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.

704 lines
35KB

  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 "internal.h"
  38. #include "me_cmp.h"
  39. #include "motion_est.h"
  40. #include "mpegpicture.h"
  41. #include "mpegvideodsp.h"
  42. #include "mpegvideoencdsp.h"
  43. #include "pixblockdsp.h"
  44. #include "put_bits.h"
  45. #include "ratecontrol.h"
  46. #include "parser.h"
  47. #include "mpegutils.h"
  48. #include "mpeg12data.h"
  49. #include "qpeldsp.h"
  50. #include "thread.h"
  51. #include "videodsp.h"
  52. #include "libavutil/opt.h"
  53. #define MAX_THREADS 16
  54. #define MAX_B_FRAMES 16
  55. /* Start codes. */
  56. #define SEQ_END_CODE 0x000001b7
  57. #define SEQ_START_CODE 0x000001b3
  58. #define GOP_START_CODE 0x000001b8
  59. #define PICTURE_START_CODE 0x00000100
  60. #define SLICE_MIN_START_CODE 0x00000101
  61. #define SLICE_MAX_START_CODE 0x000001af
  62. #define EXT_START_CODE 0x000001b5
  63. #define USER_START_CODE 0x000001b2
  64. /**
  65. * MpegEncContext.
  66. */
  67. typedef struct MpegEncContext {
  68. AVClass *class;
  69. int y_dc_scale, c_dc_scale;
  70. int ac_pred;
  71. int block_last_index[12]; ///< last non zero coefficient in block
  72. int h263_aic; ///< Advanced INTRA Coding (AIC)
  73. /* scantables */
  74. ScanTable inter_scantable; ///< if inter == intra then intra should be used to reduce the cache usage
  75. ScanTable intra_scantable;
  76. ScanTable intra_h_scantable;
  77. ScanTable intra_v_scantable;
  78. /* WARNING: changes above this line require updates to hardcoded
  79. * offsets used in ASM. */
  80. struct AVCodecContext *avctx;
  81. /* the following parameters must be initialized before encoding */
  82. int width, height;///< picture size. must be a multiple of 16
  83. int gop_size;
  84. int intra_only; ///< if true, only intra pictures are generated
  85. int bit_rate; ///< wanted bit rate
  86. enum OutputFormat out_format; ///< output format
  87. int h263_pred; ///< use MPEG-4/H.263 ac/dc predictions
  88. int pb_frame; ///< PB-frame mode (0 = none, 1 = base, 2 = improved)
  89. /* the following codec id fields are deprecated in favor of codec_id */
  90. int h263_plus; ///< H.263+ headers
  91. int h263_flv; ///< use flv H.263 header
  92. enum AVCodecID codec_id; /* see AV_CODEC_ID_xxx */
  93. int fixed_qscale; ///< fixed qscale if non zero
  94. int encoding; ///< true if we are encoding (vs decoding)
  95. int max_b_frames; ///< max number of B-frames for encoding
  96. int luma_elim_threshold;
  97. int chroma_elim_threshold;
  98. int strict_std_compliance; ///< strictly follow the std (MPEG-4, ...)
  99. int workaround_bugs; ///< workaround bugs in encoders which cannot be detected automatically
  100. int codec_tag; ///< internal codec_tag upper case converted from avctx codec_tag
  101. /* the following fields are managed internally by the encoder */
  102. /* sequence parameters */
  103. int context_initialized;
  104. int input_picture_number; ///< used to set pic->display_picture_number, should not be used for/by anything else
  105. int coded_picture_number; ///< used to set pic->coded_picture_number, should not be used for/by anything else
  106. int picture_number; //FIXME remove, unclear definition
  107. int picture_in_gop_number; ///< 0-> first pic in gop, ...
  108. int mb_width, mb_height; ///< number of MBs horizontally & vertically
  109. int mb_stride; ///< mb_width+1 used for some arrays to allow simple addressing of left & top MBs without sig11
  110. int b8_stride; ///< 2*mb_width+1 used for some 8x8 block arrays to allow simple addressing
  111. int h_edge_pos, v_edge_pos;///< horizontal / vertical position of the right/bottom edge (pixel replication)
  112. int mb_num; ///< number of MBs of a picture
  113. ptrdiff_t linesize; ///< line size, in bytes, may be different from width
  114. ptrdiff_t uvlinesize; ///< line size, for chroma in bytes, may be different from width
  115. Picture *picture; ///< main picture buffer
  116. Picture **input_picture; ///< next pictures on display order for encoding
  117. Picture **reordered_input_picture; ///< pointer to the next pictures in coded order for encoding
  118. int64_t user_specified_pts; ///< last non-zero pts from AVFrame which was passed into avcodec_encode_video2()
  119. /**
  120. * pts difference between the first and second input frame, used for
  121. * calculating dts of the first frame when there's a delay */
  122. int64_t dts_delta;
  123. /**
  124. * reordered pts to be used as dts for the next output frame when there's
  125. * a delay */
  126. int64_t reordered_pts;
  127. /** bit output */
  128. PutBitContext pb;
  129. int start_mb_y; ///< start mb_y of this thread (so current thread should process start_mb_y <= row < end_mb_y)
  130. int end_mb_y; ///< end mb_y of this thread (so current thread should process start_mb_y <= row < end_mb_y)
  131. struct MpegEncContext *thread_context[MAX_THREADS];
  132. int slice_context_count; ///< number of used thread_contexts
  133. /**
  134. * copy of the previous picture structure.
  135. * note, linesize & data, might not match the previous picture (for field pictures)
  136. */
  137. Picture last_picture;
  138. /**
  139. * copy of the next picture structure.
  140. * note, linesize & data, might not match the next picture (for field pictures)
  141. */
  142. Picture next_picture;
  143. /**
  144. * copy of the source picture structure for encoding.
  145. * note, linesize & data, might not match the source picture (for field pictures)
  146. */
  147. Picture new_picture;
  148. /**
  149. * copy of the current picture structure.
  150. * note, linesize & data, might not match the current picture (for field pictures)
  151. */
  152. Picture current_picture; ///< buffer to store the decompressed current picture
  153. Picture *last_picture_ptr; ///< pointer to the previous picture.
  154. Picture *next_picture_ptr; ///< pointer to the next picture (for bidir pred)
  155. Picture *current_picture_ptr; ///< pointer to the current picture
  156. int last_dc[3]; ///< last DC values for MPEG-1
  157. int16_t *dc_val_base;
  158. int16_t *dc_val[3]; ///< used for MPEG-4 DC prediction, all 3 arrays must be continuous
  159. const uint8_t *y_dc_scale_table; ///< qscale -> y_dc_scale table
  160. const uint8_t *c_dc_scale_table; ///< qscale -> c_dc_scale table
  161. const uint8_t *chroma_qscale_table; ///< qscale -> chroma_qscale (H.263)
  162. uint8_t *coded_block_base;
  163. uint8_t *coded_block; ///< used for coded block pattern prediction (msmpeg4v3, wmv1)
  164. int16_t (*ac_val_base)[16];
  165. int16_t (*ac_val[3])[16]; ///< used for for MPEG-4 AC prediction, all 3 arrays must be continuous
  166. int mb_skipped; ///< MUST BE SET only during DECODING
  167. uint8_t *mbskip_table; /**< used to avoid copy if macroblock skipped (for black regions for example)
  168. and used for B-frame encoding & decoding (contains skip table of next P-frame) */
  169. uint8_t *mbintra_table; ///< used to avoid setting {ac, dc, cbp}-pred stuff to zero on inter MB decoding
  170. uint8_t *cbp_table; ///< used to store cbp, ac_pred for partitioned decoding
  171. uint8_t *pred_dir_table; ///< used to store pred_dir for partitioned decoding
  172. ScratchpadContext sc;
  173. int qscale; ///< QP
  174. int chroma_qscale; ///< chroma QP
  175. unsigned int lambda; ///< Lagrange multiplier used in rate distortion
  176. unsigned int lambda2; ///< (lambda*lambda) >> FF_LAMBDA_SHIFT
  177. int *lambda_table;
  178. int adaptive_quant; ///< use adaptive quantization
  179. int dquant; ///< qscale difference to prev qscale
  180. int pict_type; ///< AV_PICTURE_TYPE_I, AV_PICTURE_TYPE_P, AV_PICTURE_TYPE_B, ...
  181. int last_pict_type; //FIXME removes
  182. int last_non_b_pict_type; ///< used for MPEG-4 gmc B-frames & ratecontrol
  183. int droppable;
  184. int frame_rate_index;
  185. int last_lambda_for[5]; ///< last lambda for a specific pict type
  186. int skipdct; ///< skip dct and code zero residual
  187. /* motion compensation */
  188. int unrestricted_mv; ///< mv can point outside of the coded picture
  189. int h263_long_vectors; ///< use horrible H.263v1 long vector mode
  190. BlockDSPContext bdsp;
  191. FDCTDSPContext fdsp;
  192. HpelDSPContext hdsp;
  193. IDCTDSPContext idsp;
  194. MECmpContext mecc;
  195. MpegVideoDSPContext mdsp;
  196. MpegvideoEncDSPContext mpvencdsp;
  197. PixblockDSPContext pdsp;
  198. QpelDSPContext qdsp;
  199. VideoDSPContext vdsp;
  200. H263DSPContext h263dsp;
  201. int f_code; ///< forward MV resolution
  202. int b_code; ///< backward MV resolution for B-frames (MPEG-4)
  203. int16_t (*p_mv_table_base)[2];
  204. int16_t (*b_forw_mv_table_base)[2];
  205. int16_t (*b_back_mv_table_base)[2];
  206. int16_t (*b_bidir_forw_mv_table_base)[2];
  207. int16_t (*b_bidir_back_mv_table_base)[2];
  208. int16_t (*b_direct_mv_table_base)[2];
  209. int16_t (*p_field_mv_table_base[2][2])[2];
  210. int16_t (*b_field_mv_table_base[2][2][2])[2];
  211. int16_t (*p_mv_table)[2]; ///< MV table (1MV per MB) P-frame encoding
  212. int16_t (*b_forw_mv_table)[2]; ///< MV table (1MV per MB) forward mode B-frame encoding
  213. int16_t (*b_back_mv_table)[2]; ///< MV table (1MV per MB) backward mode B-frame encoding
  214. int16_t (*b_bidir_forw_mv_table)[2]; ///< MV table (1MV per MB) bidir mode B-frame encoding
  215. int16_t (*b_bidir_back_mv_table)[2]; ///< MV table (1MV per MB) bidir mode B-frame encoding
  216. int16_t (*b_direct_mv_table)[2]; ///< MV table (1MV per MB) direct mode B-frame encoding
  217. int16_t (*p_field_mv_table[2][2])[2]; ///< MV table (2MV per MB) interlaced P-frame encoding
  218. int16_t (*b_field_mv_table[2][2][2])[2];///< MV table (4MV per MB) interlaced B-frame encoding
  219. uint8_t (*p_field_select_table[2]);
  220. uint8_t (*b_field_select_table[2][2]);
  221. #if FF_API_MOTION_EST
  222. int me_method; ///< ME algorithm
  223. #endif
  224. int motion_est; ///< ME algorithm
  225. int me_penalty_compensation;
  226. int me_pre; ///< prepass for motion estimation
  227. int mv_dir;
  228. #define MV_DIR_FORWARD 1
  229. #define MV_DIR_BACKWARD 2
  230. #define MV_DIRECT 4 ///< bidirectional mode where the difference equals the MV of the last P/S/I-Frame (MPEG-4)
  231. int mv_type;
  232. #define MV_TYPE_16X16 0 ///< 1 vector for the whole mb
  233. #define MV_TYPE_8X8 1 ///< 4 vectors (H.263, MPEG-4 4MV)
  234. #define MV_TYPE_16X8 2 ///< 2 vectors, one per 16x8 block
  235. #define MV_TYPE_FIELD 3 ///< 2 vectors, one per field
  236. #define MV_TYPE_DMV 4 ///< 2 vectors, special mpeg2 Dual Prime Vectors
  237. /**motion vectors for a macroblock
  238. first coordinate : 0 = forward 1 = backward
  239. second " : depend on type
  240. third " : 0 = x, 1 = y
  241. */
  242. int mv[2][4][2];
  243. int field_select[2][2];
  244. int last_mv[2][2][2]; ///< last MV, used for MV prediction in MPEG-1 & B-frame MPEG-4
  245. uint8_t *fcode_tab; ///< smallest fcode needed for each MV
  246. int16_t direct_scale_mv[2][64]; ///< precomputed to avoid divisions in ff_mpeg4_set_direct_mv
  247. MotionEstContext me;
  248. int no_rounding; /**< apply no rounding to motion compensation (MPEG-4, msmpeg4, ...)
  249. for B-frames rounding mode is always 0 */
  250. /* macroblock layer */
  251. int mb_x, mb_y;
  252. int mb_skip_run;
  253. int mb_intra;
  254. uint16_t *mb_type; ///< Table for candidate MB types for encoding (defines in mpegutils.h)
  255. int block_index[6]; ///< index to current MB in block based arrays with edges
  256. int block_wrap[6];
  257. uint8_t *dest[3];
  258. int *mb_index2xy; ///< mb_index -> mb_x + mb_y*mb_stride
  259. /** matrix transmitted in the bitstream */
  260. uint16_t intra_matrix[64];
  261. uint16_t chroma_intra_matrix[64];
  262. uint16_t inter_matrix[64];
  263. uint16_t chroma_inter_matrix[64];
  264. int intra_quant_bias; ///< bias for the quantizer
  265. int inter_quant_bias; ///< bias for the quantizer
  266. int min_qcoeff; ///< minimum encodable coefficient
  267. int max_qcoeff; ///< maximum encodable coefficient
  268. int ac_esc_length; ///< num of bits needed to encode the longest esc
  269. uint8_t *intra_ac_vlc_length;
  270. uint8_t *intra_ac_vlc_last_length;
  271. uint8_t *inter_ac_vlc_length;
  272. uint8_t *inter_ac_vlc_last_length;
  273. uint8_t *luma_dc_vlc_length;
  274. #define UNI_AC_ENC_INDEX(run,level) ((run)*128 + (level))
  275. int coded_score[8];
  276. /** precomputed matrix (combine qscale and DCT renorm) */
  277. int (*q_intra_matrix)[64];
  278. int (*q_inter_matrix)[64];
  279. /** identical to the above but for MMX & these are not permutated, second 64 entries are bias*/
  280. uint16_t (*q_intra_matrix16)[2][64];
  281. uint16_t (*q_inter_matrix16)[2][64];
  282. /* noise reduction */
  283. int (*dct_error_sum)[64];
  284. int dct_count[2];
  285. uint16_t (*dct_offset)[64];
  286. /* bit rate control */
  287. int64_t total_bits;
  288. int frame_bits; ///< bits used for the current frame
  289. int next_lambda; ///< next lambda used for retrying to encode a frame
  290. RateControlContext rc_context; ///< contains stuff only accessed in ratecontrol.c
  291. /* statistics, used for 2-pass encoding */
  292. int mv_bits;
  293. int header_bits;
  294. int i_tex_bits;
  295. int p_tex_bits;
  296. int i_count;
  297. int f_count;
  298. int b_count;
  299. int skip_count;
  300. int misc_bits; ///< cbp, mb_type
  301. int last_bits; ///< temp var used for calculating the above vars
  302. /* error concealment / resync */
  303. int resync_mb_x; ///< x position of last resync marker
  304. int resync_mb_y; ///< y position of last resync marker
  305. GetBitContext last_resync_gb; ///< used to search for the next resync marker
  306. int mb_num_left; ///< number of MBs left in this video packet (for partitioned Slices only)
  307. int next_p_frame_damaged; ///< set if the next p frame is damaged, to avoid showing trashed B-frames
  308. ParseContext parse_context;
  309. /* H.263 specific */
  310. int gob_index;
  311. int obmc; ///< overlapped block motion compensation
  312. int mb_info; ///< interval for outputting info about mb offsets as side data
  313. int prev_mb_info, last_mb_info;
  314. uint8_t *mb_info_ptr;
  315. int mb_info_size;
  316. int rc_strategy;
  317. /* H.263+ specific */
  318. int umvplus; ///< == H.263+ && unrestricted_mv
  319. int h263_aic_dir; ///< AIC direction: 0 = left, 1 = top
  320. int h263_slice_structured;
  321. int alt_inter_vlc; ///< alternative inter vlc
  322. int modified_quant;
  323. int loop_filter;
  324. int custom_pcf;
  325. /* MPEG-4 specific */
  326. ///< number of bits to represent the fractional part of time (encoder only)
  327. int time_increment_bits;
  328. int last_time_base;
  329. int time_base; ///< time in seconds of last I,P,S Frame
  330. int64_t time; ///< time of current frame
  331. int64_t last_non_b_time;
  332. uint16_t pp_time; ///< time distance between the last 2 p,s,i frames
  333. uint16_t pb_time; ///< time distance between the last b and p,s,i frame
  334. uint16_t pp_field_time;
  335. uint16_t pb_field_time; ///< like above, just for interlaced
  336. int real_sprite_warping_points;
  337. int sprite_offset[2][2]; ///< sprite offset[isChroma][isMVY]
  338. int sprite_delta[2][2]; ///< sprite_delta [isY][isMVY]
  339. int mcsel;
  340. int quant_precision;
  341. int quarter_sample; ///< 1->qpel, 0->half pel ME/MC
  342. int aspect_ratio_info; //FIXME remove
  343. int sprite_warping_accuracy;
  344. int data_partitioning; ///< data partitioning flag from header
  345. int partitioned_frame; ///< is current frame partitioned
  346. int low_delay; ///< no reordering needed / has no B-frames
  347. int vo_type;
  348. PutBitContext tex_pb; ///< used for data partitioned VOPs
  349. PutBitContext pb2; ///< used for data partitioned VOPs
  350. int mpeg_quant;
  351. int padding_bug_score; ///< used to detect the VERY common padding bug in MPEG-4
  352. /* divx specific, used to workaround (many) bugs in divx5 */
  353. int divx_packed;
  354. uint8_t *bitstream_buffer; //Divx 5.01 puts several frames in a single one, this is used to reorder them
  355. int bitstream_buffer_size;
  356. unsigned int allocated_bitstream_buffer_size;
  357. /* RV10 specific */
  358. int rv10_version; ///< RV10 version: 0 or 3
  359. int rv10_first_dc_coded[3];
  360. /* MJPEG specific */
  361. struct MJpegContext *mjpeg_ctx;
  362. int pred;
  363. /* MSMPEG4 specific */
  364. int mv_table_index;
  365. int rl_table_index;
  366. int rl_chroma_table_index;
  367. int dc_table_index;
  368. int use_skip_mb_code;
  369. int slice_height; ///< in macroblocks
  370. int first_slice_line; ///< used in MPEG-4 too to handle resync markers
  371. int flipflop_rounding;
  372. int msmpeg4_version; ///< 0=not msmpeg4, 1=mp41, 2=mp42, 3=mp43/divx3 4=wmv1/7 5=wmv2/8
  373. int per_mb_rl_table;
  374. int esc3_level_length;
  375. int esc3_run_length;
  376. /** [mb_intra][isChroma][level][run][last] */
  377. int (*ac_stats)[2][MAX_LEVEL+1][MAX_RUN+1][2];
  378. int inter_intra_pred;
  379. int mspel;
  380. /* decompression specific */
  381. GetBitContext gb;
  382. /* MPEG-1 specific */
  383. int gop_picture_number; ///< index of the first picture of a GOP based on fake_pic_num & MPEG-1 specific
  384. int last_mv_dir; ///< last mv_dir, used for B-frame encoding
  385. uint8_t *vbv_delay_ptr; ///< pointer to vbv_delay in the bitstream
  386. /* MPEG-2-specific - I wished not to have to support this mess. */
  387. int progressive_sequence;
  388. int mpeg_f_code[2][2];
  389. // picture structure defines are loaded from mpegutils.h
  390. int picture_structure;
  391. int64_t timecode_frame_start; ///< GOP timecode frame start number, in non drop frame format
  392. int intra_dc_precision;
  393. int frame_pred_frame_dct;
  394. int top_field_first;
  395. int concealment_motion_vectors;
  396. int q_scale_type;
  397. int brd_scale;
  398. int intra_vlc_format;
  399. int alternate_scan;
  400. int repeat_first_field;
  401. int chroma_420_type;
  402. int chroma_format;
  403. #define CHROMA_420 1
  404. #define CHROMA_422 2
  405. int chroma_x_shift;//depend on pix_format, that depend on chroma_format
  406. int chroma_y_shift;
  407. int progressive_frame;
  408. int full_pel[2];
  409. int interlaced_dct;
  410. int first_field; ///< is 1 for the first field of a field picture 0 otherwise
  411. int drop_frame_timecode; ///< timecode is in drop frame format.
  412. int scan_offset; ///< reserve space for SVCD scan offset user data.
  413. /* RTP specific */
  414. int rtp_mode;
  415. int rtp_payload_size;
  416. uint8_t *ptr_lastgob;
  417. int16_t (*pblocks[12])[64];
  418. int16_t (*block)[64]; ///< points to one of the following blocks
  419. int16_t (*blocks)[12][64]; // for HQ mode we need to keep the best block
  420. int (*decode_mb)(struct MpegEncContext *s, int16_t block[6][64]); // used by some codecs to avoid a switch()
  421. #define SLICE_OK 0
  422. #define SLICE_ERROR -1
  423. #define SLICE_END -2 ///<end marker found
  424. #define SLICE_NOEND -3 ///<no end marker or error found but mb count exceeded
  425. void (*dct_unquantize_mpeg1_intra)(struct MpegEncContext *s,
  426. int16_t *block/*align 16*/, int n, int qscale);
  427. void (*dct_unquantize_mpeg1_inter)(struct MpegEncContext *s,
  428. int16_t *block/*align 16*/, int n, int qscale);
  429. void (*dct_unquantize_mpeg2_intra)(struct MpegEncContext *s,
  430. int16_t *block/*align 16*/, int n, int qscale);
  431. void (*dct_unquantize_mpeg2_inter)(struct MpegEncContext *s,
  432. int16_t *block/*align 16*/, int n, int qscale);
  433. void (*dct_unquantize_h263_intra)(struct MpegEncContext *s,
  434. int16_t *block/*align 16*/, int n, int qscale);
  435. void (*dct_unquantize_h263_inter)(struct MpegEncContext *s,
  436. int16_t *block/*align 16*/, int n, int qscale);
  437. void (*dct_unquantize_intra)(struct MpegEncContext *s, // unquantizer to use (MPEG-4 can use both)
  438. int16_t *block/*align 16*/, int n, int qscale);
  439. void (*dct_unquantize_inter)(struct MpegEncContext *s, // unquantizer to use (MPEG-4 can use both)
  440. int16_t *block/*align 16*/, int n, int qscale);
  441. int (*dct_quantize)(struct MpegEncContext *s, int16_t *block/*align 16*/, int n, int qscale, int *overflow);
  442. int (*fast_dct_quantize)(struct MpegEncContext *s, int16_t *block/*align 16*/, int n, int qscale, int *overflow);
  443. void (*denoise_dct)(struct MpegEncContext *s, int16_t *block);
  444. int mpv_flags; ///< flags set by private options
  445. int quantizer_noise_shaping;
  446. /**
  447. * ratecontrol qmin qmax limiting method
  448. * 0-> clipping, 1-> use a nice continuous function to limit qscale within qmin/qmax.
  449. */
  450. float rc_qsquish;
  451. float rc_qmod_amp;
  452. int rc_qmod_freq;
  453. float rc_initial_cplx;
  454. float rc_buffer_aggressivity;
  455. float border_masking;
  456. int lmin, lmax;
  457. char *rc_eq;
  458. /* temp buffers for rate control */
  459. float *cplx_tab, *bits_tab;
  460. /* flag to indicate a reinitialization is required, e.g. after
  461. * a frame size change */
  462. int context_reinit;
  463. ERContext er;
  464. int error_rate;
  465. /* temporary frames used by b_frame_strategy = 2 */
  466. AVFrame *tmp_frames[MAX_B_FRAMES + 2];
  467. int b_frame_strategy;
  468. int b_sensitivity;
  469. /* frame skip options for encoding */
  470. int frame_skip_threshold;
  471. int frame_skip_factor;
  472. int frame_skip_exp;
  473. int frame_skip_cmp;
  474. int scenechange_threshold;
  475. int noise_reduction;
  476. } MpegEncContext;
  477. /* mpegvideo_enc common options */
  478. #define FF_MPV_FLAG_SKIP_RD 0x0001
  479. #define FF_MPV_FLAG_STRICT_GOP 0x0002
  480. #define FF_MPV_FLAG_QP_RD 0x0004
  481. #define FF_MPV_FLAG_CBP_RD 0x0008
  482. #define FF_MPV_FLAG_NAQ 0x0010
  483. #define FF_MPV_FLAG_MV0 0x0020
  484. #define FF_MPV_OPT_CMP_FUNC \
  485. { "sad", "Sum of absolute differences, fast", 0, AV_OPT_TYPE_CONST, {.i64 = FF_CMP_SAD }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS, "cmp_func" }, \
  486. { "sse", "Sum of squared errors", 0, AV_OPT_TYPE_CONST, {.i64 = FF_CMP_SSE }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS, "cmp_func" }, \
  487. { "satd", "Sum of absolute Hadamard transformed differences", 0, AV_OPT_TYPE_CONST, {.i64 = FF_CMP_SATD }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS, "cmp_func" }, \
  488. { "dct", "Sum of absolute DCT transformed differences", 0, AV_OPT_TYPE_CONST, {.i64 = FF_CMP_DCT }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS, "cmp_func" }, \
  489. { "psnr", "Sum of squared quantization errors, low quality", 0, AV_OPT_TYPE_CONST, {.i64 = FF_CMP_PSNR }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS, "cmp_func" }, \
  490. { "bit", "Number of bits needed for the block", 0, AV_OPT_TYPE_CONST, {.i64 = FF_CMP_BIT }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS, "cmp_func" }, \
  491. { "rd", "Rate distortion optimal, slow", 0, AV_OPT_TYPE_CONST, {.i64 = FF_CMP_RD }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS, "cmp_func" }, \
  492. { "zero", "Zero", 0, AV_OPT_TYPE_CONST, {.i64 = FF_CMP_ZERO }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS, "cmp_func" }, \
  493. { "vsad", "Sum of absolute vertical differences", 0, AV_OPT_TYPE_CONST, {.i64 = FF_CMP_VSAD }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS, "cmp_func" }, \
  494. { "vsse", "Sum of squared vertical differences", 0, AV_OPT_TYPE_CONST, {.i64 = FF_CMP_VSSE }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS, "cmp_func" }, \
  495. { "nsse", "Noise preserving sum of squared differences", 0, AV_OPT_TYPE_CONST, {.i64 = FF_CMP_NSSE }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS, "cmp_func" }, \
  496. { "dct264", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = FF_CMP_DCT264 }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS, "cmp_func" }, \
  497. { "dctmax", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = FF_CMP_DCTMAX }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS, "cmp_func" }, \
  498. { "chroma", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = FF_CMP_CHROMA }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS, "cmp_func" }
  499. #define FF_MPV_OFFSET(x) offsetof(MpegEncContext, x)
  500. #define FF_MPV_OPT_FLAGS (AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM)
  501. #define FF_MPV_COMMON_OPTS \
  502. FF_MPV_OPT_CMP_FUNC, \
  503. { "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" },\
  504. { "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" },\
  505. { "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" },\
  506. { "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" },\
  507. { "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" },\
  508. { "naq", "normalize adaptive quantization", 0, AV_OPT_TYPE_CONST, { .i64 = FF_MPV_FLAG_NAQ }, 0, 0, FF_MPV_OPT_FLAGS, "mpv_flags" },\
  509. { "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" },\
  510. { "luma_elim_threshold", "single coefficient elimination threshold for luminance (negative values also consider dc coefficient)",\
  511. FF_MPV_OFFSET(luma_elim_threshold), AV_OPT_TYPE_INT, { .i64 = 0 }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS },\
  512. { "chroma_elim_threshold", "single coefficient elimination threshold for chrominance (negative values also consider dc coefficient)",\
  513. FF_MPV_OFFSET(chroma_elim_threshold), AV_OPT_TYPE_INT, { .i64 = 0 }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS },\
  514. { "quantizer_noise_shaping", NULL, FF_MPV_OFFSET(quantizer_noise_shaping), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, INT_MAX, FF_MPV_OPT_FLAGS },\
  515. { "error_rate", "Simulate errors in the bitstream to test error concealment.", \
  516. FF_MPV_OFFSET(error_rate), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, INT_MAX, FF_MPV_OPT_FLAGS },\
  517. {"qsquish", "how to keep quantizer between qmin and qmax (0 = clip, 1 = use differentiable function)", \
  518. FF_MPV_OFFSET(rc_qsquish), AV_OPT_TYPE_FLOAT, {.dbl = 0 }, 0, 99, FF_MPV_OPT_FLAGS}, \
  519. {"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}, \
  520. {"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}, \
  521. {"rc_eq", "Set rate control equation. When computing the expression, besides the standard functions " \
  522. "defined in the section 'Expression Evaluation', the following functions are available: " \
  523. "bits2qp(bits), qp2bits(qp). Also the following constants are available: iTex pTex tex mv " \
  524. "fCode iCount mcVar var isI isP isB avgQP qComp avgIITex avgPITex avgPPTex avgBPTex avgTex.", \
  525. FF_MPV_OFFSET(rc_eq), AV_OPT_TYPE_STRING, .flags = FF_MPV_OPT_FLAGS }, \
  526. {"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}, \
  527. {"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}, \
  528. {"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}, \
  529. {"lmin", "minimum Lagrange factor (VBR)", FF_MPV_OFFSET(lmin), AV_OPT_TYPE_INT, {.i64 = 2*FF_QP2LAMBDA }, 0, INT_MAX, FF_MPV_OPT_FLAGS }, \
  530. {"lmax", "maximum Lagrange factor (VBR)", FF_MPV_OFFSET(lmax), AV_OPT_TYPE_INT, {.i64 = 31*FF_QP2LAMBDA }, 0, INT_MAX, FF_MPV_OPT_FLAGS }, \
  531. {"ibias", "intra quant bias", FF_MPV_OFFSET(intra_quant_bias), AV_OPT_TYPE_INT, {.i64 = FF_DEFAULT_QUANT_BIAS }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS }, \
  532. {"pbias", "inter quant bias", FF_MPV_OFFSET(inter_quant_bias), AV_OPT_TYPE_INT, {.i64 = FF_DEFAULT_QUANT_BIAS }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS }, \
  533. {"rc_strategy", "ratecontrol method", FF_MPV_OFFSET(rc_strategy), AV_OPT_TYPE_INT, {.i64 = 0 }, 0, 1, FF_MPV_OPT_FLAGS }, \
  534. {"motion_est", "motion estimation algorithm", FF_MPV_OFFSET(motion_est), AV_OPT_TYPE_INT, {.i64 = FF_ME_EPZS }, FF_ME_ZERO, FF_ME_XONE, FF_MPV_OPT_FLAGS, "motion_est" }, \
  535. { "zero", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FF_ME_ZERO }, 0, 0, FF_MPV_OPT_FLAGS, "motion_est" }, \
  536. { "epzs", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FF_ME_EPZS }, 0, 0, FF_MPV_OPT_FLAGS, "motion_est" }, \
  537. { "xone", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FF_ME_XONE }, 0, 0, FF_MPV_OPT_FLAGS, "motion_est" }, \
  538. {"b_strategy", "Strategy to choose between I/P/B-frames", FF_MPV_OFFSET(b_frame_strategy), AV_OPT_TYPE_INT, {.i64 = 0 }, 0, 2, FF_MPV_OPT_FLAGS }, \
  539. {"b_sensitivity", "Adjust sensitivity of b_frame_strategy 1", FF_MPV_OFFSET(b_sensitivity), AV_OPT_TYPE_INT, {.i64 = 40 }, 1, INT_MAX, FF_MPV_OPT_FLAGS }, \
  540. {"brd_scale", "Downscale frames for dynamic B-frame decision", FF_MPV_OFFSET(brd_scale), AV_OPT_TYPE_INT, {.i64 = 0 }, 0, 3, FF_MPV_OPT_FLAGS }, \
  541. {"skip_threshold", "Frame skip threshold", FF_MPV_OFFSET(frame_skip_threshold), AV_OPT_TYPE_INT, {.i64 = 0 }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS }, \
  542. {"skip_factor", "Frame skip factor", FF_MPV_OFFSET(frame_skip_factor), AV_OPT_TYPE_INT, {.i64 = 0 }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS }, \
  543. {"skip_exp", "Frame skip exponent", FF_MPV_OFFSET(frame_skip_exp), AV_OPT_TYPE_INT, {.i64 = 0 }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS }, \
  544. {"skip_cmp", "Frame skip compare function", FF_MPV_OFFSET(frame_skip_cmp), AV_OPT_TYPE_INT, {.i64 = FF_CMP_DCTMAX }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS, "cmp_func" }, \
  545. {"sc_threshold", "Scene change threshold", FF_MPV_OFFSET(scenechange_threshold), AV_OPT_TYPE_INT, {.i64 = 0 }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS }, \
  546. {"noise_reduction", "Noise reduction", FF_MPV_OFFSET(noise_reduction), AV_OPT_TYPE_INT, {.i64 = 0 }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS }, \
  547. {"mpeg_quant", "Use MPEG quantizers instead of H.263", FF_MPV_OFFSET(mpeg_quant), AV_OPT_TYPE_INT, {.i64 = 0 }, 0, 1, FF_MPV_OPT_FLAGS }, \
  548. {"ps", "RTP payload size in bytes", FF_MPV_OFFSET(rtp_payload_size), AV_OPT_TYPE_INT, {.i64 = 0 }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS }, \
  549. {"mepc", "Motion estimation bitrate penalty compensation (1.0 = 256)", FF_MPV_OFFSET(me_penalty_compensation), AV_OPT_TYPE_INT, {.i64 = 256 }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS }, \
  550. {"mepre", "pre motion estimation", FF_MPV_OFFSET(me_pre), AV_OPT_TYPE_INT, {.i64 = 0 }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS }, \
  551. extern const AVOption ff_mpv_generic_options[];
  552. /**
  553. * Set the given MpegEncContext to common defaults (same for encoding
  554. * and decoding). The changed fields will not depend upon the prior
  555. * state of the MpegEncContext.
  556. */
  557. void ff_mpv_common_defaults(MpegEncContext *s);
  558. int ff_mpv_common_init(MpegEncContext *s);
  559. void ff_mpv_common_init_arm(MpegEncContext *s);
  560. void ff_mpv_common_init_neon(MpegEncContext *s);
  561. void ff_mpv_common_init_ppc(MpegEncContext *s);
  562. void ff_mpv_common_init_x86(MpegEncContext *s);
  563. int ff_mpv_common_frame_size_change(MpegEncContext *s);
  564. void ff_mpv_common_end(MpegEncContext *s);
  565. void ff_mpv_decode_defaults(MpegEncContext *s);
  566. void ff_mpv_decode_mb(MpegEncContext *s, int16_t block[12][64]);
  567. void ff_mpv_report_decode_progress(MpegEncContext *s);
  568. int ff_mpv_frame_start(MpegEncContext *s, AVCodecContext *avctx);
  569. void ff_mpv_frame_end(MpegEncContext *s);
  570. int ff_mpv_encode_init(AVCodecContext *avctx);
  571. void ff_mpv_encode_init_x86(MpegEncContext *s);
  572. int ff_mpv_encode_end(AVCodecContext *avctx);
  573. int ff_mpv_encode_picture(AVCodecContext *avctx, AVPacket *pkt,
  574. const AVFrame *frame, int *got_packet);
  575. void ff_clean_intra_table_entries(MpegEncContext *s);
  576. void ff_mpeg_draw_horiz_band(MpegEncContext *s, int y, int h);
  577. void ff_mpeg_flush(AVCodecContext *avctx);
  578. void ff_print_debug_info(MpegEncContext *s, Picture *p);
  579. void ff_write_quant_matrix(PutBitContext *pb, uint16_t *matrix);
  580. int ff_update_duplicate_context(MpegEncContext *dst, MpegEncContext *src);
  581. int ff_mpeg_update_thread_context(AVCodecContext *dst, const AVCodecContext *src);
  582. void ff_set_qscale(MpegEncContext * s, int qscale);
  583. void ff_mpv_idct_init(MpegEncContext *s);
  584. void ff_convert_matrix(MpegEncContext *s, int (*qmat)[64], uint16_t (*qmat16)[2][64],
  585. const uint16_t *quant_matrix, int bias, int qmin, int qmax, int intra);
  586. int ff_dct_quantize_c(MpegEncContext *s, int16_t *block, int n, int qscale, int *overflow);
  587. void ff_init_block_index(MpegEncContext *s);
  588. void ff_mpv_motion(MpegEncContext *s,
  589. uint8_t *dest_y, uint8_t *dest_cb,
  590. uint8_t *dest_cr, int dir,
  591. uint8_t **ref_picture,
  592. op_pixels_func (*pix_op)[4],
  593. qpel_mc_func (*qpix_op)[16]);
  594. static inline void ff_update_block_index(MpegEncContext *s){
  595. const int block_size = 8;
  596. s->block_index[0]+=2;
  597. s->block_index[1]+=2;
  598. s->block_index[2]+=2;
  599. s->block_index[3]+=2;
  600. s->block_index[4]++;
  601. s->block_index[5]++;
  602. s->dest[0]+= 2*block_size;
  603. s->dest[1]+= block_size;
  604. s->dest[2]+= block_size;
  605. }
  606. static inline int get_bits_diff(MpegEncContext *s){
  607. const int bits= put_bits_count(&s->pb);
  608. const int last= s->last_bits;
  609. s->last_bits = bits;
  610. return bits - last;
  611. }
  612. #endif /* AVCODEC_MPEGVIDEO_H */