libavcodec/vc1.h

   1 /*
   2  * VC-1 and WMV3 decoder
   3  * Copyright (c) 2006-2007 Konstantin Shishkov
   4  * Partly based on vc9.c (c) 2005 Anonymous, Alex Beregszaszi, Michael Niedermayer
   5  *
   6  * This file is part of Libav.
   7  *
   8  * Libav is free software; you can redistribute it and/or
   9  * modify it under the terms of the GNU Lesser General Public
  10  * License as published by the Free Software Foundation; either
  11  * version 2.1 of the License, or (at your option) any later version.
  12  *
  13  * Libav is distributed in the hope that it will be useful,
  14  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  15  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  16  * Lesser General Public License for more details.
  17  *
  18  * You should have received a copy of the GNU Lesser General Public
  19  * License along with Libav; if not, write to the Free Software
  20  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  21  */
  22
  23 #ifndef AVCODEC_VC1_H
  24 #define AVCODEC_VC1_H
  25
  26 #include "avcodec.h"
  27 #include "h264chroma.h"
  28 #include "mpegvideo.h"
  29 #include "intrax8.h"
  30 #include "vc1dsp.h"
  31
  32 #define AC_VLC_BITS 9
  33
  34 /** Markers used in VC-1 AP frame data */
  35 //@{
  36 enum VC1Code {
  37     VC1_CODE_RES0       = 0x00000100,
  38     VC1_CODE_ENDOFSEQ   = 0x0000010A,
  39     VC1_CODE_SLICE,
  40     VC1_CODE_FIELD,
  41     VC1_CODE_FRAME,
  42     VC1_CODE_ENTRYPOINT,
  43     VC1_CODE_SEQHDR,
  44 };
  45 //@}
  46
  47 #define IS_MARKER(x) (((x) & ~0xFF) == VC1_CODE_RES0)
  48
  49 /** Available Profiles */
  50 //@{
  51 enum Profile {
  52     PROFILE_SIMPLE,
  53     PROFILE_MAIN,
  54     PROFILE_COMPLEX, ///< TODO: WMV9 specific
  55     PROFILE_ADVANCED
  56 };
  57 //@}
  58
  59 /** Sequence quantizer mode */
  60 //@{
  61 enum QuantMode {
  62     QUANT_FRAME_IMPLICIT,    ///< Implicitly specified at frame level
  63     QUANT_FRAME_EXPLICIT,    ///< Explicitly specified at frame level
  64     QUANT_NON_UNIFORM,       ///< Non-uniform quant used for all frames
  65     QUANT_UNIFORM            ///< Uniform quant used for all frames
  66 };
  67 //@}
  68
  69 /** Where quant can be changed */
  70 //@{
  71 enum DQProfile {
  72     DQPROFILE_FOUR_EDGES,
  73     DQPROFILE_DOUBLE_EDGES,
  74     DQPROFILE_SINGLE_EDGE,
  75     DQPROFILE_ALL_MBS
  76 };
  77 //@}
  78
  79 /** @name Where quant can be changed
  80  */
  81 //@{
  82 enum DQSingleEdge {
  83     DQSINGLE_BEDGE_LEFT,
  84     DQSINGLE_BEDGE_TOP,
  85     DQSINGLE_BEDGE_RIGHT,
  86     DQSINGLE_BEDGE_BOTTOM
  87 };
  88 //@}
  89
  90 /** Which pair of edges is quantized with ALTPQUANT */
  91 //@{
  92 enum DQDoubleEdge {
  93     DQDOUBLE_BEDGE_TOPLEFT,
  94     DQDOUBLE_BEDGE_TOPRIGHT,
  95     DQDOUBLE_BEDGE_BOTTOMRIGHT,
  96     DQDOUBLE_BEDGE_BOTTOMLEFT
  97 };
  98 //@}
  99
 100 /** MV modes for P frames */
 101 //@{
 102 enum MVModes {
 103     MV_PMODE_1MV_HPEL_BILIN,
 104     MV_PMODE_1MV,
 105     MV_PMODE_1MV_HPEL,
 106     MV_PMODE_MIXED_MV,
 107     MV_PMODE_INTENSITY_COMP
 108 };
 109 //@}
 110
 111 /** MBMODE for interlaced frame P-picture */
 112 //@{
 113 enum MBModesIntfr {
 114     MV_PMODE_INTFR_1MV,
 115     MV_PMODE_INTFR_2MV_FIELD,
 116     MV_PMODE_INTFR_2MV,
 117     MV_PMODE_INTFR_4MV_FIELD,
 118     MV_PMODE_INTFR_4MV,
 119     MV_PMODE_INTFR_INTRA,
 120 };
 121 //@}
 122
 123 /** @name MV types for B frames */
 124 //@{
 125 enum BMVTypes {
 126     BMV_TYPE_BACKWARD,
 127     BMV_TYPE_FORWARD,
 128     BMV_TYPE_INTERPOLATED,
 129     BMV_TYPE_DIRECT
 130 };
 131 //@}
 132
 133 /** @name Block types for P/B frames */
 134 //@{
 135 enum TransformTypes {
 136     TT_8X8,
 137     TT_8X4_BOTTOM,
 138     TT_8X4_TOP,
 139     TT_8X4,         // both halves
 140     TT_4X8_RIGHT,
 141     TT_4X8_LEFT,
 142     TT_4X8,         // both halves
 143     TT_4X4
 144 };
 145 //@}
 146
 147 enum CodingSet {
 148     CS_HIGH_MOT_INTRA = 0,
 149     CS_HIGH_MOT_INTER,
 150     CS_LOW_MOT_INTRA,
 151     CS_LOW_MOT_INTER,
 152     CS_MID_RATE_INTRA,
 153     CS_MID_RATE_INTER,
 154     CS_HIGH_RATE_INTRA,
 155     CS_HIGH_RATE_INTER
 156 };
 157
 158 /** @name Overlap conditions for Advanced Profile */
 159 //@{
 160 enum COTypes {
 161     CONDOVER_NONE = 0,
 162     CONDOVER_ALL,
 163     CONDOVER_SELECT
 164 };
 165 //@}
 166
 167 /**
 168  * FCM Frame Coding Mode
 169  * @note some content might be marked interlaced
 170  * but have fcm set to 0 as well (e.g. HD-DVD)
 171  */
 172 enum FrameCodingMode {
 173     PROGRESSIVE = 0,    ///<  in the bitstream is reported as 00b
 174     ILACE_FRAME,        ///<  in the bitstream is reported as 10b
 175     ILACE_FIELD         ///<  in the bitstream is reported as 11b
 176 };
 177
 178 /** The VC1 Context
 179  * @todo Change size wherever another size is more efficient
 180  * Many members are only used for Advanced Profile
 181  */
 182 typedef struct VC1Context{
 183     MpegEncContext s;
 184     IntraX8Context x8;
 185     H264ChromaContext h264chroma;
 186     VC1DSPContext vc1dsp;
 187
 188     int bits;
 189
 190     /** Simple/Main Profile sequence header */
 191     //@{
 192     int res_sprite;       ///< reserved, sprite mode
 193     int res_y411;         ///< reserved, old interlaced mode
 194     int res_x8;           ///< reserved
 195     int multires;         ///< frame-level RESPIC syntax element present
 196     int res_fasttx;       ///< reserved, always 1
 197     int res_transtab;     ///< reserved, always 0
 198     int rangered;         ///< RANGEREDFRM (range reduction) syntax element present
 199                           ///< at frame level
 200     int res_rtm_flag;     ///< reserved, set to 1
 201     int reserved;         ///< reserved
 202     //@}
 203
 204     /** Advanced Profile */
 205     //@{
 206     int level;            ///< 3bits, for Advanced/Simple Profile, provided by TS layer
 207     int chromaformat;     ///< 2bits, 2=4:2:0, only defined
 208     int postprocflag;     ///< Per-frame processing suggestion flag present
 209     int broadcast;        ///< TFF/RFF present
 210     int interlace;        ///< Progressive/interlaced (RPTFTM syntax element)
 211     int tfcntrflag;       ///< TFCNTR present
 212     int panscanflag;      ///< NUMPANSCANWIN, TOPLEFT{X,Y}, BOTRIGHT{X,Y} present
 213     int refdist_flag;     ///< REFDIST syntax element present in II, IP, PI or PP field picture headers
 214     int extended_dmv;     ///< Additional extended dmv range at P/B frame-level
 215     int color_prim;       ///< 8bits, chroma coordinates of the color primaries
 216     int transfer_char;    ///< 8bits, Opto-electronic transfer characteristics
 217     int matrix_coef;      ///< 8bits, Color primaries->YCbCr transform matrix
 218     int hrd_param_flag;   ///< Presence of Hypothetical Reference
 219                           ///< Decoder parameters
 220     int psf;              ///< Progressive Segmented Frame
 221     //@}
 222
 223     /** Sequence header data for all Profiles
 224      * TODO: choose between ints, uint8_ts and monobit flags
 225      */
 226     //@{
 227     int profile;          ///< 2bits, Profile
 228     int frmrtq_postproc;  ///< 3bits,
 229     int bitrtq_postproc;  ///< 5bits, quantized framerate-based postprocessing strength
 230     int fastuvmc;         ///< Rounding of qpel vector to hpel ? (not in Simple)
 231     int extended_mv;      ///< Ext MV in P/B (not in Simple)
 232     int dquant;           ///< How qscale varies with MBs, 2bits (not in Simple)
 233     int vstransform;      ///< variable-size [48]x[48] transform type + info
 234     int overlap;          ///< overlapped transforms in use
 235     int quantizer_mode;   ///< 2bits, quantizer mode used for sequence, see QUANT_*
 236     int finterpflag;      ///< INTERPFRM present
 237     //@}
 238
 239     /** Frame decoding info for all profiles */
 240     //@{
 241     uint8_t mv_mode;             ///< MV coding monde
 242     uint8_t mv_mode2;            ///< Secondary MV coding mode (B frames)
 243     int k_x;                     ///< Number of bits for MVs (depends on MV range)
 244     int k_y;                     ///< Number of bits for MVs (depends on MV range)
 245     int range_x, range_y;        ///< MV range
 246     uint8_t pq, altpq;           ///< Current/alternate frame quantizer scale
 247     uint8_t zz_8x8[4][64];       ///< Zigzag table for TT_8x8, permuted for IDCT
 248     int left_blk_sh, top_blk_sh; ///< Either 3 or 0, positions of l/t in blk[]
 249     const uint8_t* zz_8x4;       ///< Zigzag scan table for TT_8x4 coding mode
 250     const uint8_t* zz_4x8;       ///< Zigzag scan table for TT_4x8 coding mode
 251     /** pquant parameters */
 252     //@{
 253     uint8_t dquantfrm;
 254     uint8_t dqprofile;
 255     uint8_t dqsbedge;
 256     uint8_t dqbilevel;
 257     //@}
 258     /** AC coding set indexes
 259      * @see 8.1.1.10, p(1)10
 260      */
 261     //@{
 262     int c_ac_table_index;    ///< Chroma index from ACFRM element
 263     int y_ac_table_index;    ///< Luma index from AC2FRM element
 264     //@}
 265     int ttfrm;               ///< Transform type info present at frame level
 266     uint8_t ttmbf;           ///< Transform type flag
 267     int *ttblk_base, *ttblk; ///< Transform type at the block level
 268     int codingset;           ///< index of current table set from 11.8 to use for luma block decoding
 269     int codingset2;          ///< index of current table set from 11.8 to use for chroma block decoding
 270     int pqindex;             ///< raw pqindex used in coding set selection
 271     int a_avail, c_avail;
 272     uint8_t *mb_type_base, *mb_type[3];
 273
 274
 275     /** Luma compensation parameters */
 276     //@{
 277     uint8_t lumscale;
 278     uint8_t lumshift;
 279     //@}
 280     int16_t bfraction;    ///< Relative position % anchors=> how to scale MVs
 281     uint8_t halfpq;       ///< Uniform quant over image and qp+.5
 282     uint8_t respic;       ///< Frame-level flag for resized images
 283     int buffer_fullness;  ///< HRD info
 284     /** Ranges:
 285      * -# 0 -> [-64n 63.f] x [-32, 31.f]
 286      * -# 1 -> [-128, 127.f] x [-64, 63.f]
 287      * -# 2 -> [-512, 511.f] x [-128, 127.f]
 288      * -# 3 -> [-1024, 1023.f] x [-256, 255.f]
 289      */
 290     uint8_t mvrange;                ///< Extended MV range flag
 291     uint8_t pquantizer;             ///< Uniform (over sequence) quantizer in use
 292     VLC *cbpcy_vlc;                 ///< CBPCY VLC table
 293     int tt_index;                   ///< Index for Transform Type tables (to decode TTMB)
 294     uint8_t* mv_type_mb_plane;      ///< bitplane for mv_type == (4MV)
 295     uint8_t* direct_mb_plane;       ///< bitplane for "direct" MBs
 296     uint8_t* forward_mb_plane;      ///< bitplane for "forward" MBs
 297     int mv_type_is_raw;             ///< mv type mb plane is not coded
 298     int dmb_is_raw;                 ///< direct mb plane is raw
 299     int fmb_is_raw;                 ///< forward mb plane is raw
 300     int skip_is_raw;                ///< skip mb plane is not coded
 301     uint8_t luty[256], lutuv[256];  ///< lookup tables used for intensity compensation
 302     int use_ic;                     ///< use intensity compensation in B-frames
 303     int rnd;                        ///< rounding control
 304
 305     /** Frame decoding info for S/M profiles only */
 306     //@{
 307     uint8_t rangeredfrm;            ///< out_sample = CLIP((in_sample-128)*2+128)
 308     uint8_t interpfrm;
 309     //@}
 310
 311     /** Frame decoding info for Advanced profile */
 312     //@{
 313     enum FrameCodingMode fcm;
 314     uint8_t numpanscanwin;
 315     uint8_t tfcntr;
 316     uint8_t rptfrm, tff, rff;
 317     uint16_t topleftx;
 318     uint16_t toplefty;
 319     uint16_t bottomrightx;
 320     uint16_t bottomrighty;
 321     uint8_t uvsamp;
 322     uint8_t postproc;
 323     int hrd_num_leaky_buckets;
 324     uint8_t bit_rate_exponent;
 325     uint8_t buffer_size_exponent;
 326     uint8_t* acpred_plane;       ///< AC prediction flags bitplane
 327     int acpred_is_raw;
 328     uint8_t* over_flags_plane;   ///< Overflags bitplane
 329     int overflg_is_raw;
 330     uint8_t condover;
 331     uint16_t *hrd_rate, *hrd_buffer;
 332     uint8_t *hrd_fullness;
 333     uint8_t range_mapy_flag;
 334     uint8_t range_mapuv_flag;
 335     uint8_t range_mapy;
 336     uint8_t range_mapuv;
 337     //@}
 338
 339     /** Frame decoding info for interlaced picture */
 340     uint8_t dmvrange;   ///< Extended differential MV range flag
 341     int fourmvswitch;
 342     int intcomp;
 343     uint8_t lumscale2;  ///< for interlaced field P picture
 344     uint8_t lumshift2;
 345     uint8_t luty2[256], lutuv2[256]; // lookup tables used for intensity compensation
 346     VLC* mbmode_vlc;
 347     VLC* imv_vlc;
 348     VLC* twomvbp_vlc;
 349     VLC* fourmvbp_vlc;
 350     uint8_t twomvbp;
 351     uint8_t fourmvbp;
 352     uint8_t* fieldtx_plane;
 353     int fieldtx_is_raw;
 354     int8_t zzi_8x8[64];
 355     uint8_t *blk_mv_type_base, *blk_mv_type;    ///< 0: frame MV, 1: field MV (interlaced frame)
 356     uint8_t *mv_f_base, *mv_f[2];               ///< 0: MV obtained from same field, 1: opposite field
 357     uint8_t *mv_f_last_base, *mv_f_last[2];
 358     uint8_t *mv_f_next_base, *mv_f_next[2];
 359     int field_mode;         ///< 1 for interlaced field pictures
 360     int fptype;
 361     int second_field;
 362     int refdist;            ///< distance of the current picture from reference
 363     int numref;             ///< number of past field pictures used as reference
 364                             // 0 corresponds to 1 and 1 corresponds to 2 references
 365     int reffield;           ///< if numref = 0 (1 reference) then reffield decides which
 366                             // field to use among the two fields from previous frame
 367     int intcompfield;       ///< which of the two fields to be intensity compensated
 368                             // 0: both fields, 1: bottom field, 2: top field
 369     int cur_field_type;     ///< 0: top, 1: bottom
 370     int ref_field_type[2];  ///< forward and backward reference field type (top or bottom)
 371     int blocks_off, mb_off;
 372     int qs_last;            ///< if qpel has been used in the previous (tr.) picture
 373     int bmvtype;
 374     int frfd, brfd;         ///< reference frame distance (forward or backward)
 375     int pic_header_flag;
 376
 377     /** Frame decoding info for sprite modes */
 378     //@{
 379     int new_sprite;
 380     int two_sprites;
 381     AVFrame sprite_output_frame;
 382     int output_width, output_height, sprite_width, sprite_height;
 383     uint8_t* sr_rows[2][2];      ///< Sprite resizer line cache
 384     //@}
 385
 386     int p_frame_skipped;
 387     int bi_type;
 388     int x8_type;
 389
 390     int16_t (*block)[6][64];
 391     int n_allocated_blks, cur_blk_idx, left_blk_idx, topleft_blk_idx, top_blk_idx;
 392     uint32_t *cbp_base, *cbp;
 393     uint8_t *is_intra_base, *is_intra;
 394     int16_t (*luma_mv_base)[2], (*luma_mv)[2];
 395     uint8_t bfraction_lut_index; ///< Index for BFRACTION value (see Table 40, reproduced into ff_vc1_bfraction_lut[])
 396     uint8_t broken_link;         ///< Broken link flag (BROKEN_LINK syntax element)
 397     uint8_t closed_entry;        ///< Closed entry point flag (CLOSED_ENTRY syntax element)
 398
 399     int end_mb_x;                ///< Horizontal macroblock limit (used only by mss2)
 400
 401     int parse_only;              ///< Context is used within parser
 402
 403     int warn_interlaced;
 404 } VC1Context;
 405
 406 /** Find VC-1 marker in buffer
 407  * @return position where next marker starts or end of buffer if no marker found
 408  */
 409 static av_always_inline const uint8_t* find_next_marker(const uint8_t *src, const uint8_t *end)
 410 {
 411     uint32_t mrk = 0xFFFFFFFF;
 412
 413     if (end-src < 4)
 414         return end;
 415     while (src < end) {
 416         mrk = (mrk << 8) | *src++;
 417         if (IS_MARKER(mrk))
 418             return src - 4;
 419     }
 420     return end;
 421 }
 422
 423 static av_always_inline int vc1_unescape_buffer(const uint8_t *src, int size, uint8_t *dst)
 424 {
 425     int dsize = 0, i;
 426
 427     if (size < 4) {
 428         for (dsize = 0; dsize < size; dsize++)
 429             *dst++ = *src++;
 430         return size;
 431     }
 432     for (i = 0; i < size; i++, src++) {
 433         if (src[0] == 3 && i >= 2 && !src[-1] && !src[-2] && i < size-1 && src[1] < 4) {
 434             dst[dsize++] = src[1];
 435             src++;
 436             i++;
 437         } else
 438             dst[dsize++] = *src;
 439     }
 440     return dsize;
 441 }
 442
 443 /**
 444  * Decode Simple/Main Profiles sequence header
 445  * @see Figure 7-8, p16-17
 446  * @param avctx Codec context
 447  * @param gb GetBit context initialized from Codec context extra_data
 448  * @return Status
 449  */
 450 int ff_vc1_decode_sequence_header(AVCodecContext *avctx, VC1Context *v, GetBitContext *gb);
 451
 452 int ff_vc1_decode_entry_point(AVCodecContext *avctx, VC1Context *v, GetBitContext *gb);
 453
 454 int ff_vc1_parse_frame_header    (VC1Context *v, GetBitContext *gb);
 455 int ff_vc1_parse_frame_header_adv(VC1Context *v, GetBitContext *gb);
 456 int ff_vc1_init_common(VC1Context *v);
 457
 458 av_cold int  ff_vc1_decode_init_alloc_tables(VC1Context *v);
 459 av_cold void ff_vc1_init_transposed_scantables(VC1Context *v);
 460 av_cold int  ff_vc1_decode_end(AVCodecContext *avctx);
 461 void ff_vc1_decode_blocks(VC1Context *v);
 462
 463 #endif /* AVCODEC_VC1_H */