316   // Size in bits of dec_ref_pic_marking() syntax element.
    317   size_t dec_ref_pic_marking_bit_size;
    318   size_t pic_order_cnt_bit_size;
    319 };
    320 
    321 struct H264SEIRecoveryPoint {
    322   int recovery_frame_cnt;
    323   bool exact_match_flag;
    324   bool broken_link_flag;
    325   int changing_slice_group_idc;
    326 };
    327 
    328 struct H264SEIMessage {
    329   H264SEIMessage();
    330 
    331   enum Type {
    332     kSEIRecoveryPoint = 6,
    333   };
    334 
    335   int type;
    336   int payload_size;
    337   union {
    338     // Placeholder; in future more supported types will contribute to more
    339     // union members here.
    340     H264SEIRecoveryPoint recovery_point;
    341   };
    342 };
    343 
    344 // Class to parse an Annex-B H.264 stream,
    345 // as specified in chapters 7 and Annex B of the H.264 spec.
    346 class H264Parser {
    347  public:
    348   enum Result {
    349     kOk,
    350     kInvalidStream,      // error in stream
    351     kUnsupportedStream,  // stream not supported by the parser
    352     kEOStream,           // end of stream
    353   };
    354 
    355   // Find offset from start of data to next NALU start code
    356   // and size of found start code (3 or 4 bytes).
    357   // If no start code is found, offset is pointing to the first unprocessed byte
    358   // (i.e. the first byte that was not considered as a possible start of a start
    359   // code) and |*start_code_size| is set to 0.
    360   // Preconditions:
    361   // - |data_size| >= 0
    362   // Postconditions:
    363   // - |*offset| is between 0 and |data_size| included.
    364   //   It is strictly less than |data_size| if |data_size| > 0.
    365   // - |*start_code_size| is either 0, 3 or 4.
    366   static bool FindStartCode(const uint8_t* data,
    367                             off_t data_size,
    368                             off_t* offset,
    369                             off_t* start_code_size);
    370 
    371   // Wrapper for FindStartCode() that skips over start codes that
    372   // may appear inside of |encrypted_ranges_|.
    373   // Returns true if a start code was found. Otherwise returns false.
    374   static bool FindStartCodeInClearRanges(const uint8_t* data,
    375                                          off_t data_size,
    376                                          const Ranges<const uint8_t*>& ranges,
    377                                          off_t* offset,
    378                                          off_t* start_code_size);
    379 
    380   H264Parser();
    381   ~H264Parser();
    382 
    383   void Reset();
    384   // Set current stream pointer to |stream| of |stream_size| in bytes,
    385   // |stream| owned by caller.
    386   // |subsamples| contains information about what parts of |stream| are
    387   // encrypted.
    388   void SetStream(const uint8_t* stream, off_t stream_size);
    389   void SetEncryptedStream(const uint8_t* stream,
    390                           off_t stream_size,
    391                           const std::vector<SubsampleEntry>& subsamples);
    392 
    393   // Read the stream to find the next NALU, identify it and return
    394   // that information in |*nalu|. This advances the stream to the beginning
    395   // of this NALU, but not past it, so subsequent calls to NALU-specific
    396   // parsing functions (ParseSPS, etc.)  will parse this NALU.
    397   // If the caller wishes to skip the current NALU, it can call this function
    398   // again, instead of any NALU-type specific parse functions below.
    399   Result AdvanceToNextNALU(H264NALU* nalu);
    400 
    401   // NALU-specific parsing functions.
    402   // These should be called after AdvanceToNextNALU().
    403 
    404   // SPSes and PPSes are owned by the parser class and the memory for their
    405   // structures is managed here, not by the caller, as they are reused
    406   // across NALUs.
    407   //
    408   // Parse an SPS/PPS NALU and save their data in the parser, returning id
    409   // of the parsed structure in |*pps_id|/|*sps_id|.
    410   // To get a pointer to a given SPS/PPS structure, use GetSPS()/GetPPS(),
    411   // passing the returned |*sps_id|/|*pps_id| as parameter.
    412   // TODO(posciak,fischman): consider replacing returning Result from Parse*()
    413   // methods with a scoped_ptr and adding an AtEOS() function to check for EOS
    414   // if Parse*() return NULL.
    415   Result ParseSPS(int* sps_id);
    416   Result ParsePPS(int* pps_id);
    417 
    418   // Return a pointer to SPS/PPS with given |sps_id|/|pps_id| or NULL if not
    419   // present.
    420   const H264SPS* GetSPS(int sps_id) const;
    421   const H264PPS* GetPPS(int pps_id) const;
    422 
    423   // Slice headers and SEI messages are not used across NALUs by the parser
    424   // and can be discarded after current NALU, so the parser does not store
    425   // them, nor does it manage their memory.
    426   // The caller has to provide and manage it instead.
    427 
    428   // Parse a slice header, returning it in |*shdr|. |*nalu| must be set to
    429   // the NALU returned from AdvanceToNextNALU() and corresponding to |*shdr|.
    430   Result ParseSliceHeader(const H264NALU& nalu, H264SliceHeader* shdr);
    431 
    432   // Parse a SEI message, returning it in |*sei_msg|, provided and managed
    433   // by the caller.
    434   Result ParseSEI(H264SEIMessage* sei_msg);
    435 
    436  private:
    437   // Move the stream pointer to the beginning of the next NALU,
    438   // i.e. pointing at the next start code.
    439   // Return true if a NALU has been found.
    440   // If a NALU is found:
    441   // - its size in bytes is returned in |*nalu_size| and includes
    442   //   the start code as well as the trailing zero bits.
    443   // - the size in bytes of the start code is returned in |*start_code_size|.
    444   bool LocateNALU(off_t* nalu_size, off_t* start_code_size);
    445 
    446   // Exp-Golomb code parsing as specified in chapter 9.1 of the spec.
    447   // Read one unsigned exp-Golomb code from the stream and return in |*val|.
    448   Result ReadUE(int* val);
    449 
    450   // Read one signed exp-Golomb code from the stream and return in |*val|.
    451   Result ReadSE(int* val);
    452 
    453   // Parse scaling lists (see spec).
    454   Result ParseScalingList(int size, int* scaling_list, bool* use_default);
    455   Result ParseSPSScalingLists(H264SPS* sps);
    456   Result ParsePPSScalingLists(const H264SPS& sps, H264PPS* pps);
    457 
    458   // Parse optional VUI parameters in SPS (see spec).
    459   Result ParseVUIParameters(H264SPS* sps);
    460   // Set |hrd_parameters_present| to true only if they are present.
    461   Result ParseAndIgnoreHRDParameters(bool* hrd_parameters_present);
    462 
    463   // Parse reference picture lists' modifications (see spec).
    464   Result ParseRefPicListModifications(H264SliceHeader* shdr);
    465   Result ParseRefPicListModification(int num_ref_idx_active_minus1,
    466                                      H264ModificationOfPicNum* ref_list_mods);
    467 
    468   // Parse prediction weight table (see spec).
    469   Result ParsePredWeightTable(const H264SPS& sps, H264SliceHeader* shdr);
    470 
    471   // Parse weighting factors (see spec).
    472   Result ParseWeightingFactors(int num_ref_idx_active_minus1,
    473                                int chroma_array_type,
    474                                int luma_log2_weight_denom,
    475                                int chroma_log2_weight_denom,
    476                                H264WeightingFactors* w_facts);
    477 
    478   // Parse decoded reference picture marking information (see spec).
    479   Result ParseDecRefPicMarking(H264SliceHeader* shdr);
    480 
    481   // Pointer to the current NALU in the stream.
    482   const uint8_t* stream_;
    483 
    484   // Bytes left in the stream after the current NALU.
    485   off_t bytes_left_;
    486 
    487   H264BitReader br_;
    488 
    489   // PPSes and SPSes stored for future reference.
    490   std::map<int, std::unique_ptr<H264SPS>> active_SPSes_;
    491   std::map<int, std::unique_ptr<H264PPS>> active_PPSes_;
    492 
    493   // Ranges of encrypted bytes in the buffer passed to
    494   // SetEncryptedStream().
    495   Ranges<const uint8_t*> encrypted_ranges_;
    496 
    497   DISALLOW_COPY_AND_ASSIGN(H264Parser);
    498 };
    499 
    500 }  // namespace media
    501 
    502 #endif  // H264_PARSER_H_
    503