You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

899 lines
37KB

  1. /*
  2. * Generic DCT based hybrid video encoder
  3. * Copyright (c) 2000, 2001, 2002 Fabrice Bellard.
  4. *
  5. * This library is free software; you can redistribute it and/or
  6. * modify it under the terms of the GNU Lesser General Public
  7. * License as published by the Free Software Foundation; either
  8. * version 2 of the License, or (at your option) any later version.
  9. *
  10. * This library is distributed in the hope that it will be useful,
  11. * but WITHOUT ANY WARRANTY; without even the implied warranty of
  12. * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
  13. * Lesser General Public License for more details.
  14. *
  15. * You should have received a copy of the GNU Lesser General Public
  16. * License along with this library; if not, write to the Free Software
  17. * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
  18. */
  19. /**
  20. * @file mpegvideo.h
  21. * mpegvideo header.
  22. */
  23. #ifndef AVCODEC_MPEGVIDEO_H
  24. #define AVCODEC_MPEGVIDEO_H
  25. #include "dsputil.h"
  26. #define FRAME_SKIPED 100 ///< return value for header parsers if frame is not coded
  27. enum OutputFormat {
  28. FMT_MPEG1,
  29. FMT_H263,
  30. FMT_MJPEG,
  31. FMT_H264,
  32. };
  33. #define EDGE_WIDTH 16
  34. #define MPEG_BUF_SIZE (16 * 1024)
  35. #define QMAT_SHIFT_MMX 16
  36. #define QMAT_SHIFT 22
  37. #define MAX_FCODE 7
  38. #define MAX_MV 2048
  39. #define MAX_PICTURE_COUNT 15
  40. #define ME_MAP_SIZE 64
  41. #define ME_MAP_SHIFT 3
  42. #define ME_MAP_MV_BITS 11
  43. /* run length table */
  44. #define MAX_RUN 64
  45. #define MAX_LEVEL 64
  46. #define I_TYPE FF_I_TYPE ///< Intra
  47. #define P_TYPE FF_P_TYPE ///< Predicted
  48. #define B_TYPE FF_B_TYPE ///< Bi-dir predicted
  49. #define S_TYPE FF_S_TYPE ///< S(GMC)-VOP MPEG4
  50. #define SI_TYPE FF_SI_TYPE ///< Switching Intra
  51. #define SP_TYPE FF_SP_TYPE ///< Switching Predicted
  52. typedef struct Predictor{
  53. double coeff;
  54. double count;
  55. double decay;
  56. } Predictor;
  57. typedef struct RateControlEntry{
  58. int pict_type;
  59. float qscale;
  60. int mv_bits;
  61. int i_tex_bits;
  62. int p_tex_bits;
  63. int misc_bits;
  64. uint64_t expected_bits;
  65. int new_pict_type;
  66. float new_qscale;
  67. int mc_mb_var_sum;
  68. int mb_var_sum;
  69. int i_count;
  70. int f_code;
  71. int b_code;
  72. }RateControlEntry;
  73. /**
  74. * rate control context.
  75. */
  76. typedef struct RateControlContext{
  77. FILE *stats_file;
  78. int num_entries; ///< number of RateControlEntries
  79. RateControlEntry *entry;
  80. int buffer_index; ///< amount of bits in the video/audio buffer
  81. Predictor pred[5];
  82. double short_term_qsum; ///< sum of recent qscales
  83. double short_term_qcount; ///< count of recent qscales
  84. double pass1_rc_eq_output_sum;///< sum of the output of the rc equation, this is used for normalization
  85. double pass1_wanted_bits; ///< bits which should have been outputed by the pass1 code (including complexity init)
  86. double last_qscale;
  87. double last_qscale_for[5]; ///< last qscale for a specific pict type, used for max_diff & ipb factor stuff
  88. int last_mc_mb_var_sum;
  89. int last_mb_var_sum;
  90. uint64_t i_cplx_sum[5];
  91. uint64_t p_cplx_sum[5];
  92. uint64_t mv_bits_sum[5];
  93. uint64_t qscale_sum[5];
  94. int frame_count[5];
  95. int last_non_b_pict_type;
  96. }RateControlContext;
  97. /**
  98. * Scantable.
  99. */
  100. typedef struct ScanTable{
  101. const uint8_t *scantable;
  102. uint8_t permutated[64];
  103. uint8_t raster_end[64];
  104. #ifdef ARCH_POWERPC
  105. /** Used by dct_quantise_alitvec to find last-non-zero */
  106. uint8_t __align8 inverse[64];
  107. #endif
  108. } ScanTable;
  109. /**
  110. * Picture.
  111. */
  112. typedef struct Picture{
  113. FF_COMMON_FRAME
  114. /**
  115. * halfpel luma planes.
  116. */
  117. uint8_t *interpolated[3];
  118. int16_t (*motion_val[2])[2];
  119. int8_t *ref_index[2];
  120. uint32_t *mb_type_base;
  121. uint32_t *mb_type; ///< mb_type_base + mb_width + 2, note: only used for decoding currently
  122. #define MB_TYPE_INTRA4x4 0x0001
  123. #define MB_TYPE_INTRA16x16 0x0002 //FIXME h264 specific
  124. #define MB_TYPE_INTRA_PCM 0x0004 //FIXME h264 specific
  125. #define MB_TYPE_16x16 0x0008
  126. #define MB_TYPE_16x8 0x0010
  127. #define MB_TYPE_8x16 0x0020
  128. #define MB_TYPE_8x8 0x0040
  129. #define MB_TYPE_INTERLACED 0x0080
  130. #define MB_TYPE_DIRECT2 0x0100 //FIXME
  131. #define MB_TYPE_ACPRED 0x0200
  132. #define MB_TYPE_GMC 0x0400 //FIXME mpeg4 specific
  133. #define MB_TYPE_SKIP 0x0800
  134. #define MB_TYPE_P0L0 0x1000
  135. #define MB_TYPE_P1L0 0x2000
  136. #define MB_TYPE_P0L1 0x4000
  137. #define MB_TYPE_P1L1 0x8000
  138. #define MB_TYPE_L0 (MB_TYPE_P0L0 | MB_TYPE_P1L0)
  139. #define MB_TYPE_L1 (MB_TYPE_P0L1 | MB_TYPE_P1L1)
  140. #define MB_TYPE_L0L1 (MB_TYPE_L0 | MB_TYPE_L1)
  141. #define MB_TYPE_QUANT 0x00010000
  142. //Note bits 24-31 are reserved for codec specific use (h264 ref0, mpeg1 pat, ...)
  143. #define IS_INTRA4x4(a) ((a)&MB_TYPE_INTRA4x4)
  144. #define IS_INTRA16x16(a) ((a)&MB_TYPE_INTRA16x16)
  145. #define IS_PCM(a) ((a)&MB_TYPE_INTRA_PCM)
  146. #define IS_INTRA(a) ((a)&7)
  147. #define IS_INTER(a) ((a)&(MB_TYPE_16x16|MB_TYPE_16x8|MB_TYPE_8x16|MB_TYPE_8x8))
  148. #define IS_SKIP(a) ((a)&MB_TYPE_SKIP)
  149. #define IS_INTRA_PCM(a) ((a)&MB_TYPE_INTRA_PCM)
  150. #define IS_INTERLACED(a) ((a)&MB_TYPE_INTERLACED)
  151. #define IS_DIRECT(a) ((a)&MB_TYPE_DIRECT2)
  152. #define IS_GMC(a) ((a)&MB_TYPE_GMC)
  153. #define IS_16X16(a) ((a)&MB_TYPE_16x16)
  154. #define IS_16X8(a) ((a)&MB_TYPE_16x8)
  155. #define IS_8X16(a) ((a)&MB_TYPE_8x16)
  156. #define IS_8X8(a) ((a)&MB_TYPE_8x8)
  157. #define IS_SUB_8X8(a) ((a)&MB_TYPE_16x16) //note reused
  158. #define IS_SUB_8X4(a) ((a)&MB_TYPE_16x8) //note reused
  159. #define IS_SUB_4X8(a) ((a)&MB_TYPE_8x16) //note reused
  160. #define IS_SUB_4X4(a) ((a)&MB_TYPE_8x8) //note reused
  161. #define IS_ACPRED(a) ((a)&MB_TYPE_ACPRED)
  162. #define IS_QUANT(a) ((a)&MB_TYPE_QUANT)
  163. #define IS_DIR(a, part, list) ((a) & (MB_TYPE_P0L0<<((part)+2*(list))))
  164. #define USES_LIST(a, list) ((a) & ((MB_TYPE_P0L0|MB_TYPE_P1L0)<<(2*(list)))) ///< does this mb use listX, note doesnt work if subMBs
  165. int field_poc[2]; ///< h264 top/bottom POC
  166. int poc; ///< h264 frame POC
  167. int frame_num; ///< h264 frame_num
  168. int pic_id; ///< h264 pic_num or long_term_pic_idx
  169. int long_ref; ///< 1->long term reference 0->short term reference
  170. int mb_var_sum; ///< sum of MB variance for current frame
  171. int mc_mb_var_sum; ///< motion compensated MB variance for current frame
  172. uint16_t *mb_var; ///< Table for MB variances
  173. uint16_t *mc_mb_var; ///< Table for motion compensated MB variances
  174. uint8_t *mb_mean; ///< Table for MB luminance
  175. int32_t *mb_cmp_score; ///< Table for MB cmp scores, for mb decission
  176. int b_frame_score; /* */
  177. } Picture;
  178. typedef struct ParseContext{
  179. uint8_t *buffer;
  180. int index;
  181. int last_index;
  182. int buffer_size;
  183. int state;
  184. int frame_start_found;
  185. } ParseContext;
  186. struct MpegEncContext;
  187. /**
  188. * Motion estimation context.
  189. */
  190. typedef struct MotionEstContext{
  191. int skip; ///< set if ME is skiped for the current MB
  192. int co_located_mv[4][2]; ///< mv from last p frame for direct mode ME
  193. int direct_basis_mv[4][2];
  194. uint8_t *scratchpad; ///< data area for the me algo, so that the ME doesnt need to malloc/free
  195. uint32_t *map; ///< map to avoid duplicate evaluations
  196. uint32_t *score_map; ///< map to store the scores
  197. int map_generation;
  198. int pre_penalty_factor;
  199. int penalty_factor;
  200. int sub_penalty_factor;
  201. int mb_penalty_factor;
  202. int pre_pass; ///< = 1 for the pre pass
  203. int dia_size;
  204. uint8_t (*mv_penalty)[MAX_MV*2+1]; ///< amount of bits needed to encode a MV
  205. int (*sub_motion_search)(struct MpegEncContext * s,
  206. int *mx_ptr, int *my_ptr, int dmin,
  207. int xmin, int ymin, int xmax, int ymax,
  208. int pred_x, int pred_y, Picture *ref_picture,
  209. int n, int size, uint8_t * const mv_penalty);
  210. int (*motion_search[7])(struct MpegEncContext * s, int block,
  211. int *mx_ptr, int *my_ptr,
  212. int P[10][2], int pred_x, int pred_y,
  213. int xmin, int ymin, int xmax, int ymax, Picture *ref_picture, int16_t (*last_mv)[2],
  214. int ref_mv_scale, uint8_t * const mv_penalty);
  215. int (*pre_motion_search)(struct MpegEncContext * s, int block,
  216. int *mx_ptr, int *my_ptr,
  217. int P[10][2], int pred_x, int pred_y,
  218. int xmin, int ymin, int xmax, int ymax, Picture *ref_picture, int16_t (*last_mv)[2],
  219. int ref_mv_scale, uint8_t * const mv_penalty);
  220. int (*get_mb_score)(struct MpegEncContext * s, int mx, int my, int pred_x, int pred_y, Picture *ref_picture,
  221. uint8_t * const mv_penalty);
  222. }MotionEstContext;
  223. /**
  224. * MpegEncContext.
  225. */
  226. typedef struct MpegEncContext {
  227. struct AVCodecContext *avctx;
  228. /* the following parameters must be initialized before encoding */
  229. int width, height;///< picture size. must be a multiple of 16
  230. int gop_size;
  231. int intra_only; ///< if true, only intra pictures are generated
  232. int bit_rate; ///< wanted bit rate
  233. int bit_rate_tolerance; ///< amount of +- bits (>0)
  234. enum OutputFormat out_format; ///< output format
  235. int h263_pred; ///< use mpeg4/h263 ac/dc predictions
  236. /* the following codec id fields are deprecated in favor of codec_id */
  237. int h263_plus; ///< h263 plus headers
  238. int h263_rv10; ///< use RV10 variation for H263
  239. int h263_msmpeg4; ///< generate MSMPEG4 compatible stream (deprecated, use msmpeg4_version instead)
  240. int h263_intel; ///< use I263 intel h263 header
  241. int codec_id; /* see CODEC_ID_xxx */
  242. int fixed_qscale; ///< fixed qscale if non zero
  243. float qcompress; ///< amount of qscale change between easy & hard scenes (0.0-1.0)
  244. float qblur; ///< amount of qscale smoothing over time (0.0-1.0)
  245. int max_qdiff; ///< max qscale difference between frames
  246. int encoding; ///< true if we are encoding (vs decoding)
  247. int flags; ///< AVCodecContext.flags (HQ, MV4, ...)
  248. int max_b_frames; ///< max number of b-frames for encoding
  249. int b_frame_strategy;
  250. int luma_elim_threshold;
  251. int chroma_elim_threshold;
  252. int strict_std_compliance; ///< strictly follow the std (MPEG4, ...)
  253. int workaround_bugs; ///< workaround bugs in encoders which cannot be detected automatically
  254. /* the following fields are managed internally by the encoder */
  255. /** bit output */
  256. PutBitContext pb;
  257. /* sequence parameters */
  258. int context_initialized;
  259. int input_picture_number;
  260. int picture_number;
  261. int picture_in_gop_number; ///< 0-> first pic in gop, ...
  262. int b_frames_since_non_b; ///< used for encoding, relative to not yet reordered input
  263. int mb_width, mb_height; ///< number of MBs horizontally & vertically
  264. int mb_stride; ///< mb_width+1 used for some arrays to allow simple addressng of left & top MBs withoutt sig11
  265. int h_edge_pos, v_edge_pos;///< horizontal / vertical position of the right/bottom edge (pixel replicateion)
  266. int mb_num; ///< number of MBs of a picture
  267. int linesize; ///< line size, in bytes, may be different from width
  268. int uvlinesize; ///< line size, for chroma in bytes, may be different from width
  269. Picture picture[MAX_PICTURE_COUNT]; ///< main picture buffer
  270. Picture *input_picture[MAX_PICTURE_COUNT]; ///< next pictures on display order for encoding
  271. Picture *reordered_input_picture[MAX_PICTURE_COUNT]; ///< pointer to the next pictures in codedorder for encoding
  272. /**
  273. * copy of the previous picture structure.
  274. * note, linesize & data, might not match the previous picture (for field pictures)
  275. */
  276. Picture last_picture;
  277. /**
  278. * copy of the next picture structure.
  279. * note, linesize & data, might not match the next picture (for field pictures)
  280. */
  281. Picture next_picture;
  282. /**
  283. * copy of the source picture structure for encoding.
  284. * note, linesize & data, might not match the source picture (for field pictures)
  285. */
  286. Picture new_picture;
  287. /**
  288. * copy of the current picture structure.
  289. * note, linesize & data, might not match the current picture (for field pictures)
  290. */
  291. Picture current_picture; ///< buffer to store the decompressed current picture
  292. Picture *last_picture_ptr; ///< pointer to the previous picture.
  293. Picture *next_picture_ptr; ///< pointer to the next picture (for bidir pred)
  294. Picture *new_picture_ptr; ///< pointer to the source picture for encoding
  295. Picture *current_picture_ptr; ///< pointer to the current picture
  296. int last_dc[3]; ///< last DC values for MPEG1
  297. int16_t *dc_val[3]; ///< used for mpeg4 DC prediction, all 3 arrays must be continuous
  298. int16_t dc_cache[4*5];
  299. int y_dc_scale, c_dc_scale;
  300. uint8_t *y_dc_scale_table; ///< qscale -> y_dc_scale table
  301. uint8_t *c_dc_scale_table; ///< qscale -> c_dc_scale table
  302. uint8_t *coded_block; ///< used for coded block pattern prediction (msmpeg4v3, wmv1)
  303. int16_t (*ac_val[3])[16]; ///< used for for mpeg4 AC prediction, all 3 arrays must be continuous
  304. int ac_pred;
  305. uint8_t *prev_pict_types; ///< previous picture types in bitstream order, used for mb skip
  306. #define PREV_PICT_TYPES_BUFFER_SIZE 256
  307. int mb_skiped; ///< MUST BE SET only during DECODING
  308. uint8_t *mbskip_table; /**< used to avoid copy if macroblock skipped (for black regions for example)
  309. and used for b-frame encoding & decoding (contains skip table of next P Frame) */
  310. uint8_t *mbintra_table; ///< used to avoid setting {ac, dc, cbp}-pred stuff to zero on inter MB decoding
  311. uint8_t *cbp_table; ///< used to store cbp, ac_pred for partitioned decoding
  312. uint8_t *pred_dir_table; ///< used to store pred_dir for partitioned decoding
  313. uint8_t *allocated_edge_emu_buffer;
  314. uint8_t *edge_emu_buffer; ///< points into the middle of allocated_edge_emu_buffer
  315. int qscale; ///< QP
  316. float frame_qscale; ///< qscale from the frame level rc FIXME remove
  317. int adaptive_quant; ///< use adaptive quantization
  318. int dquant; ///< qscale difference to prev qscale
  319. int pict_type; ///< I_TYPE, P_TYPE, B_TYPE, ...
  320. int last_pict_type;
  321. int last_non_b_pict_type; ///< used for mpeg4 gmc b-frames & ratecontrol
  322. int frame_rate_index;
  323. /* motion compensation */
  324. int unrestricted_mv; ///< mv can point outside of the coded picture
  325. int h263_long_vectors; ///< use horrible h263v1 long vector mode
  326. int decode; ///< if 0 then decoding will be skiped (for encoding b frames for example)
  327. DSPContext dsp; ///< pointers for accelerated dsp fucntions
  328. int f_code; ///< forward MV resolution
  329. int b_code; ///< backward MV resolution for B Frames (mpeg4)
  330. int16_t (*motion_val)[2];
  331. int16_t (*p_mv_table_base)[2];
  332. int16_t (*b_forw_mv_table_base)[2];
  333. int16_t (*b_back_mv_table_base)[2];
  334. int16_t (*b_bidir_forw_mv_table_base)[2];
  335. int16_t (*b_bidir_back_mv_table_base)[2];
  336. int16_t (*b_direct_mv_table_base)[2];
  337. int16_t (*p_mv_table)[2]; ///< MV table (1MV per MB) p-frame encoding
  338. int16_t (*b_forw_mv_table)[2]; ///< MV table (1MV per MB) forward mode b-frame encoding
  339. int16_t (*b_back_mv_table)[2]; ///< MV table (1MV per MB) backward mode b-frame encoding
  340. int16_t (*b_bidir_forw_mv_table)[2]; ///< MV table (1MV per MB) bidir mode b-frame encoding
  341. int16_t (*b_bidir_back_mv_table)[2]; ///< MV table (1MV per MB) bidir mode b-frame encoding
  342. int16_t (*b_direct_mv_table)[2]; ///< MV table (1MV per MB) direct mode b-frame encoding
  343. int me_method; ///< ME algorithm
  344. int scene_change_score;
  345. int mv_dir;
  346. #define MV_DIR_BACKWARD 1
  347. #define MV_DIR_FORWARD 2
  348. #define MV_DIRECT 4 ///< bidirectional mode where the difference equals the MV of the last P/S/I-Frame (mpeg4)
  349. int mv_type;
  350. #define MV_TYPE_16X16 0 ///< 1 vector for the whole mb
  351. #define MV_TYPE_8X8 1 ///< 4 vectors (h263, mpeg4 4MV)
  352. #define MV_TYPE_16X8 2 ///< 2 vectors, one per 16x8 block
  353. #define MV_TYPE_FIELD 3 ///< 2 vectors, one per field
  354. #define MV_TYPE_DMV 4 ///< 2 vectors, special mpeg2 Dual Prime Vectors
  355. /**motion vectors for a macroblock
  356. first coordinate : 0 = forward 1 = backward
  357. second " : depend on type
  358. third " : 0 = x, 1 = y
  359. */
  360. int mv[2][4][2];
  361. int field_select[2][2];
  362. int last_mv[2][2][2]; ///< last MV, used for MV prediction in MPEG1 & B-frame MPEG4
  363. uint8_t *fcode_tab; ///< smallest fcode needed for each MV
  364. MotionEstContext me;
  365. int no_rounding; /**< apply no rounding to motion compensation (MPEG4, msmpeg4, ...)
  366. for b-frames rounding mode is allways 0 */
  367. int hurry_up; /**< when set to 1 during decoding, b frames will be skiped
  368. when set to 2 idct/dequant will be skipped too */
  369. /* macroblock layer */
  370. int mb_x, mb_y;
  371. int mb_skip_run;
  372. int mb_intra;
  373. uint8_t *mb_type; ///< Table for MB type FIXME remove and use picture->mb_type
  374. #define MB_TYPE_INTRA 0x01
  375. #define MB_TYPE_INTER 0x02
  376. #define MB_TYPE_INTER4V 0x04
  377. #define MB_TYPE_SKIPED 0x08
  378. //#define MB_TYPE_GMC 0x10
  379. #define MB_TYPE_DIRECT 0x10
  380. #define MB_TYPE_FORWARD 0x20
  381. #define MB_TYPE_BACKWARD 0x40
  382. #define MB_TYPE_BIDIR 0x80
  383. int block_index[6]; ///< index to current MB in block based arrays with edges
  384. int block_wrap[6];
  385. int *mb_index2xy; ///< mb_index -> mb_x + mb_y*mb_stride
  386. /** matrix transmitted in the bitstream */
  387. uint16_t intra_matrix[64];
  388. uint16_t chroma_intra_matrix[64];
  389. uint16_t inter_matrix[64];
  390. uint16_t chroma_inter_matrix[64];
  391. #define QUANT_BIAS_SHIFT 8
  392. int intra_quant_bias; ///< bias for the quantizer
  393. int inter_quant_bias; ///< bias for the quantizer
  394. int min_qcoeff; ///< minimum encodable coefficient
  395. int max_qcoeff; ///< maximum encodable coefficient
  396. int ac_esc_length; ///< num of bits needed to encode the longest esc
  397. uint8_t *intra_ac_vlc_length;
  398. uint8_t *intra_ac_vlc_last_length;
  399. uint8_t *inter_ac_vlc_length;
  400. uint8_t *inter_ac_vlc_last_length;
  401. uint8_t *luma_dc_vlc_length;
  402. uint8_t *chroma_dc_vlc_length;
  403. #define UNI_AC_ENC_INDEX(run,level) ((run)*128 + (level))
  404. /** precomputed matrix (combine qscale and DCT renorm) */
  405. int __align8 q_intra_matrix[32][64];
  406. int __align8 q_inter_matrix[32][64];
  407. /** identical to the above but for MMX & these are not permutated */
  408. uint16_t __align8 q_intra_matrix16[32][64];
  409. uint16_t __align8 q_inter_matrix16[32][64];
  410. uint16_t __align8 q_intra_matrix16_bias[32][64];
  411. uint16_t __align8 q_inter_matrix16_bias[32][64];
  412. int block_last_index[6]; ///< last non zero coefficient in block
  413. /* scantables */
  414. ScanTable __align8 intra_scantable;
  415. ScanTable intra_h_scantable;
  416. ScanTable intra_v_scantable;
  417. ScanTable inter_scantable; ///< if inter == intra then intra should be used to reduce tha cache usage
  418. void *opaque; ///< private data for the user
  419. /* bit rate control */
  420. int I_frame_bits; //FIXME used in mpeg12 ...
  421. int64_t wanted_bits;
  422. int64_t total_bits;
  423. int frame_bits; ///< bits used for the current frame
  424. RateControlContext rc_context; ///< contains stuff only accessed in ratecontrol.c
  425. /* statistics, used for 2-pass encoding */
  426. int mv_bits;
  427. int header_bits;
  428. int i_tex_bits;
  429. int p_tex_bits;
  430. int i_count;
  431. int f_count;
  432. int b_count;
  433. int skip_count;
  434. int misc_bits; ///< cbp, mb_type
  435. int last_bits; ///< temp var used for calculating the above vars
  436. /* error concealment / resync */
  437. int error_count;
  438. uint8_t *error_status_table; ///< table of the error status of each MB
  439. #define VP_START 1 ///< current MB is the first after a resync marker
  440. #define AC_ERROR 2
  441. #define DC_ERROR 4
  442. #define MV_ERROR 8
  443. #define AC_END 16
  444. #define DC_END 32
  445. #define MV_END 64
  446. //FIXME some prefix?
  447. int resync_mb_x; ///< x position of last resync marker
  448. int resync_mb_y; ///< y position of last resync marker
  449. GetBitContext last_resync_gb; ///< used to search for the next resync marker
  450. int mb_num_left; ///< number of MBs left in this video packet (for partitioned Slices only)
  451. int next_p_frame_damaged; ///< set if the next p frame is damaged, to avoid showing trashed b frames
  452. int error_resilience;
  453. ParseContext parse_context;
  454. /* H.263 specific */
  455. int gob_number;
  456. int gob_index;
  457. /* H.263+ specific */
  458. int umvplus; ///< == H263+ && unrestricted_mv
  459. int h263_aic; ///< Advanded INTRA Coding (AIC)
  460. int h263_aic_dir; ///< AIC direction: 0 = left, 1 = top
  461. /* mpeg4 specific */
  462. int time_increment_resolution;
  463. int time_increment_bits; ///< number of bits to represent the fractional part of time
  464. int last_time_base;
  465. int time_base; ///< time in seconds of last I,P,S Frame
  466. int64_t time; ///< time of current frame
  467. int64_t last_non_b_time;
  468. uint16_t pp_time; ///< time distance between the last 2 p,s,i frames
  469. uint16_t pb_time; ///< time distance between the last b and p,s,i frame
  470. uint16_t pp_field_time;
  471. uint16_t pb_field_time; ///< like above, just for interlaced
  472. int shape;
  473. int vol_sprite_usage;
  474. int sprite_width;
  475. int sprite_height;
  476. int sprite_left;
  477. int sprite_top;
  478. int sprite_brightness_change;
  479. int num_sprite_warping_points;
  480. int real_sprite_warping_points;
  481. int sprite_offset[2][2]; ///< sprite offset[isChroma][isMVY]
  482. int sprite_delta[2][2]; ///< sprite_delta [isY][isMVY]
  483. int sprite_shift[2]; ///< sprite shift [isChroma]
  484. int mcsel;
  485. int quant_precision;
  486. int quarter_sample; ///< 1->qpel, 0->half pel ME/MC
  487. int scalability;
  488. int hierachy_type;
  489. int enhancement_type;
  490. int new_pred;
  491. int reduced_res_vop;
  492. int aspect_ratio_info; //FIXME remove
  493. int aspected_width; //FIXME remove
  494. int aspected_height; //FIXME remove
  495. int sprite_warping_accuracy;
  496. int low_latency_sprite;
  497. int data_partitioning; ///< data partitioning flag from header
  498. int partitioned_frame; ///< is current frame partitioned
  499. int rvlc; ///< reversible vlc
  500. int resync_marker; ///< could this stream contain resync markers
  501. int low_delay; ///< no reordering needed / has no b-frames
  502. int vo_type;
  503. int vol_control_parameters; ///< does the stream contain the low_delay flag, used to workaround buggy encoders
  504. PutBitContext tex_pb; ///< used for data partitioned VOPs
  505. PutBitContext pb2; ///< used for data partitioned VOPs
  506. #define PB_BUFFER_SIZE 1024*256
  507. uint8_t *tex_pb_buffer;
  508. uint8_t *pb2_buffer;
  509. int mpeg_quant;
  510. int16_t (*field_mv_table)[2][2]; ///< used for interlaced b frame decoding
  511. int8_t (*field_select_table)[2]; ///< wtf, no really another table for interlaced b frames
  512. int t_frame; ///< time distance of first I -> B, used for interlaced b frames
  513. int padding_bug_score; ///< used to detect the VERY common padding bug in MPEG4
  514. /* divx specific, used to workaround (many) bugs in divx5 */
  515. int divx_version;
  516. int divx_build;
  517. int divx_packed;
  518. #define BITSTREAM_BUFFER_SIZE 1024*256
  519. uint8_t *bitstream_buffer; //Divx 5.01 puts several frames in a single one, this is used to reorder them
  520. int bitstream_buffer_size;
  521. int xvid_build;
  522. /* lavc specific stuff, used to workaround bugs in libavcodec */
  523. int ffmpeg_version;
  524. int lavc_build;
  525. /* RV10 specific */
  526. int rv10_version; ///< RV10 version: 0 or 3
  527. int rv10_first_dc_coded[3];
  528. /* MJPEG specific */
  529. struct MJpegContext *mjpeg_ctx;
  530. int mjpeg_vsample[3]; ///< vertical sampling factors, default = {2, 1, 1}
  531. int mjpeg_hsample[3]; ///< horizontal sampling factors, default = {2, 1, 1}
  532. int mjpeg_write_tables; ///< do we want to have quantisation- and huffmantables in the jpeg file ?
  533. int mjpeg_data_only_frames; ///< frames only with SOI, SOS and EOI markers
  534. /* MSMPEG4 specific */
  535. int mv_table_index;
  536. int rl_table_index;
  537. int rl_chroma_table_index;
  538. int dc_table_index;
  539. int use_skip_mb_code;
  540. int slice_height; ///< in macroblocks
  541. int first_slice_line; ///< used in mpeg4 too to handle resync markers
  542. int flipflop_rounding;
  543. int msmpeg4_version; ///< 0=not msmpeg4, 1=mp41, 2=mp42, 3=mp43/divx3 4=wmv1/7 5=wmv2/8
  544. int per_mb_rl_table;
  545. int esc3_level_length;
  546. int esc3_run_length;
  547. /** [mb_intra][isChroma][level][run][last] */
  548. int (*ac_stats)[2][MAX_LEVEL+1][MAX_RUN+1][2];
  549. int inter_intra_pred;
  550. int mspel;
  551. /* decompression specific */
  552. GetBitContext gb;
  553. /* Mpeg1 specific */
  554. int fake_picture_number; ///< picture number at the bitstream frame rate
  555. int gop_picture_number; ///< index of the first picture of a GOP based on fake_pic_num & mpeg1 specific
  556. int last_mv_dir; ///< last mv_dir, used for b frame encoding
  557. int broken_link; ///< no_output_of_prior_pics_flag
  558. /* MPEG2 specific - I wish I had not to support this mess. */
  559. int progressive_sequence;
  560. int mpeg_f_code[2][2];
  561. int picture_structure;
  562. /* picture type */
  563. #define PICT_TOP_FIELD 1
  564. #define PICT_BOTTOM_FIELD 2
  565. #define PICT_FRAME 3
  566. int intra_dc_precision;
  567. int frame_pred_frame_dct;
  568. int top_field_first;
  569. int concealment_motion_vectors;
  570. int q_scale_type;
  571. int intra_vlc_format;
  572. int alternate_scan;
  573. int repeat_first_field;
  574. int chroma_420_type;
  575. int progressive_frame;
  576. int mpeg2;
  577. int full_pel[2];
  578. int interlaced_dct;
  579. int first_slice;
  580. int first_field;
  581. /* RTP specific */
  582. /* These are explained on avcodec.h */
  583. int rtp_mode;
  584. int rtp_payload_size;
  585. void (*rtp_callback)(void *data, int size, int packet_number);
  586. uint8_t *ptr_lastgob;
  587. DCTELEM (*block)[64]; ///< points to one of the following blocks
  588. DCTELEM blocks[2][6][64] __align8; // for HQ mode we need to keep the best block
  589. int (*decode_mb)(struct MpegEncContext *s, DCTELEM block[6][64]); // used by some codecs to avoid a switch()
  590. #define SLICE_OK 0
  591. #define SLICE_ERROR -1
  592. #define SLICE_END -2 //end marker found
  593. #define SLICE_NOEND -3 //no end marker or error found but mb count exceeded
  594. void (*dct_unquantize_mpeg1)(struct MpegEncContext *s,
  595. DCTELEM *block/*align 16*/, int n, int qscale);
  596. void (*dct_unquantize_mpeg2)(struct MpegEncContext *s,
  597. DCTELEM *block/*align 16*/, int n, int qscale);
  598. void (*dct_unquantize_h263)(struct MpegEncContext *s,
  599. DCTELEM *block/*align 16*/, int n, int qscale);
  600. void (*dct_unquantize)(struct MpegEncContext *s, // unquantizer to use (mpeg4 can use both)
  601. DCTELEM *block/*align 16*/, int n, int qscale);
  602. int (*dct_quantize)(struct MpegEncContext *s, DCTELEM *block/*align 16*/, int n, int qscale, int *overflow);
  603. int (*fast_dct_quantize)(struct MpegEncContext *s, DCTELEM *block/*align 16*/, int n, int qscale, int *overflow);
  604. } MpegEncContext;
  605. int DCT_common_init(MpegEncContext *s);
  606. int MPV_common_init(MpegEncContext *s);
  607. void MPV_common_end(MpegEncContext *s);
  608. void MPV_decode_mb(MpegEncContext *s, DCTELEM block[6][64]);
  609. int MPV_frame_start(MpegEncContext *s, AVCodecContext *avctx);
  610. void MPV_frame_end(MpegEncContext *s);
  611. int MPV_encode_init(AVCodecContext *avctx);
  612. int MPV_encode_end(AVCodecContext *avctx);
  613. int MPV_encode_picture(AVCodecContext *avctx, unsigned char *buf, int buf_size, void *data);
  614. #ifdef HAVE_MMX
  615. void MPV_common_init_mmx(MpegEncContext *s);
  616. #endif
  617. #ifdef ARCH_ALPHA
  618. void MPV_common_init_axp(MpegEncContext *s);
  619. #endif
  620. #ifdef HAVE_MLIB
  621. void MPV_common_init_mlib(MpegEncContext *s);
  622. #endif
  623. #ifdef HAVE_MMI
  624. void MPV_common_init_mmi(MpegEncContext *s);
  625. #endif
  626. #ifdef ARCH_ARMV4L
  627. void MPV_common_init_armv4l(MpegEncContext *s);
  628. #endif
  629. #ifdef ARCH_POWERPC
  630. void MPV_common_init_ppc(MpegEncContext *s);
  631. #endif
  632. extern void (*draw_edges)(uint8_t *buf, int wrap, int width, int height, int w);
  633. void ff_copy_bits(PutBitContext *pb, uint8_t *src, int length);
  634. void ff_clean_intra_table_entries(MpegEncContext *s);
  635. void ff_init_scantable(MpegEncContext *s, ScanTable *st, const uint8_t *src_scantable);
  636. void ff_draw_horiz_band(MpegEncContext *s, int y, int h);
  637. void ff_emulated_edge_mc(MpegEncContext *s, uint8_t *src, int linesize, int block_w, int block_h,
  638. int src_x, int src_y, int w, int h);
  639. char ff_get_pict_type_char(int pict_type);
  640. int ff_combine_frame( MpegEncContext *s, int next, uint8_t **buf, int *buf_size);
  641. void ff_print_debug_info(MpegEncContext *s, Picture *pict);
  642. void ff_er_frame_start(MpegEncContext *s);
  643. void ff_er_frame_end(MpegEncContext *s);
  644. void ff_er_add_slice(MpegEncContext *s, int startx, int starty, int endx, int endy, int status);
  645. extern enum PixelFormat ff_yuv420p_list[2];
  646. static inline void ff_init_block_index(MpegEncContext *s){
  647. s->block_index[0]= s->block_wrap[0]*(s->mb_y*2 + 1) - 1 + s->mb_x*2;
  648. s->block_index[1]= s->block_wrap[0]*(s->mb_y*2 + 1) + s->mb_x*2;
  649. s->block_index[2]= s->block_wrap[0]*(s->mb_y*2 + 2) - 1 + s->mb_x*2;
  650. s->block_index[3]= s->block_wrap[0]*(s->mb_y*2 + 2) + s->mb_x*2;
  651. s->block_index[4]= s->block_wrap[4]*(s->mb_y + 1) + s->block_wrap[0]*(s->mb_height*2 + 2) + s->mb_x;
  652. s->block_index[5]= s->block_wrap[4]*(s->mb_y + 1 + s->mb_height + 2) + s->block_wrap[0]*(s->mb_height*2 + 2) + s->mb_x;
  653. }
  654. static inline void ff_update_block_index(MpegEncContext *s){
  655. s->block_index[0]+=2;
  656. s->block_index[1]+=2;
  657. s->block_index[2]+=2;
  658. s->block_index[3]+=2;
  659. s->block_index[4]++;
  660. s->block_index[5]++;
  661. }
  662. static inline int get_bits_diff(MpegEncContext *s){
  663. const int bits= get_bit_count(&s->pb);
  664. const int last= s->last_bits;
  665. s->last_bits = bits;
  666. return bits - last;
  667. }
  668. /* motion_est.c */
  669. void ff_estimate_p_frame_motion(MpegEncContext * s,
  670. int mb_x, int mb_y);
  671. void ff_estimate_b_frame_motion(MpegEncContext * s,
  672. int mb_x, int mb_y);
  673. int ff_get_best_fcode(MpegEncContext * s, int16_t (*mv_table)[2], int type);
  674. void ff_fix_long_p_mvs(MpegEncContext * s);
  675. void ff_fix_long_b_mvs(MpegEncContext * s, int16_t (*mv_table)[2], int f_code, int type);
  676. void ff_init_me(MpegEncContext *s);
  677. int ff_pre_estimate_p_frame_motion(MpegEncContext * s, int mb_x, int mb_y);
  678. /* mpeg12.c */
  679. extern const int16_t ff_mpeg1_default_intra_matrix[64];
  680. extern const int16_t ff_mpeg1_default_non_intra_matrix[64];
  681. extern uint8_t ff_mpeg1_dc_scale_table[128];
  682. void mpeg1_encode_picture_header(MpegEncContext *s, int picture_number);
  683. void mpeg1_encode_mb(MpegEncContext *s,
  684. DCTELEM block[6][64],
  685. int motion_x, int motion_y);
  686. void ff_mpeg1_encode_init(MpegEncContext *s);
  687. void ff_mpeg1_encode_slice_header(MpegEncContext *s);
  688. void ff_mpeg1_clean_buffers(MpegEncContext *s);
  689. /** RLTable. */
  690. typedef struct RLTable {
  691. int n; ///< number of entries of table_vlc minus 1
  692. int last; ///< number of values for last = 0
  693. const uint16_t (*table_vlc)[2];
  694. const int8_t *table_run;
  695. const int8_t *table_level;
  696. uint8_t *index_run[2]; ///< encoding only
  697. int8_t *max_level[2]; ///< encoding & decoding
  698. int8_t *max_run[2]; ///< encoding & decoding
  699. VLC vlc; ///< decoding only deprected FIXME remove
  700. RL_VLC_ELEM *rl_vlc[32]; ///< decoding only
  701. } RLTable;
  702. void init_rl(RLTable *rl);
  703. void init_vlc_rl(RLTable *rl);
  704. static inline int get_rl_index(const RLTable *rl, int last, int run, int level)
  705. {
  706. int index;
  707. index = rl->index_run[last][run];
  708. if (index >= rl->n)
  709. return rl->n;
  710. if (level > rl->max_level[last][run])
  711. return rl->n;
  712. return index + level - 1;
  713. }
  714. extern uint8_t ff_mpeg4_y_dc_scale_table[32];
  715. extern uint8_t ff_mpeg4_c_dc_scale_table[32];
  716. extern const int16_t ff_mpeg4_default_intra_matrix[64];
  717. extern const int16_t ff_mpeg4_default_non_intra_matrix[64];
  718. int ff_h263_decode_init(AVCodecContext *avctx);
  719. int ff_h263_decode_frame(AVCodecContext *avctx,
  720. void *data, int *data_size,
  721. uint8_t *buf, int buf_size);
  722. int ff_h263_decode_end(AVCodecContext *avctx);
  723. void h263_encode_mb(MpegEncContext *s,
  724. DCTELEM block[6][64],
  725. int motion_x, int motion_y);
  726. void mpeg4_encode_mb(MpegEncContext *s,
  727. DCTELEM block[6][64],
  728. int motion_x, int motion_y);
  729. void h263_encode_picture_header(MpegEncContext *s, int picture_number);
  730. int h263_encode_gob_header(MpegEncContext * s, int mb_line);
  731. int16_t *h263_pred_motion(MpegEncContext * s, int block,
  732. int *px, int *py);
  733. void mpeg4_pred_ac(MpegEncContext * s, DCTELEM *block, int n,
  734. int dir);
  735. void ff_set_mpeg4_time(MpegEncContext * s, int picture_number);
  736. void mpeg4_encode_picture_header(MpegEncContext *s, int picture_number);
  737. void h263_encode_init(MpegEncContext *s);
  738. void h263_decode_init_vlc(MpegEncContext *s);
  739. int h263_decode_picture_header(MpegEncContext *s);
  740. int ff_h263_decode_gob_header(MpegEncContext *s);
  741. int ff_mpeg4_decode_picture_header(MpegEncContext * s, GetBitContext *gb);
  742. int intel_h263_decode_picture_header(MpegEncContext *s);
  743. int ff_h263_decode_mb(MpegEncContext *s,
  744. DCTELEM block[6][64]);
  745. int h263_get_picture_format(int width, int height);
  746. void ff_mpeg4_encode_video_packet_header(MpegEncContext *s);
  747. void ff_mpeg4_clean_buffers(MpegEncContext *s);
  748. void ff_mpeg4_stuffing(PutBitContext * pbc);
  749. void ff_mpeg4_init_partitions(MpegEncContext *s);
  750. void ff_mpeg4_merge_partitions(MpegEncContext *s);
  751. void ff_clean_mpeg4_qscales(MpegEncContext *s);
  752. void ff_clean_h263_qscales(MpegEncContext *s);
  753. int ff_mpeg4_decode_partitions(MpegEncContext *s);
  754. int ff_mpeg4_get_video_packet_prefix_length(MpegEncContext *s);
  755. int ff_h263_resync(MpegEncContext *s);
  756. int ff_h263_get_gob_height(MpegEncContext *s);
  757. int ff_mpeg4_set_direct_mv(MpegEncContext *s, int mx, int my);
  758. inline int ff_h263_round_chroma(int x);
  759. /* rv10.c */
  760. void rv10_encode_picture_header(MpegEncContext *s, int picture_number);
  761. int rv_decode_dc(MpegEncContext *s, int n);
  762. /* msmpeg4.c */
  763. void msmpeg4_encode_picture_header(MpegEncContext * s, int picture_number);
  764. void msmpeg4_encode_ext_header(MpegEncContext * s);
  765. void msmpeg4_encode_mb(MpegEncContext * s,
  766. DCTELEM block[6][64],
  767. int motion_x, int motion_y);
  768. int msmpeg4_decode_picture_header(MpegEncContext * s);
  769. int msmpeg4_decode_ext_header(MpegEncContext * s, int buf_size);
  770. int ff_msmpeg4_decode_init(MpegEncContext *s);
  771. void ff_msmpeg4_encode_init(MpegEncContext *s);
  772. int ff_wmv2_decode_picture_header(MpegEncContext * s);
  773. void ff_wmv2_add_mb(MpegEncContext *s, DCTELEM block[6][64], uint8_t *dest_y, uint8_t *dest_cb, uint8_t *dest_cr);
  774. void ff_mspel_motion(MpegEncContext *s,
  775. uint8_t *dest_y, uint8_t *dest_cb, uint8_t *dest_cr,
  776. uint8_t **ref_picture, op_pixels_func (*pix_op)[4],
  777. int motion_x, int motion_y, int h);
  778. int ff_wmv2_encode_picture_header(MpegEncContext * s, int picture_number);
  779. void ff_wmv2_encode_mb(MpegEncContext * s,
  780. DCTELEM block[6][64],
  781. int motion_x, int motion_y);
  782. /* mjpegenc.c */
  783. int mjpeg_init(MpegEncContext *s);
  784. void mjpeg_close(MpegEncContext *s);
  785. void mjpeg_encode_mb(MpegEncContext *s,
  786. DCTELEM block[6][64]);
  787. void mjpeg_picture_header(MpegEncContext *s);
  788. void mjpeg_picture_trailer(MpegEncContext *s);
  789. /* rate control */
  790. int ff_rate_control_init(MpegEncContext *s);
  791. float ff_rate_estimate_qscale(MpegEncContext *s);
  792. void ff_write_pass1_stats(MpegEncContext *s);
  793. void ff_rate_control_uninit(MpegEncContext *s);
  794. double ff_eval(char *s, double *const_value, const char **const_name,
  795. double (**func1)(void *, double), const char **func1_name,
  796. double (**func2)(void *, double, double), char **func2_name,
  797. void *opaque);
  798. #endif /* AVCODEC_MPEGVIDEO_H */