FFmpeg: libavcodec/h264.h Source File

00001 /*
00002  * H.26L/H.264/AVC/JVT/14496-10/... encoder/decoder
00003  * Copyright (c) 2003 Michael Niedermayer <michaelni@gmx.at>
00004  *
00005  * This file is part of FFmpeg.
00006  *
00007  * FFmpeg is free software; you can redistribute it and/or
00008  * modify it under the terms of the GNU Lesser General Public
00009  * License as published by the Free Software Foundation; either
00010  * version 2.1 of the License, or (at your option) any later version.
00011  *
00012  * FFmpeg is distributed in the hope that it will be useful,
00013  * but WITHOUT ANY WARRANTY; without even the implied warranty of
00014  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
00015  * Lesser General Public License for more details.
00016  *
00017  * You should have received a copy of the GNU Lesser General Public
00018  * License along with FFmpeg; if not, write to the Free Software
00019  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
00020  */
00021 
00028 #ifndef AVCODEC_H264_H
00029 #define AVCODEC_H264_H
00030 
00031 #include "dsputil.h"
00032 #include "cabac.h"
00033 #include "mpegvideo.h"
00034 #include "h264pred.h"
00035 
00036 #define interlaced_dct interlaced_dct_is_a_bad_name
00037 #define mb_intra mb_intra_is_not_initialized_see_mb_type
00038 
00039 #define LUMA_DC_BLOCK_INDEX   25
00040 #define CHROMA_DC_BLOCK_INDEX 26
00041 
00042 #define CHROMA_DC_COEFF_TOKEN_VLC_BITS 8
00043 #define COEFF_TOKEN_VLC_BITS           8
00044 #define TOTAL_ZEROS_VLC_BITS           9
00045 #define CHROMA_DC_TOTAL_ZEROS_VLC_BITS 3
00046 #define RUN_VLC_BITS                   3
00047 #define RUN7_VLC_BITS                  6
00048 
00049 #define MAX_SPS_COUNT 32
00050 #define MAX_PPS_COUNT 256
00051 
00052 #define MAX_MMCO_COUNT 66
00053 
00054 #define MAX_DELAYED_PIC_COUNT 16
00055 
00056 /* Compiling in interlaced support reduces the speed
00057  * of progressive decoding by about 2%. */
00058 #define ALLOW_INTERLACE
00059 
00060 #define ALLOW_NOCHROMA
00061 
00066 #define MAX_SLICES 16
00067 
00068 #ifdef ALLOW_INTERLACE
00069 #define MB_MBAFF h->mb_mbaff
00070 #define MB_FIELD h->mb_field_decoding_flag
00071 #define FRAME_MBAFF h->mb_aff_frame
00072 #define FIELD_PICTURE (s->picture_structure != PICT_FRAME)
00073 #else
00074 #define MB_MBAFF 0
00075 #define MB_FIELD 0
00076 #define FRAME_MBAFF 0
00077 #define FIELD_PICTURE 0
00078 #undef  IS_INTERLACED
00079 #define IS_INTERLACED(mb_type) 0
00080 #endif
00081 #define FIELD_OR_MBAFF_PICTURE (FRAME_MBAFF || FIELD_PICTURE)
00082 
00083 #ifdef ALLOW_NOCHROMA
00084 #define CHROMA h->sps.chroma_format_idc
00085 #else
00086 #define CHROMA 1
00087 #endif
00088 
00089 #define EXTENDED_SAR          255
00090 
00091 #define MB_TYPE_REF0       MB_TYPE_ACPRED //dirty but it fits in 16 bit
00092 #define MB_TYPE_8x8DCT     0x01000000
00093 #define IS_REF0(a)         ((a) & MB_TYPE_REF0)
00094 #define IS_8x8DCT(a)       ((a) & MB_TYPE_8x8DCT)
00095 
00096 /* NAL unit types */
00097 enum {
00098     NAL_SLICE=1,
00099     NAL_DPA,
00100     NAL_DPB,
00101     NAL_DPC,
00102     NAL_IDR_SLICE,
00103     NAL_SEI,
00104     NAL_SPS,
00105     NAL_PPS,
00106     NAL_AUD,
00107     NAL_END_SEQUENCE,
00108     NAL_END_STREAM,
00109     NAL_FILLER_DATA,
00110     NAL_SPS_EXT,
00111     NAL_AUXILIARY_SLICE=19
00112 };
00113 
00117 typedef enum {
00118     SEI_BUFFERING_PERIOD             =  0, 
00119     SEI_TYPE_PIC_TIMING              =  1, 
00120     SEI_TYPE_USER_DATA_UNREGISTERED  =  5, 
00121     SEI_TYPE_RECOVERY_POINT          =  6  
00122 } SEI_Type;
00123 
00127 typedef enum {
00128     SEI_PIC_STRUCT_FRAME             = 0, 
00129     SEI_PIC_STRUCT_TOP_FIELD         = 1, 
00130     SEI_PIC_STRUCT_BOTTOM_FIELD      = 2, 
00131     SEI_PIC_STRUCT_TOP_BOTTOM        = 3, 
00132     SEI_PIC_STRUCT_BOTTOM_TOP        = 4, 
00133     SEI_PIC_STRUCT_TOP_BOTTOM_TOP    = 5, 
00134     SEI_PIC_STRUCT_BOTTOM_TOP_BOTTOM = 6, 
00135     SEI_PIC_STRUCT_FRAME_DOUBLING    = 7, 
00136     SEI_PIC_STRUCT_FRAME_TRIPLING    = 8  
00137 } SEI_PicStructType;
00138 
00142 typedef struct SPS{
00143 
00144     int profile_idc;
00145     int level_idc;
00146     int chroma_format_idc;
00147     int transform_bypass;              
00148     int log2_max_frame_num;            
00149     int poc_type;                      
00150     int log2_max_poc_lsb;              
00151     int delta_pic_order_always_zero_flag;
00152     int offset_for_non_ref_pic;
00153     int offset_for_top_to_bottom_field;
00154     int poc_cycle_length;              
00155     int ref_frame_count;               
00156     int gaps_in_frame_num_allowed_flag;
00157     int mb_width;                      
00158     int mb_height;                     
00159     int frame_mbs_only_flag;
00160     int mb_aff;                        
00161     int direct_8x8_inference_flag;
00162     int crop;                   
00163     unsigned int crop_left;            
00164     unsigned int crop_right;           
00165     unsigned int crop_top;             
00166     unsigned int crop_bottom;          
00167     int vui_parameters_present_flag;
00168     AVRational sar;
00169     int timing_info_present_flag;
00170     uint32_t num_units_in_tick;
00171     uint32_t time_scale;
00172     int fixed_frame_rate_flag;
00173     short offset_for_ref_frame[256]; //FIXME dyn aloc?
00174     int bitstream_restriction_flag;
00175     int num_reorder_frames;
00176     int scaling_matrix_present;
00177     uint8_t scaling_matrix4[6][16];
00178     uint8_t scaling_matrix8[2][64];
00179     int nal_hrd_parameters_present_flag;
00180     int vcl_hrd_parameters_present_flag;
00181     int pic_struct_present_flag;
00182     int time_offset_length;
00183     int cpb_cnt;                       
00184     int initial_cpb_removal_delay_length; 
00185     int cpb_removal_delay_length;      
00186     int dpb_output_delay_length;       
00187     int bit_depth_luma;                
00188     int bit_depth_chroma;              
00189     int residual_color_transform_flag; 
00190 }SPS;
00191 
00195 typedef struct PPS{
00196     unsigned int sps_id;
00197     int cabac;                  
00198     int pic_order_present;      
00199     int slice_group_count;      
00200     int mb_slice_group_map_type;
00201     unsigned int ref_count[2];  
00202     int weighted_pred;          
00203     int weighted_bipred_idc;
00204     int init_qp;                
00205     int init_qs;                
00206     int chroma_qp_index_offset[2];
00207     int deblocking_filter_parameters_present; 
00208     int constrained_intra_pred; 
00209     int redundant_pic_cnt_present; 
00210     int transform_8x8_mode;     
00211     uint8_t scaling_matrix4[6][16];
00212     uint8_t scaling_matrix8[2][64];
00213     uint8_t chroma_qp_table[2][64];  
00214     int chroma_qp_diff;
00215 }PPS;
00216 
00220 typedef enum MMCOOpcode{
00221     MMCO_END=0,
00222     MMCO_SHORT2UNUSED,
00223     MMCO_LONG2UNUSED,
00224     MMCO_SHORT2LONG,
00225     MMCO_SET_MAX_LONG,
00226     MMCO_RESET,
00227     MMCO_LONG,
00228 } MMCOOpcode;
00229 
00233 typedef struct MMCO{
00234     MMCOOpcode opcode;
00235     int short_pic_num;  
00236     int long_arg;       
00237 } MMCO;
00238 
00242 typedef struct H264Context{
00243     MpegEncContext s;
00244     int nal_ref_idc;
00245     int nal_unit_type;
00246     uint8_t *rbsp_buffer[2];
00247     unsigned int rbsp_buffer_size[2];
00248 
00252     int is_avc; 
00253     int got_avcC; 
00254     int nal_length_size; 
00255 
00256     int chroma_qp[2]; //QPc
00257 
00258     int prev_mb_skipped;
00259     int next_mb_skipped;
00260 
00261     //prediction stuff
00262     int chroma_pred_mode;
00263     int intra16x16_pred_mode;
00264 
00265     int top_mb_xy;
00266     int left_mb_xy[2];
00267 
00268     int8_t intra4x4_pred_mode_cache[5*8];
00269     int8_t (*intra4x4_pred_mode)[8];
00270     H264PredContext hpc;
00271     unsigned int topleft_samples_available;
00272     unsigned int top_samples_available;
00273     unsigned int topright_samples_available;
00274     unsigned int left_samples_available;
00275     uint8_t (*top_borders[2])[16+2*8];
00276     uint8_t left_border[2*(17+2*9)];
00277 
00282     DECLARE_ALIGNED_8(uint8_t, non_zero_count_cache[6*8]);
00283     uint8_t (*non_zero_count)[16];
00284 
00288     DECLARE_ALIGNED_8(int16_t, mv_cache[2][5*8][2]);
00289     DECLARE_ALIGNED_8(int8_t, ref_cache[2][5*8]);
00290 #define LIST_NOT_USED -1 //FIXME rename?
00291 #define PART_NOT_AVAILABLE -2
00292 
00296     int mv_cache_clean[2];
00297 
00301     int neighbor_transform_size;
00302 
00307     int block_offset[2*(16+8)];
00308 
00309     uint32_t *mb2b_xy; //FIXME are these 4 a good idea?
00310     uint32_t *mb2b8_xy;
00311     int b_stride; //FIXME use s->b4_stride
00312     int b8_stride;
00313 
00314     int mb_linesize;   
00315     int mb_uvlinesize;
00316 
00317     int emu_edge_width;
00318     int emu_edge_height;
00319 
00320     int halfpel_flag;
00321     int thirdpel_flag;
00322 
00323     int unknown_svq3_flag;
00324     int next_slice_index;
00325 
00326     SPS *sps_buffers[MAX_SPS_COUNT];
00327     SPS sps; 
00328 
00329     PPS *pps_buffers[MAX_PPS_COUNT];
00333     PPS pps; //FIXME move to Picture perhaps? (->no) do we need that?
00334 
00335     uint32_t dequant4_buffer[6][52][16];
00336     uint32_t dequant8_buffer[2][52][64];
00337     uint32_t (*dequant4_coeff[6])[16];
00338     uint32_t (*dequant8_coeff[2])[64];
00339     int dequant_coeff_pps;     
00340 
00341     int slice_num;
00342     uint16_t *slice_table_base;
00343     uint16_t *slice_table;     
00344     int slice_type;
00345     int slice_type_nos;        
00346     int slice_type_fixed;
00347 
00348     //interlacing specific flags
00349     int mb_aff_frame;
00350     int mb_field_decoding_flag;
00351     int mb_mbaff;              
00352 
00353     DECLARE_ALIGNED_8(uint16_t, sub_mb_type[4]);
00354 
00355     //POC stuff
00356     int poc_lsb;
00357     int poc_msb;
00358     int delta_poc_bottom;
00359     int delta_poc[2];
00360     int frame_num;
00361     int prev_poc_msb;             
00362     int prev_poc_lsb;             
00363     int frame_num_offset;         
00364     int prev_frame_num_offset;    
00365     int prev_frame_num;           
00366 
00370     int curr_pic_num;
00371 
00375     int max_pic_num;
00376 
00377     //Weighted pred stuff
00378     int use_weight;
00379     int use_weight_chroma;
00380     int luma_log2_weight_denom;
00381     int chroma_log2_weight_denom;
00382     int luma_weight[2][48];
00383     int luma_offset[2][48];
00384     int chroma_weight[2][48][2];
00385     int chroma_offset[2][48][2];
00386     int implicit_weight[48][48];
00387 
00388     //deblock
00389     int deblocking_filter;         
00390     int slice_alpha_c0_offset;
00391     int slice_beta_offset;
00392 
00393     int redundant_pic_count;
00394 
00395     int direct_spatial_mv_pred;
00396     int dist_scale_factor[16];
00397     int dist_scale_factor_field[2][32];
00398     int map_col_to_list0[2][16+32];
00399     int map_col_to_list0_field[2][2][16+32];
00400 
00404     unsigned int ref_count[2];   
00405     unsigned int list_count;
00406     Picture *short_ref[32];
00407     Picture *long_ref[32];
00408     Picture default_ref_list[2][32]; 
00409     Picture ref_list[2][48];         
00412     int ref2frm[MAX_SLICES][2][64];  
00413     Picture *delayed_pic[MAX_DELAYED_PIC_COUNT+2]; //FIXME size?
00414     int outputed_poc;
00415 
00419     MMCO mmco[MAX_MMCO_COUNT];
00420     int mmco_index;
00421 
00422     int long_ref_count;  
00423     int short_ref_count; 
00424 
00425     //data partitioning
00426     GetBitContext intra_gb;
00427     GetBitContext inter_gb;
00428     GetBitContext *intra_gb_ptr;
00429     GetBitContext *inter_gb_ptr;
00430 
00431     DECLARE_ALIGNED_16(DCTELEM, mb[16*24]);
00432     DCTELEM mb_padding[256];        
00433 
00437     CABACContext cabac;
00438     uint8_t      cabac_state[460];
00439     int          cabac_init_idc;
00440 
00441     /* 0x100 -> non null luma_dc, 0x80/0x40 -> non null chroma_dc (cb/cr), 0x?0 -> chroma_cbp(0,1,2), 0x0? luma_cbp */
00442     uint16_t     *cbp_table;
00443     int cbp;
00444     int top_cbp;
00445     int left_cbp;
00446     /* chroma_pred_mode for i4x4 or i16x16, else 0 */
00447     uint8_t     *chroma_pred_mode_table;
00448     int         last_qscale_diff;
00449     int16_t     (*mvd_table[2])[2];
00450     DECLARE_ALIGNED_8(int16_t, mvd_cache[2][5*8][2]);
00451     uint8_t     *direct_table;
00452     uint8_t     direct_cache[5*8];
00453 
00454     uint8_t zigzag_scan[16];
00455     uint8_t zigzag_scan8x8[64];
00456     uint8_t zigzag_scan8x8_cavlc[64];
00457     uint8_t field_scan[16];
00458     uint8_t field_scan8x8[64];
00459     uint8_t field_scan8x8_cavlc[64];
00460     const uint8_t *zigzag_scan_q0;
00461     const uint8_t *zigzag_scan8x8_q0;
00462     const uint8_t *zigzag_scan8x8_cavlc_q0;
00463     const uint8_t *field_scan_q0;
00464     const uint8_t *field_scan8x8_q0;
00465     const uint8_t *field_scan8x8_cavlc_q0;
00466 
00467     int x264_build;
00468 
00473     struct H264Context *thread_context[MAX_THREADS];
00474 
00478     int current_slice;
00479 
00486     int max_contexts;
00487 
00492     int single_decode_warning;
00493 
00494     int last_slice_type;
00497     int mb_xy;
00498 
00499     uint32_t svq3_watermark_key;
00500 
00504     SEI_PicStructType sei_pic_struct;
00505 
00509     int sei_dpb_output_delay;
00510 
00514     int sei_cpb_removal_delay;
00515 
00523     int sei_recovery_frame_cnt;
00524 
00525     int is_complex;
00526 
00527     int luma_weight_flag[2];   
00528     int chroma_weight_flag[2]; 
00529 
00530     // Timestamp stuff
00531     int sei_buffering_period_present;  
00532     int initial_cpb_removal_delay[32]; 
00533 }H264Context;
00534 
00538 int ff_h264_decode_sei(H264Context *h);
00539 
00543 int ff_h264_decode_seq_parameter_set(H264Context *h);
00544 
00548 int ff_h264_decode_picture_parameter_set(H264Context *h, int bit_length);
00549 
00557 const uint8_t *ff_h264_decode_nal(H264Context *h, const uint8_t *src, int *dst_length, int *consumed, int length);
00558 
00563 int ff_h264_decode_rbsp_trailing(H264Context *h, const uint8_t *src);
00564 
00565 #endif /* AVCODEC_H264_H */