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.

871 lines
38KB

  1. /*
  2. * Generic DCT based hybrid video encoder
  3. * Copyright (c) 2000, 2001, 2002 Fabrice Bellard
  4. * Copyright (c) 2002-2004 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. * @file
  24. * mpegvideo header.
  25. */
  26. #ifndef AVCODEC_MPEGVIDEO_H
  27. #define AVCODEC_MPEGVIDEO_H
  28. #include <float.h>
  29. #include "avcodec.h"
  30. #include "blockdsp.h"
  31. #include "error_resilience.h"
  32. #include "fdctdsp.h"
  33. #include "get_bits.h"
  34. #include "h263dsp.h"
  35. #include "hpeldsp.h"
  36. #include "idctdsp.h"
  37. #include "me_cmp.h"
  38. #include "mpegvideodsp.h"
  39. #include "mpegvideoencdsp.h"
  40. #include "pixblockdsp.h"
  41. #include "put_bits.h"
  42. #include "ratecontrol.h"
  43. #include "parser.h"
  44. #include "mpeg12data.h"
  45. #include "qpeldsp.h"
  46. #include "thread.h"
  47. #include "videodsp.h"
  48. #include "libavutil/opt.h"
  49. #define FRAME_SKIPPED 100 ///< return value for header parsers if frame is not coded
  50. enum OutputFormat {
  51. FMT_MPEG1,
  52. FMT_H261,
  53. FMT_H263,
  54. FMT_MJPEG,
  55. };
  56. #define MAX_FCODE 7
  57. #define MAX_MV 2048
  58. #define MAX_THREADS 16
  59. #define MAX_PICTURE_COUNT 32
  60. #define MAX_B_FRAMES 16
  61. #define ME_MAP_SIZE 64
  62. #define MAX_MB_BYTES (30*16*16*3/8 + 120)
  63. #define INPLACE_OFFSET 16
  64. #define EDGE_WIDTH 16
  65. /* Start codes. */
  66. #define SEQ_END_CODE 0x000001b7
  67. #define SEQ_START_CODE 0x000001b3
  68. #define GOP_START_CODE 0x000001b8
  69. #define PICTURE_START_CODE 0x00000100
  70. #define SLICE_MIN_START_CODE 0x00000101
  71. #define SLICE_MAX_START_CODE 0x000001af
  72. #define EXT_START_CODE 0x000001b5
  73. #define USER_START_CODE 0x000001b2
  74. /* encoding scans */
  75. extern const uint8_t ff_alternate_horizontal_scan[64];
  76. extern const uint8_t ff_alternate_vertical_scan[64];
  77. struct MpegEncContext;
  78. /**
  79. * Picture.
  80. */
  81. typedef struct Picture{
  82. struct AVFrame *f;
  83. ThreadFrame tf;
  84. AVBufferRef *qscale_table_buf;
  85. int8_t *qscale_table;
  86. AVBufferRef *motion_val_buf[2];
  87. int16_t (*motion_val[2])[2];
  88. AVBufferRef *mb_type_buf;
  89. uint32_t *mb_type; ///< types and macros are defined in mpegutils.h
  90. AVBufferRef *mbskip_table_buf;
  91. uint8_t *mbskip_table;
  92. AVBufferRef *ref_index_buf[2];
  93. int8_t *ref_index[2];
  94. AVBufferRef *mb_var_buf;
  95. uint16_t *mb_var; ///< Table for MB variances
  96. AVBufferRef *mc_mb_var_buf;
  97. uint16_t *mc_mb_var; ///< Table for motion compensated MB variances
  98. AVBufferRef *mb_mean_buf;
  99. uint8_t *mb_mean; ///< Table for MB luminance
  100. AVBufferRef *hwaccel_priv_buf;
  101. /**
  102. * hardware accelerator private data
  103. */
  104. void *hwaccel_picture_private;
  105. int field_picture; ///< whether or not the picture was encoded in separate fields
  106. int mb_var_sum; ///< sum of MB variance for current frame
  107. int mc_mb_var_sum; ///< motion compensated MB variance for current frame
  108. int b_frame_score; /* */
  109. int needs_realloc; ///< Picture needs to be reallocated (eg due to a frame size change)
  110. int reference;
  111. int shared;
  112. } Picture;
  113. /**
  114. * Motion estimation context.
  115. */
  116. typedef struct MotionEstContext{
  117. AVCodecContext *avctx;
  118. int skip; ///< set if ME is skipped for the current MB
  119. int co_located_mv[4][2]; ///< mv from last P-frame for direct mode ME
  120. int direct_basis_mv[4][2];
  121. uint8_t *scratchpad; ///< data area for the ME algo, so that the ME does not need to malloc/free
  122. uint8_t *best_mb;
  123. uint8_t *temp_mb[2];
  124. uint8_t *temp;
  125. int best_bits;
  126. uint32_t *map; ///< map to avoid duplicate evaluations
  127. uint32_t *score_map; ///< map to store the scores
  128. unsigned map_generation;
  129. int pre_penalty_factor;
  130. int penalty_factor; /**< an estimate of the bits required to
  131. code a given mv value, e.g. (1,0) takes
  132. more bits than (0,0). We have to
  133. estimate whether any reduction in
  134. residual is worth the extra bits. */
  135. int sub_penalty_factor;
  136. int mb_penalty_factor;
  137. int flags;
  138. int sub_flags;
  139. int mb_flags;
  140. int pre_pass; ///< = 1 for the pre pass
  141. int dia_size;
  142. int xmin;
  143. int xmax;
  144. int ymin;
  145. int ymax;
  146. int pred_x;
  147. int pred_y;
  148. uint8_t *src[4][4];
  149. uint8_t *ref[4][4];
  150. int stride;
  151. int uvstride;
  152. /* temp variables for picture complexity calculation */
  153. int mc_mb_var_sum_temp;
  154. int mb_var_sum_temp;
  155. int scene_change_score;
  156. /* cmp, chroma_cmp;*/
  157. op_pixels_func (*hpel_put)[4];
  158. op_pixels_func (*hpel_avg)[4];
  159. qpel_mc_func (*qpel_put)[16];
  160. qpel_mc_func (*qpel_avg)[16];
  161. uint8_t (*mv_penalty)[MAX_MV*2+1]; ///< amount of bits needed to encode a MV
  162. uint8_t *current_mv_penalty;
  163. int (*sub_motion_search)(struct MpegEncContext * s,
  164. int *mx_ptr, int *my_ptr, int dmin,
  165. int src_index, int ref_index,
  166. int size, int h);
  167. }MotionEstContext;
  168. /**
  169. * MpegEncContext.
  170. */
  171. typedef struct MpegEncContext {
  172. AVClass *class;
  173. int y_dc_scale, c_dc_scale;
  174. int ac_pred;
  175. int block_last_index[12]; ///< last non zero coefficient in block
  176. int h263_aic; ///< Advanded INTRA Coding (AIC)
  177. /* scantables */
  178. ScanTable inter_scantable; ///< if inter == intra then intra should be used to reduce tha cache usage
  179. ScanTable intra_scantable;
  180. ScanTable intra_h_scantable;
  181. ScanTable intra_v_scantable;
  182. /* WARNING: changes above this line require updates to hardcoded
  183. * offsets used in asm. */
  184. struct AVCodecContext *avctx;
  185. /* the following parameters must be initialized before encoding */
  186. int width, height;///< picture size. must be a multiple of 16
  187. int gop_size;
  188. int intra_only; ///< if true, only intra pictures are generated
  189. int bit_rate; ///< wanted bit rate
  190. enum OutputFormat out_format; ///< output format
  191. int h263_pred; ///< use mpeg4/h263 ac/dc predictions
  192. int pb_frame; ///< PB frame mode (0 = none, 1 = base, 2 = improved)
  193. /* the following codec id fields are deprecated in favor of codec_id */
  194. int h263_plus; ///< h263 plus headers
  195. int h263_flv; ///< use flv h263 header
  196. enum AVCodecID codec_id; /* see AV_CODEC_ID_xxx */
  197. int fixed_qscale; ///< fixed qscale if non zero
  198. int encoding; ///< true if we are encoding (vs decoding)
  199. int max_b_frames; ///< max number of b-frames for encoding
  200. int luma_elim_threshold;
  201. int chroma_elim_threshold;
  202. int strict_std_compliance; ///< strictly follow the std (MPEG4, ...)
  203. int workaround_bugs; ///< workaround bugs in encoders which cannot be detected automatically
  204. int codec_tag; ///< internal codec_tag upper case converted from avctx codec_tag
  205. /* the following fields are managed internally by the encoder */
  206. /* sequence parameters */
  207. int context_initialized;
  208. int input_picture_number; ///< used to set pic->display_picture_number, should not be used for/by anything else
  209. int coded_picture_number; ///< used to set pic->coded_picture_number, should not be used for/by anything else
  210. int picture_number; //FIXME remove, unclear definition
  211. int picture_in_gop_number; ///< 0-> first pic in gop, ...
  212. int mb_width, mb_height; ///< number of MBs horizontally & vertically
  213. int mb_stride; ///< mb_width+1 used for some arrays to allow simple addressing of left & top MBs without sig11
  214. int b8_stride; ///< 2*mb_width+1 used for some 8x8 block arrays to allow simple addressing
  215. int h_edge_pos, v_edge_pos;///< horizontal / vertical position of the right/bottom edge (pixel replication)
  216. int mb_num; ///< number of MBs of a picture
  217. ptrdiff_t linesize; ///< line size, in bytes, may be different from width
  218. ptrdiff_t uvlinesize; ///< line size, for chroma in bytes, may be different from width
  219. Picture *picture; ///< main picture buffer
  220. Picture **input_picture; ///< next pictures on display order for encoding
  221. Picture **reordered_input_picture; ///< pointer to the next pictures in codedorder for encoding
  222. int64_t user_specified_pts; ///< last non-zero pts from AVFrame which was passed into avcodec_encode_video2()
  223. /**
  224. * pts difference between the first and second input frame, used for
  225. * calculating dts of the first frame when there's a delay */
  226. int64_t dts_delta;
  227. /**
  228. * reordered pts to be used as dts for the next output frame when there's
  229. * a delay */
  230. int64_t reordered_pts;
  231. /** bit output */
  232. PutBitContext pb;
  233. int start_mb_y; ///< start mb_y of this thread (so current thread should process start_mb_y <= row < end_mb_y)
  234. int end_mb_y; ///< end mb_y of this thread (so current thread should process start_mb_y <= row < end_mb_y)
  235. struct MpegEncContext *thread_context[MAX_THREADS];
  236. int slice_context_count; ///< number of used thread_contexts
  237. /**
  238. * copy of the previous picture structure.
  239. * note, linesize & data, might not match the previous picture (for field pictures)
  240. */
  241. Picture last_picture;
  242. /**
  243. * copy of the next picture structure.
  244. * note, linesize & data, might not match the next picture (for field pictures)
  245. */
  246. Picture next_picture;
  247. /**
  248. * copy of the source picture structure for encoding.
  249. * note, linesize & data, might not match the source picture (for field pictures)
  250. */
  251. Picture new_picture;
  252. /**
  253. * copy of the current picture structure.
  254. * note, linesize & data, might not match the current picture (for field pictures)
  255. */
  256. Picture current_picture; ///< buffer to store the decompressed current picture
  257. Picture *last_picture_ptr; ///< pointer to the previous picture.
  258. Picture *next_picture_ptr; ///< pointer to the next picture (for bidir pred)
  259. Picture *current_picture_ptr; ///< pointer to the current picture
  260. int last_dc[3]; ///< last DC values for MPEG1
  261. int16_t *dc_val_base;
  262. int16_t *dc_val[3]; ///< used for mpeg4 DC prediction, all 3 arrays must be continuous
  263. const uint8_t *y_dc_scale_table; ///< qscale -> y_dc_scale table
  264. const uint8_t *c_dc_scale_table; ///< qscale -> c_dc_scale table
  265. const uint8_t *chroma_qscale_table; ///< qscale -> chroma_qscale (h263)
  266. uint8_t *coded_block_base;
  267. uint8_t *coded_block; ///< used for coded block pattern prediction (msmpeg4v3, wmv1)
  268. int16_t (*ac_val_base)[16];
  269. int16_t (*ac_val[3])[16]; ///< used for for mpeg4 AC prediction, all 3 arrays must be continuous
  270. int mb_skipped; ///< MUST BE SET only during DECODING
  271. uint8_t *mbskip_table; /**< used to avoid copy if macroblock skipped (for black regions for example)
  272. and used for b-frame encoding & decoding (contains skip table of next P Frame) */
  273. uint8_t *mbintra_table; ///< used to avoid setting {ac, dc, cbp}-pred stuff to zero on inter MB decoding
  274. uint8_t *cbp_table; ///< used to store cbp, ac_pred for partitioned decoding
  275. uint8_t *pred_dir_table; ///< used to store pred_dir for partitioned decoding
  276. uint8_t *edge_emu_buffer; ///< temporary buffer for if MVs point to out-of-frame data
  277. uint8_t *rd_scratchpad; ///< scratchpad for rate distortion mb decision
  278. uint8_t *obmc_scratchpad;
  279. uint8_t *b_scratchpad; ///< scratchpad used for writing into write only buffers
  280. int qscale; ///< QP
  281. int chroma_qscale; ///< chroma QP
  282. unsigned int lambda; ///< lagrange multipler used in rate distortion
  283. unsigned int lambda2; ///< (lambda*lambda) >> FF_LAMBDA_SHIFT
  284. int *lambda_table;
  285. int adaptive_quant; ///< use adaptive quantization
  286. int dquant; ///< qscale difference to prev qscale
  287. int pict_type; ///< AV_PICTURE_TYPE_I, AV_PICTURE_TYPE_P, AV_PICTURE_TYPE_B, ...
  288. int last_pict_type; //FIXME removes
  289. int last_non_b_pict_type; ///< used for mpeg4 gmc b-frames & ratecontrol
  290. int droppable;
  291. int frame_rate_index;
  292. int last_lambda_for[5]; ///< last lambda for a specific pict type
  293. int skipdct; ///< skip dct and code zero residual
  294. /* motion compensation */
  295. int unrestricted_mv; ///< mv can point outside of the coded picture
  296. int h263_long_vectors; ///< use horrible h263v1 long vector mode
  297. BlockDSPContext bdsp;
  298. FDCTDSPContext fdsp;
  299. HpelDSPContext hdsp;
  300. IDCTDSPContext idsp;
  301. MECmpContext mecc;
  302. MpegVideoDSPContext mdsp;
  303. MpegvideoEncDSPContext mpvencdsp;
  304. PixblockDSPContext pdsp;
  305. QpelDSPContext qdsp;
  306. VideoDSPContext vdsp;
  307. H263DSPContext h263dsp;
  308. int f_code; ///< forward MV resolution
  309. int b_code; ///< backward MV resolution for B Frames (mpeg4)
  310. int16_t (*p_mv_table_base)[2];
  311. int16_t (*b_forw_mv_table_base)[2];
  312. int16_t (*b_back_mv_table_base)[2];
  313. int16_t (*b_bidir_forw_mv_table_base)[2];
  314. int16_t (*b_bidir_back_mv_table_base)[2];
  315. int16_t (*b_direct_mv_table_base)[2];
  316. int16_t (*p_field_mv_table_base[2][2])[2];
  317. int16_t (*b_field_mv_table_base[2][2][2])[2];
  318. int16_t (*p_mv_table)[2]; ///< MV table (1MV per MB) p-frame encoding
  319. int16_t (*b_forw_mv_table)[2]; ///< MV table (1MV per MB) forward mode b-frame encoding
  320. int16_t (*b_back_mv_table)[2]; ///< MV table (1MV per MB) backward mode b-frame encoding
  321. int16_t (*b_bidir_forw_mv_table)[2]; ///< MV table (1MV per MB) bidir mode b-frame encoding
  322. int16_t (*b_bidir_back_mv_table)[2]; ///< MV table (1MV per MB) bidir mode b-frame encoding
  323. int16_t (*b_direct_mv_table)[2]; ///< MV table (1MV per MB) direct mode b-frame encoding
  324. int16_t (*p_field_mv_table[2][2])[2]; ///< MV table (2MV per MB) interlaced p-frame encoding
  325. int16_t (*b_field_mv_table[2][2][2])[2];///< MV table (4MV per MB) interlaced b-frame encoding
  326. uint8_t (*p_field_select_table[2]);
  327. uint8_t (*b_field_select_table[2][2]);
  328. int me_method; ///< ME algorithm
  329. int mv_dir;
  330. #define MV_DIR_FORWARD 1
  331. #define MV_DIR_BACKWARD 2
  332. #define MV_DIRECT 4 ///< bidirectional mode where the difference equals the MV of the last P/S/I-Frame (mpeg4)
  333. int mv_type;
  334. #define MV_TYPE_16X16 0 ///< 1 vector for the whole mb
  335. #define MV_TYPE_8X8 1 ///< 4 vectors (h263, mpeg4 4MV)
  336. #define MV_TYPE_16X8 2 ///< 2 vectors, one per 16x8 block
  337. #define MV_TYPE_FIELD 3 ///< 2 vectors, one per field
  338. #define MV_TYPE_DMV 4 ///< 2 vectors, special mpeg2 Dual Prime Vectors
  339. /**motion vectors for a macroblock
  340. first coordinate : 0 = forward 1 = backward
  341. second " : depend on type
  342. third " : 0 = x, 1 = y
  343. */
  344. int mv[2][4][2];
  345. int field_select[2][2];
  346. int last_mv[2][2][2]; ///< last MV, used for MV prediction in MPEG1 & B-frame MPEG4
  347. uint8_t *fcode_tab; ///< smallest fcode needed for each MV
  348. int16_t direct_scale_mv[2][64]; ///< precomputed to avoid divisions in ff_mpeg4_set_direct_mv
  349. MotionEstContext me;
  350. int no_rounding; /**< apply no rounding to motion compensation (MPEG4, msmpeg4, ...)
  351. for b-frames rounding mode is always 0 */
  352. /* macroblock layer */
  353. int mb_x, mb_y;
  354. int mb_skip_run;
  355. int mb_intra;
  356. uint16_t *mb_type; ///< Table for candidate MB types for encoding (defines in mpegutils.h)
  357. int block_index[6]; ///< index to current MB in block based arrays with edges
  358. int block_wrap[6];
  359. uint8_t *dest[3];
  360. int *mb_index2xy; ///< mb_index -> mb_x + mb_y*mb_stride
  361. /** matrix transmitted in the bitstream */
  362. uint16_t intra_matrix[64];
  363. uint16_t chroma_intra_matrix[64];
  364. uint16_t inter_matrix[64];
  365. uint16_t chroma_inter_matrix[64];
  366. int intra_quant_bias; ///< bias for the quantizer
  367. int inter_quant_bias; ///< bias for the quantizer
  368. int min_qcoeff; ///< minimum encodable coefficient
  369. int max_qcoeff; ///< maximum encodable coefficient
  370. int ac_esc_length; ///< num of bits needed to encode the longest esc
  371. uint8_t *intra_ac_vlc_length;
  372. uint8_t *intra_ac_vlc_last_length;
  373. uint8_t *inter_ac_vlc_length;
  374. uint8_t *inter_ac_vlc_last_length;
  375. uint8_t *luma_dc_vlc_length;
  376. #define UNI_AC_ENC_INDEX(run,level) ((run)*128 + (level))
  377. int coded_score[8];
  378. /** precomputed matrix (combine qscale and DCT renorm) */
  379. int (*q_intra_matrix)[64];
  380. int (*q_inter_matrix)[64];
  381. /** identical to the above but for MMX & these are not permutated, second 64 entries are bias*/
  382. uint16_t (*q_intra_matrix16)[2][64];
  383. uint16_t (*q_inter_matrix16)[2][64];
  384. /* noise reduction */
  385. int (*dct_error_sum)[64];
  386. int dct_count[2];
  387. uint16_t (*dct_offset)[64];
  388. /* bit rate control */
  389. int64_t total_bits;
  390. int frame_bits; ///< bits used for the current frame
  391. int next_lambda; ///< next lambda used for retrying to encode a frame
  392. RateControlContext rc_context; ///< contains stuff only accessed in ratecontrol.c
  393. /* statistics, used for 2-pass encoding */
  394. int mv_bits;
  395. int header_bits;
  396. int i_tex_bits;
  397. int p_tex_bits;
  398. int i_count;
  399. int f_count;
  400. int b_count;
  401. int skip_count;
  402. int misc_bits; ///< cbp, mb_type
  403. int last_bits; ///< temp var used for calculating the above vars
  404. /* error concealment / resync */
  405. int resync_mb_x; ///< x position of last resync marker
  406. int resync_mb_y; ///< y position of last resync marker
  407. GetBitContext last_resync_gb; ///< used to search for the next resync marker
  408. int mb_num_left; ///< number of MBs left in this video packet (for partitioned Slices only)
  409. int next_p_frame_damaged; ///< set if the next p frame is damaged, to avoid showing trashed b frames
  410. ParseContext parse_context;
  411. /* H.263 specific */
  412. int gob_index;
  413. int obmc; ///< overlapped block motion compensation
  414. int mb_info; ///< interval for outputting info about mb offsets as side data
  415. int prev_mb_info, last_mb_info;
  416. uint8_t *mb_info_ptr;
  417. int mb_info_size;
  418. /* H.263+ specific */
  419. int umvplus; ///< == H263+ && unrestricted_mv
  420. int h263_aic_dir; ///< AIC direction: 0 = left, 1 = top
  421. int h263_slice_structured;
  422. int alt_inter_vlc; ///< alternative inter vlc
  423. int modified_quant;
  424. int loop_filter;
  425. int custom_pcf;
  426. /* mpeg4 specific */
  427. ///< number of bits to represent the fractional part of time (encoder only)
  428. int time_increment_bits;
  429. int last_time_base;
  430. int time_base; ///< time in seconds of last I,P,S Frame
  431. int64_t time; ///< time of current frame
  432. int64_t last_non_b_time;
  433. uint16_t pp_time; ///< time distance between the last 2 p,s,i frames
  434. uint16_t pb_time; ///< time distance between the last b and p,s,i frame
  435. uint16_t pp_field_time;
  436. uint16_t pb_field_time; ///< like above, just for interlaced
  437. int real_sprite_warping_points;
  438. int sprite_offset[2][2]; ///< sprite offset[isChroma][isMVY]
  439. int sprite_delta[2][2]; ///< sprite_delta [isY][isMVY]
  440. int mcsel;
  441. int quant_precision;
  442. int quarter_sample; ///< 1->qpel, 0->half pel ME/MC
  443. int aspect_ratio_info; //FIXME remove
  444. int sprite_warping_accuracy;
  445. int data_partitioning; ///< data partitioning flag from header
  446. int partitioned_frame; ///< is current frame partitioned
  447. int low_delay; ///< no reordering needed / has no b-frames
  448. int vo_type;
  449. PutBitContext tex_pb; ///< used for data partitioned VOPs
  450. PutBitContext pb2; ///< used for data partitioned VOPs
  451. int mpeg_quant;
  452. int padding_bug_score; ///< used to detect the VERY common padding bug in MPEG4
  453. /* divx specific, used to workaround (many) bugs in divx5 */
  454. int divx_packed;
  455. uint8_t *bitstream_buffer; //Divx 5.01 puts several frames in a single one, this is used to reorder them
  456. int bitstream_buffer_size;
  457. unsigned int allocated_bitstream_buffer_size;
  458. /* RV10 specific */
  459. int rv10_version; ///< RV10 version: 0 or 3
  460. int rv10_first_dc_coded[3];
  461. /* MJPEG specific */
  462. struct MJpegContext *mjpeg_ctx;
  463. /* MSMPEG4 specific */
  464. int mv_table_index;
  465. int rl_table_index;
  466. int rl_chroma_table_index;
  467. int dc_table_index;
  468. int use_skip_mb_code;
  469. int slice_height; ///< in macroblocks
  470. int first_slice_line; ///< used in mpeg4 too to handle resync markers
  471. int flipflop_rounding;
  472. int msmpeg4_version; ///< 0=not msmpeg4, 1=mp41, 2=mp42, 3=mp43/divx3 4=wmv1/7 5=wmv2/8
  473. int per_mb_rl_table;
  474. int esc3_level_length;
  475. int esc3_run_length;
  476. /** [mb_intra][isChroma][level][run][last] */
  477. int (*ac_stats)[2][MAX_LEVEL+1][MAX_RUN+1][2];
  478. int inter_intra_pred;
  479. int mspel;
  480. /* decompression specific */
  481. GetBitContext gb;
  482. /* Mpeg1 specific */
  483. int gop_picture_number; ///< index of the first picture of a GOP based on fake_pic_num & mpeg1 specific
  484. int last_mv_dir; ///< last mv_dir, used for b frame encoding
  485. uint8_t *vbv_delay_ptr; ///< pointer to vbv_delay in the bitstream
  486. /* MPEG-2-specific - I wished not to have to support this mess. */
  487. int progressive_sequence;
  488. int mpeg_f_code[2][2];
  489. // picture structure defines are loaded from mpegutils.h
  490. int picture_structure;
  491. int intra_dc_precision;
  492. int frame_pred_frame_dct;
  493. int top_field_first;
  494. int concealment_motion_vectors;
  495. int q_scale_type;
  496. int intra_vlc_format;
  497. int alternate_scan;
  498. int repeat_first_field;
  499. int chroma_420_type;
  500. int chroma_format;
  501. #define CHROMA_420 1
  502. #define CHROMA_422 2
  503. int chroma_x_shift;//depend on pix_format, that depend on chroma_format
  504. int chroma_y_shift;
  505. int progressive_frame;
  506. int full_pel[2];
  507. int interlaced_dct;
  508. int first_field; ///< is 1 for the first field of a field picture 0 otherwise
  509. int drop_frame_timecode; ///< timecode is in drop frame format.
  510. int scan_offset; ///< reserve space for SVCD scan offset user data.
  511. /* RTP specific */
  512. int rtp_mode;
  513. uint8_t *ptr_lastgob;
  514. int16_t (*pblocks[12])[64];
  515. int16_t (*block)[64]; ///< points to one of the following blocks
  516. int16_t (*blocks)[12][64]; // for HQ mode we need to keep the best block
  517. int (*decode_mb)(struct MpegEncContext *s, int16_t block[6][64]); // used by some codecs to avoid a switch()
  518. #define SLICE_OK 0
  519. #define SLICE_ERROR -1
  520. #define SLICE_END -2 ///<end marker found
  521. #define SLICE_NOEND -3 ///<no end marker or error found but mb count exceeded
  522. void (*dct_unquantize_mpeg1_intra)(struct MpegEncContext *s,
  523. int16_t *block/*align 16*/, int n, int qscale);
  524. void (*dct_unquantize_mpeg1_inter)(struct MpegEncContext *s,
  525. int16_t *block/*align 16*/, int n, int qscale);
  526. void (*dct_unquantize_mpeg2_intra)(struct MpegEncContext *s,
  527. int16_t *block/*align 16*/, int n, int qscale);
  528. void (*dct_unquantize_mpeg2_inter)(struct MpegEncContext *s,
  529. int16_t *block/*align 16*/, int n, int qscale);
  530. void (*dct_unquantize_h263_intra)(struct MpegEncContext *s,
  531. int16_t *block/*align 16*/, int n, int qscale);
  532. void (*dct_unquantize_h263_inter)(struct MpegEncContext *s,
  533. int16_t *block/*align 16*/, int n, int qscale);
  534. void (*dct_unquantize_intra)(struct MpegEncContext *s, // unquantizer to use (mpeg4 can use both)
  535. int16_t *block/*align 16*/, int n, int qscale);
  536. void (*dct_unquantize_inter)(struct MpegEncContext *s, // unquantizer to use (mpeg4 can use both)
  537. int16_t *block/*align 16*/, int n, int qscale);
  538. int (*dct_quantize)(struct MpegEncContext *s, int16_t *block/*align 16*/, int n, int qscale, int *overflow);
  539. int (*fast_dct_quantize)(struct MpegEncContext *s, int16_t *block/*align 16*/, int n, int qscale, int *overflow);
  540. void (*denoise_dct)(struct MpegEncContext *s, int16_t *block);
  541. int mpv_flags; ///< flags set by private options
  542. int quantizer_noise_shaping;
  543. /**
  544. * ratecontrol qmin qmax limiting method
  545. * 0-> clipping, 1-> use a nice continuous function to limit qscale wthin qmin/qmax.
  546. */
  547. float rc_qsquish;
  548. float rc_qmod_amp;
  549. int rc_qmod_freq;
  550. float rc_initial_cplx;
  551. float rc_buffer_aggressivity;
  552. float border_masking;
  553. int lmin, lmax;
  554. char *rc_eq;
  555. /* temp buffers for rate control */
  556. float *cplx_tab, *bits_tab;
  557. /* flag to indicate a reinitialization is required, e.g. after
  558. * a frame size change */
  559. int context_reinit;
  560. ERContext er;
  561. int error_rate;
  562. /* temporary frames used by b_frame_strategy = 2 */
  563. AVFrame *tmp_frames[MAX_B_FRAMES + 2];
  564. } MpegEncContext;
  565. /* mpegvideo_enc common options */
  566. #define FF_MPV_FLAG_SKIP_RD 0x0001
  567. #define FF_MPV_FLAG_STRICT_GOP 0x0002
  568. #define FF_MPV_FLAG_QP_RD 0x0004
  569. #define FF_MPV_FLAG_CBP_RD 0x0008
  570. #define FF_MPV_FLAG_NAQ 0x0010
  571. #define FF_MPV_FLAG_MV0 0x0020
  572. #define FF_MPV_OFFSET(x) offsetof(MpegEncContext, x)
  573. #define FF_MPV_OPT_FLAGS (AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM)
  574. #define FF_MPV_COMMON_OPTS \
  575. { "mpv_flags", "Flags common for all mpegvideo-based encoders.", FF_MPV_OFFSET(mpv_flags), AV_OPT_TYPE_FLAGS, { .i64 = 0 }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS, "mpv_flags" },\
  576. { "skip_rd", "RD optimal MB level residual skipping", 0, AV_OPT_TYPE_CONST, { .i64 = FF_MPV_FLAG_SKIP_RD }, 0, 0, FF_MPV_OPT_FLAGS, "mpv_flags" },\
  577. { "strict_gop", "Strictly enforce gop size", 0, AV_OPT_TYPE_CONST, { .i64 = FF_MPV_FLAG_STRICT_GOP }, 0, 0, FF_MPV_OPT_FLAGS, "mpv_flags" },\
  578. { "qp_rd", "Use rate distortion optimization for qp selection", 0, AV_OPT_TYPE_CONST, { .i64 = FF_MPV_FLAG_QP_RD }, 0, 0, FF_MPV_OPT_FLAGS, "mpv_flags" },\
  579. { "cbp_rd", "use rate distortion optimization for CBP", 0, AV_OPT_TYPE_CONST, { .i64 = FF_MPV_FLAG_CBP_RD }, 0, 0, FF_MPV_OPT_FLAGS, "mpv_flags" },\
  580. { "naq", "normalize adaptive quantization", 0, AV_OPT_TYPE_CONST, { .i64 = FF_MPV_FLAG_NAQ }, 0, 0, FF_MPV_OPT_FLAGS, "mpv_flags" },\
  581. { "mv0", "always try a mb with mv=<0,0>", 0, AV_OPT_TYPE_CONST, { .i64 = FF_MPV_FLAG_MV0 }, 0, 0, FF_MPV_OPT_FLAGS, "mpv_flags" },\
  582. { "luma_elim_threshold", "single coefficient elimination threshold for luminance (negative values also consider dc coefficient)",\
  583. FF_MPV_OFFSET(luma_elim_threshold), AV_OPT_TYPE_INT, { .i64 = 0 }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS },\
  584. { "chroma_elim_threshold", "single coefficient elimination threshold for chrominance (negative values also consider dc coefficient)",\
  585. FF_MPV_OFFSET(chroma_elim_threshold), AV_OPT_TYPE_INT, { .i64 = 0 }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS },\
  586. { "quantizer_noise_shaping", NULL, FF_MPV_OFFSET(quantizer_noise_shaping), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, INT_MAX, FF_MPV_OPT_FLAGS },\
  587. { "error_rate", "Simulate errors in the bitstream to test error concealment.", \
  588. FF_MPV_OFFSET(error_rate), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, INT_MAX, FF_MPV_OPT_FLAGS },\
  589. {"qsquish", "how to keep quantizer between qmin and qmax (0 = clip, 1 = use differentiable function)", \
  590. FF_MPV_OFFSET(rc_qsquish), AV_OPT_TYPE_FLOAT, {.dbl = 0 }, 0, 99, FF_MPV_OPT_FLAGS}, \
  591. {"rc_qmod_amp", "experimental quantizer modulation", FF_MPV_OFFSET(rc_qmod_amp), AV_OPT_TYPE_FLOAT, {.dbl = 0 }, -FLT_MAX, FLT_MAX, FF_MPV_OPT_FLAGS}, \
  592. {"rc_qmod_freq", "experimental quantizer modulation", FF_MPV_OFFSET(rc_qmod_freq), AV_OPT_TYPE_INT, {.i64 = 0 }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS}, \
  593. {"rc_eq", "Set rate control equation. When computing the expression, besides the standard functions " \
  594. "defined in the section 'Expression Evaluation', the following functions are available: " \
  595. "bits2qp(bits), qp2bits(qp). Also the following constants are available: iTex pTex tex mv " \
  596. "fCode iCount mcVar var isI isP isB avgQP qComp avgIITex avgPITex avgPPTex avgBPTex avgTex.", \
  597. FF_MPV_OFFSET(rc_eq), AV_OPT_TYPE_STRING, .flags = FF_MPV_OPT_FLAGS }, \
  598. {"rc_init_cplx", "initial complexity for 1-pass encoding", FF_MPV_OFFSET(rc_initial_cplx), AV_OPT_TYPE_FLOAT, {.dbl = 0 }, -FLT_MAX, FLT_MAX, FF_MPV_OPT_FLAGS}, \
  599. {"rc_buf_aggressivity", "currently useless", FF_MPV_OFFSET(rc_buffer_aggressivity), AV_OPT_TYPE_FLOAT, {.dbl = 1.0 }, -FLT_MAX, FLT_MAX, FF_MPV_OPT_FLAGS}, \
  600. {"border_mask", "increase the quantizer for macroblocks close to borders", FF_MPV_OFFSET(border_masking), AV_OPT_TYPE_FLOAT, {.dbl = 0 }, -FLT_MAX, FLT_MAX, FF_MPV_OPT_FLAGS}, \
  601. {"lmin", "minimum Lagrange factor (VBR)", FF_MPV_OFFSET(lmin), AV_OPT_TYPE_INT, {.i64 = 2*FF_QP2LAMBDA }, 0, INT_MAX, FF_MPV_OPT_FLAGS }, \
  602. {"lmax", "maximum Lagrange factor (VBR)", FF_MPV_OFFSET(lmax), AV_OPT_TYPE_INT, {.i64 = 31*FF_QP2LAMBDA }, 0, INT_MAX, FF_MPV_OPT_FLAGS }, \
  603. extern const AVOption ff_mpv_generic_options[];
  604. #define FF_MPV_GENERIC_CLASS(name) \
  605. static const AVClass name ## _class = {\
  606. .class_name = #name " encoder",\
  607. .item_name = av_default_item_name,\
  608. .option = ff_mpv_generic_options,\
  609. .version = LIBAVUTIL_VERSION_INT,\
  610. };
  611. /**
  612. * Set the given MpegEncContext to common defaults (same for encoding
  613. * and decoding). The changed fields will not depend upon the prior
  614. * state of the MpegEncContext.
  615. */
  616. void ff_mpv_common_defaults(MpegEncContext *s);
  617. int ff_mpv_common_init(MpegEncContext *s);
  618. void ff_mpv_common_init_arm(MpegEncContext *s);
  619. void ff_mpv_common_init_neon(MpegEncContext *s);
  620. void ff_mpv_common_init_ppc(MpegEncContext *s);
  621. void ff_mpv_common_init_x86(MpegEncContext *s);
  622. int ff_mpv_common_frame_size_change(MpegEncContext *s);
  623. void ff_mpv_common_end(MpegEncContext *s);
  624. void ff_mpv_decode_defaults(MpegEncContext *s);
  625. void ff_mpv_decode_mb(MpegEncContext *s, int16_t block[12][64]);
  626. void ff_mpv_report_decode_progress(MpegEncContext *s);
  627. int ff_mpv_frame_start(MpegEncContext *s, AVCodecContext *avctx);
  628. void ff_mpv_frame_end(MpegEncContext *s);
  629. int ff_mpv_lowest_referenced_row(MpegEncContext *s, int dir);
  630. int ff_mpv_encode_init(AVCodecContext *avctx);
  631. void ff_mpv_encode_init_x86(MpegEncContext *s);
  632. int ff_mpv_encode_end(AVCodecContext *avctx);
  633. int ff_mpv_encode_picture(AVCodecContext *avctx, AVPacket *pkt,
  634. const AVFrame *frame, int *got_packet);
  635. void ff_clean_intra_table_entries(MpegEncContext *s);
  636. void ff_mpeg_draw_horiz_band(MpegEncContext *s, int y, int h);
  637. void ff_mpeg_flush(AVCodecContext *avctx);
  638. void ff_print_debug_info(MpegEncContext *s, Picture *p);
  639. void ff_write_quant_matrix(PutBitContext *pb, uint16_t *matrix);
  640. int ff_find_unused_picture(AVCodecContext *avctx, Picture *picture, int shared);
  641. int ff_update_duplicate_context(MpegEncContext *dst, MpegEncContext *src);
  642. int ff_mpeg_update_thread_context(AVCodecContext *dst, const AVCodecContext *src);
  643. void ff_set_qscale(MpegEncContext * s, int qscale);
  644. void ff_mpv_idct_init(MpegEncContext *s);
  645. void ff_convert_matrix(MpegEncContext *s, int (*qmat)[64], uint16_t (*qmat16)[2][64],
  646. const uint16_t *quant_matrix, int bias, int qmin, int qmax, int intra);
  647. int ff_dct_quantize_c(MpegEncContext *s, int16_t *block, int n, int qscale, int *overflow);
  648. void ff_init_block_index(MpegEncContext *s);
  649. void ff_mpv_motion(MpegEncContext *s,
  650. uint8_t *dest_y, uint8_t *dest_cb,
  651. uint8_t *dest_cr, int dir,
  652. uint8_t **ref_picture,
  653. op_pixels_func (*pix_op)[4],
  654. qpel_mc_func (*qpix_op)[16]);
  655. /**
  656. * Allocate a Picture.
  657. * The pixels are allocated/set by calling get_buffer() if shared = 0.
  658. */
  659. int ff_alloc_picture(MpegEncContext *s, Picture *pic, int shared);
  660. /**
  661. * permute block according to permuatation.
  662. * @param last last non zero element in scantable order
  663. */
  664. void ff_block_permute(int16_t *block, uint8_t *permutation, const uint8_t *scantable, int last);
  665. static inline void ff_update_block_index(MpegEncContext *s){
  666. const int block_size = 8;
  667. s->block_index[0]+=2;
  668. s->block_index[1]+=2;
  669. s->block_index[2]+=2;
  670. s->block_index[3]+=2;
  671. s->block_index[4]++;
  672. s->block_index[5]++;
  673. s->dest[0]+= 2*block_size;
  674. s->dest[1]+= block_size;
  675. s->dest[2]+= block_size;
  676. }
  677. static inline int get_bits_diff(MpegEncContext *s){
  678. const int bits= put_bits_count(&s->pb);
  679. const int last= s->last_bits;
  680. s->last_bits = bits;
  681. return bits - last;
  682. }
  683. static inline int ff_h263_round_chroma(int x){
  684. static const uint8_t h263_chroma_roundtab[16] = {
  685. // 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
  686. 0, 0, 0, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 1, 1,
  687. };
  688. return h263_chroma_roundtab[x & 0xf] + (x >> 3);
  689. }
  690. /* motion_est.c */
  691. void ff_estimate_p_frame_motion(MpegEncContext * s,
  692. int mb_x, int mb_y);
  693. void ff_estimate_b_frame_motion(MpegEncContext * s,
  694. int mb_x, int mb_y);
  695. int ff_get_best_fcode(MpegEncContext * s, int16_t (*mv_table)[2], int type);
  696. void ff_fix_long_p_mvs(MpegEncContext * s);
  697. void ff_fix_long_mvs(MpegEncContext * s, uint8_t *field_select_table, int field_select,
  698. int16_t (*mv_table)[2], int f_code, int type, int truncate);
  699. int ff_init_me(MpegEncContext *s);
  700. int ff_pre_estimate_p_frame_motion(MpegEncContext * s, int mb_x, int mb_y);
  701. int ff_epzs_motion_search(MpegEncContext * s, int *mx_ptr, int *my_ptr,
  702. int P[10][2], int src_index, int ref_index, int16_t (*last_mv)[2],
  703. int ref_mv_scale, int size, int h);
  704. int ff_get_mb_score(MpegEncContext * s, int mx, int my, int src_index,
  705. int ref_index, int size, int h, int add_rate);
  706. /* mpeg12.c */
  707. extern const uint8_t ff_mpeg1_dc_scale_table[128];
  708. extern const uint8_t * const ff_mpeg2_dc_scale_table[4];
  709. void ff_mpeg1_encode_picture_header(MpegEncContext *s, int picture_number);
  710. void ff_mpeg1_encode_mb(MpegEncContext *s,
  711. int16_t block[8][64],
  712. int motion_x, int motion_y);
  713. void ff_mpeg1_encode_init(MpegEncContext *s);
  714. void ff_mpeg1_encode_slice_header(MpegEncContext *s);
  715. extern const uint8_t ff_aic_dc_scale_table[32];
  716. extern const uint8_t ff_h263_chroma_qscale_table[32];
  717. /* rv10.c */
  718. int ff_rv10_encode_picture_header(MpegEncContext *s, int picture_number);
  719. int ff_rv_decode_dc(MpegEncContext *s, int n);
  720. void ff_rv20_encode_picture_header(MpegEncContext *s, int picture_number);
  721. /* msmpeg4.c */
  722. void ff_msmpeg4_encode_picture_header(MpegEncContext * s, int picture_number);
  723. void ff_msmpeg4_encode_ext_header(MpegEncContext * s);
  724. void ff_msmpeg4_encode_mb(MpegEncContext * s,
  725. int16_t block[6][64],
  726. int motion_x, int motion_y);
  727. int ff_msmpeg4_decode_picture_header(MpegEncContext * s);
  728. int ff_msmpeg4_decode_ext_header(MpegEncContext * s, int buf_size);
  729. int ff_msmpeg4_decode_init(AVCodecContext *avctx);
  730. int ff_msmpeg4_encode_init(MpegEncContext *s);
  731. int ff_wmv2_decode_picture_header(MpegEncContext * s);
  732. int ff_wmv2_decode_secondary_picture_header(MpegEncContext * s);
  733. void ff_wmv2_add_mb(MpegEncContext *s, int16_t block[6][64], uint8_t *dest_y, uint8_t *dest_cb, uint8_t *dest_cr);
  734. void ff_mspel_motion(MpegEncContext *s,
  735. uint8_t *dest_y, uint8_t *dest_cb, uint8_t *dest_cr,
  736. uint8_t **ref_picture, op_pixels_func (*pix_op)[4],
  737. int motion_x, int motion_y, int h);
  738. int ff_wmv2_encode_picture_header(MpegEncContext * s, int picture_number);
  739. void ff_wmv2_encode_mb(MpegEncContext * s,
  740. int16_t block[6][64],
  741. int motion_x, int motion_y);
  742. int ff_mpeg_ref_picture(AVCodecContext *avctx, Picture *dst, Picture *src);
  743. void ff_mpeg_unref_picture(AVCodecContext *avctx, Picture *picture);
  744. void ff_free_picture_tables(Picture *pic);
  745. #endif /* AVCODEC_MPEGVIDEO_H */