e7914fd0548ab1f1d85bac7ce9a9c535596a6952
[libav.git] / libavcodec / mpegvideo.h
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 /**
24 * @file
25 * mpegvideo header.
26 */
27
28 #ifndef AVCODEC_MPEGVIDEO_H
29 #define AVCODEC_MPEGVIDEO_H
30
31 #include <float.h>
32
33 #include "avcodec.h"
34 #include "blockdsp.h"
35 #include "error_resilience.h"
36 #include "fdctdsp.h"
37 #include "get_bits.h"
38 #include "h263dsp.h"
39 #include "hpeldsp.h"
40 #include "idctdsp.h"
41 #include "internal.h"
42 #include "me_cmp.h"
43 #include "motion_est.h"
44 #include "mpegpicture.h"
45 #include "mpegvideodsp.h"
46 #include "mpegvideoencdsp.h"
47 #include "pixblockdsp.h"
48 #include "put_bits.h"
49 #include "ratecontrol.h"
50 #include "parser.h"
51 #include "mpegutils.h"
52 #include "mpeg12data.h"
53 #include "qpeldsp.h"
54 #include "thread.h"
55 #include "videodsp.h"
56
57 #include "libavutil/opt.h"
58
59 #define MAX_THREADS 16
60
61 #define MAX_B_FRAMES 16
62
63 /* Start codes. */
64 #define SEQ_END_CODE 0x000001b7
65 #define SEQ_START_CODE 0x000001b3
66 #define GOP_START_CODE 0x000001b8
67 #define PICTURE_START_CODE 0x00000100
68 #define SLICE_MIN_START_CODE 0x00000101
69 #define SLICE_MAX_START_CODE 0x000001af
70 #define EXT_START_CODE 0x000001b5
71 #define USER_START_CODE 0x000001b2
72
73 /**
74 * MpegEncContext.
75 */
76 typedef struct MpegEncContext {
77 AVClass *class;
78
79 int y_dc_scale, c_dc_scale;
80 int ac_pred;
81 int block_last_index[12]; ///< last non zero coefficient in block
82 int h263_aic; ///< Advanded INTRA Coding (AIC)
83
84 /* scantables */
85 ScanTable inter_scantable; ///< if inter == intra then intra should be used to reduce tha cache usage
86 ScanTable intra_scantable;
87 ScanTable intra_h_scantable;
88 ScanTable intra_v_scantable;
89
90 /* WARNING: changes above this line require updates to hardcoded
91 * offsets used in asm. */
92
93 struct AVCodecContext *avctx;
94 /* the following parameters must be initialized before encoding */
95 int width, height;///< picture size. must be a multiple of 16
96 int gop_size;
97 int intra_only; ///< if true, only intra pictures are generated
98 int bit_rate; ///< wanted bit rate
99 enum OutputFormat out_format; ///< output format
100 int h263_pred; ///< use mpeg4/h263 ac/dc predictions
101 int pb_frame; ///< PB frame mode (0 = none, 1 = base, 2 = improved)
102
103 /* the following codec id fields are deprecated in favor of codec_id */
104 int h263_plus; ///< h263 plus headers
105 int h263_flv; ///< use flv h263 header
106
107 enum AVCodecID codec_id; /* see AV_CODEC_ID_xxx */
108 int fixed_qscale; ///< fixed qscale if non zero
109 int encoding; ///< true if we are encoding (vs decoding)
110 int max_b_frames; ///< max number of b-frames for encoding
111 int luma_elim_threshold;
112 int chroma_elim_threshold;
113 int strict_std_compliance; ///< strictly follow the std (MPEG4, ...)
114 int workaround_bugs; ///< workaround bugs in encoders which cannot be detected automatically
115 int codec_tag; ///< internal codec_tag upper case converted from avctx codec_tag
116 /* the following fields are managed internally by the encoder */
117
118 /* sequence parameters */
119 int context_initialized;
120 int input_picture_number; ///< used to set pic->display_picture_number, should not be used for/by anything else
121 int coded_picture_number; ///< used to set pic->coded_picture_number, should not be used for/by anything else
122 int picture_number; //FIXME remove, unclear definition
123 int picture_in_gop_number; ///< 0-> first pic in gop, ...
124 int mb_width, mb_height; ///< number of MBs horizontally & vertically
125 int mb_stride; ///< mb_width+1 used for some arrays to allow simple addressing of left & top MBs without sig11
126 int b8_stride; ///< 2*mb_width+1 used for some 8x8 block arrays to allow simple addressing
127 int h_edge_pos, v_edge_pos;///< horizontal / vertical position of the right/bottom edge (pixel replication)
128 int mb_num; ///< number of MBs of a picture
129 ptrdiff_t linesize; ///< line size, in bytes, may be different from width
130 ptrdiff_t uvlinesize; ///< line size, for chroma in bytes, may be different from width
131 Picture *picture; ///< main picture buffer
132 Picture **input_picture; ///< next pictures on display order for encoding
133 Picture **reordered_input_picture; ///< pointer to the next pictures in codedorder for encoding
134
135 int64_t user_specified_pts; ///< last non-zero pts from AVFrame which was passed into avcodec_encode_video2()
136 /**
137 * pts difference between the first and second input frame, used for
138 * calculating dts of the first frame when there's a delay */
139 int64_t dts_delta;
140 /**
141 * reordered pts to be used as dts for the next output frame when there's
142 * a delay */
143 int64_t reordered_pts;
144
145 /** bit output */
146 PutBitContext pb;
147
148 int start_mb_y; ///< start mb_y of this thread (so current thread should process start_mb_y <= row < end_mb_y)
149 int end_mb_y; ///< end mb_y of this thread (so current thread should process start_mb_y <= row < end_mb_y)
150 struct MpegEncContext *thread_context[MAX_THREADS];
151 int slice_context_count; ///< number of used thread_contexts
152
153 /**
154 * copy of the previous picture structure.
155 * note, linesize & data, might not match the previous picture (for field pictures)
156 */
157 Picture last_picture;
158
159 /**
160 * copy of the next picture structure.
161 * note, linesize & data, might not match the next picture (for field pictures)
162 */
163 Picture next_picture;
164
165 /**
166 * copy of the source picture structure for encoding.
167 * note, linesize & data, might not match the source picture (for field pictures)
168 */
169 Picture new_picture;
170
171 /**
172 * copy of the current picture structure.
173 * note, linesize & data, might not match the current picture (for field pictures)
174 */
175 Picture current_picture; ///< buffer to store the decompressed current picture
176
177 Picture *last_picture_ptr; ///< pointer to the previous picture.
178 Picture *next_picture_ptr; ///< pointer to the next picture (for bidir pred)
179 Picture *current_picture_ptr; ///< pointer to the current picture
180 int last_dc[3]; ///< last DC values for MPEG1
181 int16_t *dc_val_base;
182 int16_t *dc_val[3]; ///< used for mpeg4 DC prediction, all 3 arrays must be continuous
183 const uint8_t *y_dc_scale_table; ///< qscale -> y_dc_scale table
184 const uint8_t *c_dc_scale_table; ///< qscale -> c_dc_scale table
185 const uint8_t *chroma_qscale_table; ///< qscale -> chroma_qscale (h263)
186 uint8_t *coded_block_base;
187 uint8_t *coded_block; ///< used for coded block pattern prediction (msmpeg4v3, wmv1)
188 int16_t (*ac_val_base)[16];
189 int16_t (*ac_val[3])[16]; ///< used for for mpeg4 AC prediction, all 3 arrays must be continuous
190 int mb_skipped; ///< MUST BE SET only during DECODING
191 uint8_t *mbskip_table; /**< used to avoid copy if macroblock skipped (for black regions for example)
192 and used for b-frame encoding & decoding (contains skip table of next P Frame) */
193 uint8_t *mbintra_table; ///< used to avoid setting {ac, dc, cbp}-pred stuff to zero on inter MB decoding
194 uint8_t *cbp_table; ///< used to store cbp, ac_pred for partitioned decoding
195 uint8_t *pred_dir_table; ///< used to store pred_dir for partitioned decoding
196
197 ScratchpadContext sc;
198
199 int qscale; ///< QP
200 int chroma_qscale; ///< chroma QP
201 unsigned int lambda; ///< lagrange multipler used in rate distortion
202 unsigned int lambda2; ///< (lambda*lambda) >> FF_LAMBDA_SHIFT
203 int *lambda_table;
204 int adaptive_quant; ///< use adaptive quantization
205 int dquant; ///< qscale difference to prev qscale
206 int pict_type; ///< AV_PICTURE_TYPE_I, AV_PICTURE_TYPE_P, AV_PICTURE_TYPE_B, ...
207 int last_pict_type; //FIXME removes
208 int last_non_b_pict_type; ///< used for mpeg4 gmc b-frames & ratecontrol
209 int droppable;
210 int frame_rate_index;
211 int last_lambda_for[5]; ///< last lambda for a specific pict type
212 int skipdct; ///< skip dct and code zero residual
213
214 /* motion compensation */
215 int unrestricted_mv; ///< mv can point outside of the coded picture
216 int h263_long_vectors; ///< use horrible h263v1 long vector mode
217
218 BlockDSPContext bdsp;
219 FDCTDSPContext fdsp;
220 HpelDSPContext hdsp;
221 IDCTDSPContext idsp;
222 MECmpContext mecc;
223 MpegVideoDSPContext mdsp;
224 MpegvideoEncDSPContext mpvencdsp;
225 PixblockDSPContext pdsp;
226 QpelDSPContext qdsp;
227 VideoDSPContext vdsp;
228 H263DSPContext h263dsp;
229 int f_code; ///< forward MV resolution
230 int b_code; ///< backward MV resolution for B Frames (mpeg4)
231 int16_t (*p_mv_table_base)[2];
232 int16_t (*b_forw_mv_table_base)[2];
233 int16_t (*b_back_mv_table_base)[2];
234 int16_t (*b_bidir_forw_mv_table_base)[2];
235 int16_t (*b_bidir_back_mv_table_base)[2];
236 int16_t (*b_direct_mv_table_base)[2];
237 int16_t (*p_field_mv_table_base[2][2])[2];
238 int16_t (*b_field_mv_table_base[2][2][2])[2];
239 int16_t (*p_mv_table)[2]; ///< MV table (1MV per MB) p-frame encoding
240 int16_t (*b_forw_mv_table)[2]; ///< MV table (1MV per MB) forward mode b-frame encoding
241 int16_t (*b_back_mv_table)[2]; ///< MV table (1MV per MB) backward mode b-frame encoding
242 int16_t (*b_bidir_forw_mv_table)[2]; ///< MV table (1MV per MB) bidir mode b-frame encoding
243 int16_t (*b_bidir_back_mv_table)[2]; ///< MV table (1MV per MB) bidir mode b-frame encoding
244 int16_t (*b_direct_mv_table)[2]; ///< MV table (1MV per MB) direct mode b-frame encoding
245 int16_t (*p_field_mv_table[2][2])[2]; ///< MV table (2MV per MB) interlaced p-frame encoding
246 int16_t (*b_field_mv_table[2][2][2])[2];///< MV table (4MV per MB) interlaced b-frame encoding
247 uint8_t (*p_field_select_table[2]);
248 uint8_t (*b_field_select_table[2][2]);
249 #if FF_API_MOTION_EST
250 int me_method; ///< ME algorithm
251 #endif
252 int motion_est; ///< ME algorithm
253 int mv_dir;
254 #define MV_DIR_FORWARD 1
255 #define MV_DIR_BACKWARD 2
256 #define MV_DIRECT 4 ///< bidirectional mode where the difference equals the MV of the last P/S/I-Frame (mpeg4)
257 int mv_type;
258 #define MV_TYPE_16X16 0 ///< 1 vector for the whole mb
259 #define MV_TYPE_8X8 1 ///< 4 vectors (h263, mpeg4 4MV)
260 #define MV_TYPE_16X8 2 ///< 2 vectors, one per 16x8 block
261 #define MV_TYPE_FIELD 3 ///< 2 vectors, one per field
262 #define MV_TYPE_DMV 4 ///< 2 vectors, special mpeg2 Dual Prime Vectors
263 /**motion vectors for a macroblock
264 first coordinate : 0 = forward 1 = backward
265 second " : depend on type
266 third " : 0 = x, 1 = y
267 */
268 int mv[2][4][2];
269 int field_select[2][2];
270 int last_mv[2][2][2]; ///< last MV, used for MV prediction in MPEG1 & B-frame MPEG4
271 uint8_t *fcode_tab; ///< smallest fcode needed for each MV
272 int16_t direct_scale_mv[2][64]; ///< precomputed to avoid divisions in ff_mpeg4_set_direct_mv
273
274 MotionEstContext me;
275
276 int no_rounding; /**< apply no rounding to motion compensation (MPEG4, msmpeg4, ...)
277 for b-frames rounding mode is always 0 */
278
279 /* macroblock layer */
280 int mb_x, mb_y;
281 int mb_skip_run;
282 int mb_intra;
283 uint16_t *mb_type; ///< Table for candidate MB types for encoding (defines in mpegutils.h)
284
285 int block_index[6]; ///< index to current MB in block based arrays with edges
286 int block_wrap[6];
287 uint8_t *dest[3];
288
289 int *mb_index2xy; ///< mb_index -> mb_x + mb_y*mb_stride
290
291 /** matrix transmitted in the bitstream */
292 uint16_t intra_matrix[64];
293 uint16_t chroma_intra_matrix[64];
294 uint16_t inter_matrix[64];
295 uint16_t chroma_inter_matrix[64];
296
297 int intra_quant_bias; ///< bias for the quantizer
298 int inter_quant_bias; ///< bias for the quantizer
299 int min_qcoeff; ///< minimum encodable coefficient
300 int max_qcoeff; ///< maximum encodable coefficient
301 int ac_esc_length; ///< num of bits needed to encode the longest esc
302 uint8_t *intra_ac_vlc_length;
303 uint8_t *intra_ac_vlc_last_length;
304 uint8_t *inter_ac_vlc_length;
305 uint8_t *inter_ac_vlc_last_length;
306 uint8_t *luma_dc_vlc_length;
307 #define UNI_AC_ENC_INDEX(run,level) ((run)*128 + (level))
308
309 int coded_score[8];
310
311 /** precomputed matrix (combine qscale and DCT renorm) */
312 int (*q_intra_matrix)[64];
313 int (*q_inter_matrix)[64];
314 /** identical to the above but for MMX & these are not permutated, second 64 entries are bias*/
315 uint16_t (*q_intra_matrix16)[2][64];
316 uint16_t (*q_inter_matrix16)[2][64];
317
318 /* noise reduction */
319 int (*dct_error_sum)[64];
320 int dct_count[2];
321 uint16_t (*dct_offset)[64];
322
323 /* bit rate control */
324 int64_t total_bits;
325 int frame_bits; ///< bits used for the current frame
326 int next_lambda; ///< next lambda used for retrying to encode a frame
327 RateControlContext rc_context; ///< contains stuff only accessed in ratecontrol.c
328
329 /* statistics, used for 2-pass encoding */
330 int mv_bits;
331 int header_bits;
332 int i_tex_bits;
333 int p_tex_bits;
334 int i_count;
335 int f_count;
336 int b_count;
337 int skip_count;
338 int misc_bits; ///< cbp, mb_type
339 int last_bits; ///< temp var used for calculating the above vars
340
341 /* error concealment / resync */
342 int resync_mb_x; ///< x position of last resync marker
343 int resync_mb_y; ///< y position of last resync marker
344 GetBitContext last_resync_gb; ///< used to search for the next resync marker
345 int mb_num_left; ///< number of MBs left in this video packet (for partitioned Slices only)
346 int next_p_frame_damaged; ///< set if the next p frame is damaged, to avoid showing trashed b frames
347
348 ParseContext parse_context;
349
350 /* H.263 specific */
351 int gob_index;
352 int obmc; ///< overlapped block motion compensation
353 int mb_info; ///< interval for outputting info about mb offsets as side data
354 int prev_mb_info, last_mb_info;
355 uint8_t *mb_info_ptr;
356 int mb_info_size;
357 int rc_strategy;
358
359 /* H.263+ specific */
360 int umvplus; ///< == H263+ && unrestricted_mv
361 int h263_aic_dir; ///< AIC direction: 0 = left, 1 = top
362 int h263_slice_structured;
363 int alt_inter_vlc; ///< alternative inter vlc
364 int modified_quant;
365 int loop_filter;
366 int custom_pcf;
367
368 /* mpeg4 specific */
369 ///< number of bits to represent the fractional part of time (encoder only)
370 int time_increment_bits;
371 int last_time_base;
372 int time_base; ///< time in seconds of last I,P,S Frame
373 int64_t time; ///< time of current frame
374 int64_t last_non_b_time;
375 uint16_t pp_time; ///< time distance between the last 2 p,s,i frames
376 uint16_t pb_time; ///< time distance between the last b and p,s,i frame
377 uint16_t pp_field_time;
378 uint16_t pb_field_time; ///< like above, just for interlaced
379 int real_sprite_warping_points;
380 int sprite_offset[2][2]; ///< sprite offset[isChroma][isMVY]
381 int sprite_delta[2][2]; ///< sprite_delta [isY][isMVY]
382 int mcsel;
383 int quant_precision;
384 int quarter_sample; ///< 1->qpel, 0->half pel ME/MC
385 int aspect_ratio_info; //FIXME remove
386 int sprite_warping_accuracy;
387 int data_partitioning; ///< data partitioning flag from header
388 int partitioned_frame; ///< is current frame partitioned
389 int low_delay; ///< no reordering needed / has no b-frames
390 int vo_type;
391 PutBitContext tex_pb; ///< used for data partitioned VOPs
392 PutBitContext pb2; ///< used for data partitioned VOPs
393 int mpeg_quant;
394 int padding_bug_score; ///< used to detect the VERY common padding bug in MPEG4
395
396 /* divx specific, used to workaround (many) bugs in divx5 */
397 int divx_packed;
398 uint8_t *bitstream_buffer; //Divx 5.01 puts several frames in a single one, this is used to reorder them
399 int bitstream_buffer_size;
400 unsigned int allocated_bitstream_buffer_size;
401
402 /* RV10 specific */
403 int rv10_version; ///< RV10 version: 0 or 3
404 int rv10_first_dc_coded[3];
405
406 /* MJPEG specific */
407 struct MJpegContext *mjpeg_ctx;
408 int pred;
409
410 /* MSMPEG4 specific */
411 int mv_table_index;
412 int rl_table_index;
413 int rl_chroma_table_index;
414 int dc_table_index;
415 int use_skip_mb_code;
416 int slice_height; ///< in macroblocks
417 int first_slice_line; ///< used in mpeg4 too to handle resync markers
418 int flipflop_rounding;
419 int msmpeg4_version; ///< 0=not msmpeg4, 1=mp41, 2=mp42, 3=mp43/divx3 4=wmv1/7 5=wmv2/8
420 int per_mb_rl_table;
421 int esc3_level_length;
422 int esc3_run_length;
423 /** [mb_intra][isChroma][level][run][last] */
424 int (*ac_stats)[2][MAX_LEVEL+1][MAX_RUN+1][2];
425 int inter_intra_pred;
426 int mspel;
427
428 /* decompression specific */
429 GetBitContext gb;
430
431 /* Mpeg1 specific */
432 int gop_picture_number; ///< index of the first picture of a GOP based on fake_pic_num & mpeg1 specific
433 int last_mv_dir; ///< last mv_dir, used for b frame encoding
434 uint8_t *vbv_delay_ptr; ///< pointer to vbv_delay in the bitstream
435
436 /* MPEG-2-specific - I wished not to have to support this mess. */
437 int progressive_sequence;
438 int mpeg_f_code[2][2];
439
440 // picture structure defines are loaded from mpegutils.h
441 int picture_structure;
442
443 int64_t timecode_frame_start; ///< GOP timecode frame start number, in non drop frame format
444 int intra_dc_precision;
445 int frame_pred_frame_dct;
446 int top_field_first;
447 int concealment_motion_vectors;
448 int q_scale_type;
449 int brd_scale;
450 int intra_vlc_format;
451 int alternate_scan;
452 int repeat_first_field;
453 int chroma_420_type;
454 int chroma_format;
455 #define CHROMA_420 1
456 #define CHROMA_422 2
457 int chroma_x_shift;//depend on pix_format, that depend on chroma_format
458 int chroma_y_shift;
459
460 int progressive_frame;
461 int full_pel[2];
462 int interlaced_dct;
463 int first_field; ///< is 1 for the first field of a field picture 0 otherwise
464 int drop_frame_timecode; ///< timecode is in drop frame format.
465 int scan_offset; ///< reserve space for SVCD scan offset user data.
466
467 /* RTP specific */
468 int rtp_mode;
469 int rtp_payload_size;
470
471 uint8_t *ptr_lastgob;
472 int16_t (*pblocks[12])[64];
473
474 int16_t (*block)[64]; ///< points to one of the following blocks
475 int16_t (*blocks)[12][64]; // for HQ mode we need to keep the best block
476 int (*decode_mb)(struct MpegEncContext *s, int16_t block[6][64]); // used by some codecs to avoid a switch()
477 #define SLICE_OK 0
478 #define SLICE_ERROR -1
479 #define SLICE_END -2 ///<end marker found
480 #define SLICE_NOEND -3 ///<no end marker or error found but mb count exceeded
481
482 void (*dct_unquantize_mpeg1_intra)(struct MpegEncContext *s,
483 int16_t *block/*align 16*/, int n, int qscale);
484 void (*dct_unquantize_mpeg1_inter)(struct MpegEncContext *s,
485 int16_t *block/*align 16*/, int n, int qscale);
486 void (*dct_unquantize_mpeg2_intra)(struct MpegEncContext *s,
487 int16_t *block/*align 16*/, int n, int qscale);
488 void (*dct_unquantize_mpeg2_inter)(struct MpegEncContext *s,
489 int16_t *block/*align 16*/, int n, int qscale);
490 void (*dct_unquantize_h263_intra)(struct MpegEncContext *s,
491 int16_t *block/*align 16*/, int n, int qscale);
492 void (*dct_unquantize_h263_inter)(struct MpegEncContext *s,
493 int16_t *block/*align 16*/, int n, int qscale);
494 void (*dct_unquantize_intra)(struct MpegEncContext *s, // unquantizer to use (mpeg4 can use both)
495 int16_t *block/*align 16*/, int n, int qscale);
496 void (*dct_unquantize_inter)(struct MpegEncContext *s, // unquantizer to use (mpeg4 can use both)
497 int16_t *block/*align 16*/, int n, int qscale);
498 int (*dct_quantize)(struct MpegEncContext *s, int16_t *block/*align 16*/, int n, int qscale, int *overflow);
499 int (*fast_dct_quantize)(struct MpegEncContext *s, int16_t *block/*align 16*/, int n, int qscale, int *overflow);
500 void (*denoise_dct)(struct MpegEncContext *s, int16_t *block);
501
502 int mpv_flags; ///< flags set by private options
503 int quantizer_noise_shaping;
504
505 /**
506 * ratecontrol qmin qmax limiting method
507 * 0-> clipping, 1-> use a nice continuous function to limit qscale wthin qmin/qmax.
508 */
509 float rc_qsquish;
510 float rc_qmod_amp;
511 int rc_qmod_freq;
512 float rc_initial_cplx;
513 float rc_buffer_aggressivity;
514 float border_masking;
515 int lmin, lmax;
516
517 char *rc_eq;
518
519 /* temp buffers for rate control */
520 float *cplx_tab, *bits_tab;
521
522 /* flag to indicate a reinitialization is required, e.g. after
523 * a frame size change */
524 int context_reinit;
525
526 ERContext er;
527
528 int error_rate;
529
530 /* temporary frames used by b_frame_strategy = 2 */
531 AVFrame *tmp_frames[MAX_B_FRAMES + 2];
532 int b_frame_strategy;
533 int b_sensitivity;
534
535 /* frame skip options for encoding */
536 int frame_skip_threshold;
537 int frame_skip_factor;
538 int frame_skip_exp;
539 int frame_skip_cmp;
540
541 int scenechange_threshold;
542 int noise_reduction;
543 } MpegEncContext;
544
545 /* mpegvideo_enc common options */
546 #define FF_MPV_FLAG_SKIP_RD 0x0001
547 #define FF_MPV_FLAG_STRICT_GOP 0x0002
548 #define FF_MPV_FLAG_QP_RD 0x0004
549 #define FF_MPV_FLAG_CBP_RD 0x0008
550 #define FF_MPV_FLAG_NAQ 0x0010
551 #define FF_MPV_FLAG_MV0 0x0020
552
553 #define FF_MPV_OPT_CMP_FUNC \
554 { "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" }, \
555 { "sse", "Sum of squared errors", 0, AV_OPT_TYPE_CONST, {.i64 = FF_CMP_SSE }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS, "cmp_func" }, \
556 { "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" }, \
557 { "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" }, \
558 { "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" }, \
559 { "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" }, \
560 { "rd", "Rate distortion optimal, slow", 0, AV_OPT_TYPE_CONST, {.i64 = FF_CMP_RD }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS, "cmp_func" }, \
561 { "zero", "Zero", 0, AV_OPT_TYPE_CONST, {.i64 = FF_CMP_ZERO }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS, "cmp_func" }, \
562 { "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" }, \
563 { "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" }, \
564 { "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" }, \
565 { "dct264", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = FF_CMP_DCT264 }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS, "cmp_func" }, \
566 { "dctmax", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = FF_CMP_DCTMAX }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS, "cmp_func" }, \
567 { "chroma", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = FF_CMP_CHROMA }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS, "cmp_func" }
568
569 #define FF_MPV_OFFSET(x) offsetof(MpegEncContext, x)
570 #define FF_MPV_OPT_FLAGS (AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM)
571 #define FF_MPV_COMMON_OPTS \
572 FF_MPV_OPT_CMP_FUNC, \
573 { "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" },\
574 { "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" },\
575 { "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" },\
576 { "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" },\
577 { "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" },\
578 { "naq", "normalize adaptive quantization", 0, AV_OPT_TYPE_CONST, { .i64 = FF_MPV_FLAG_NAQ }, 0, 0, FF_MPV_OPT_FLAGS, "mpv_flags" },\
579 { "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" },\
580 { "luma_elim_threshold", "single coefficient elimination threshold for luminance (negative values also consider dc coefficient)",\
581 FF_MPV_OFFSET(luma_elim_threshold), AV_OPT_TYPE_INT, { .i64 = 0 }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS },\
582 { "chroma_elim_threshold", "single coefficient elimination threshold for chrominance (negative values also consider dc coefficient)",\
583 FF_MPV_OFFSET(chroma_elim_threshold), AV_OPT_TYPE_INT, { .i64 = 0 }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS },\
584 { "quantizer_noise_shaping", NULL, FF_MPV_OFFSET(quantizer_noise_shaping), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, INT_MAX, FF_MPV_OPT_FLAGS },\
585 { "error_rate", "Simulate errors in the bitstream to test error concealment.", \
586 FF_MPV_OFFSET(error_rate), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, INT_MAX, FF_MPV_OPT_FLAGS },\
587 {"qsquish", "how to keep quantizer between qmin and qmax (0 = clip, 1 = use differentiable function)", \
588 FF_MPV_OFFSET(rc_qsquish), AV_OPT_TYPE_FLOAT, {.dbl = 0 }, 0, 99, FF_MPV_OPT_FLAGS}, \
589 {"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}, \
590 {"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}, \
591 {"rc_eq", "Set rate control equation. When computing the expression, besides the standard functions " \
592 "defined in the section 'Expression Evaluation', the following functions are available: " \
593 "bits2qp(bits), qp2bits(qp). Also the following constants are available: iTex pTex tex mv " \
594 "fCode iCount mcVar var isI isP isB avgQP qComp avgIITex avgPITex avgPPTex avgBPTex avgTex.", \
595 FF_MPV_OFFSET(rc_eq), AV_OPT_TYPE_STRING, .flags = FF_MPV_OPT_FLAGS }, \
596 {"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}, \
597 {"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}, \
598 {"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}, \
599 {"lmin", "minimum Lagrange factor (VBR)", FF_MPV_OFFSET(lmin), AV_OPT_TYPE_INT, {.i64 = 2*FF_QP2LAMBDA }, 0, INT_MAX, FF_MPV_OPT_FLAGS }, \
600 {"lmax", "maximum Lagrange factor (VBR)", FF_MPV_OFFSET(lmax), AV_OPT_TYPE_INT, {.i64 = 31*FF_QP2LAMBDA }, 0, INT_MAX, FF_MPV_OPT_FLAGS }, \
601 {"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 }, \
602 {"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 }, \
603 {"rc_strategy", "ratecontrol method", FF_MPV_OFFSET(rc_strategy), AV_OPT_TYPE_INT, {.i64 = 0 }, 0, 1, FF_MPV_OPT_FLAGS }, \
604 {"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" }, \
605 { "zero", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FF_ME_ZERO }, 0, 0, FF_MPV_OPT_FLAGS, "motion_est" }, \
606 { "epzs", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FF_ME_EPZS }, 0, 0, FF_MPV_OPT_FLAGS, "motion_est" }, \
607 { "xone", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FF_ME_XONE }, 0, 0, FF_MPV_OPT_FLAGS, "motion_est" }, \
608 {"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 }, \
609 {"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 }, \
610 {"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 }, \
611 {"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 }, \
612 {"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 }, \
613 {"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 }, \
614 {"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" }, \
615 {"sc_threshold", "Scene change threshold", FF_MPV_OFFSET(scenechange_threshold), AV_OPT_TYPE_INT, {.i64 = 0 }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS }, \
616 {"noise_reduction", "Noise reduction", FF_MPV_OFFSET(noise_reduction), AV_OPT_TYPE_INT, {.i64 = 0 }, INT_MIN, INT_MAX, FF_MPV_OPT_FLAGS }, \
617 {"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 }, \
618 {"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 }, \
619
620 extern const AVOption ff_mpv_generic_options[];
621
622 /**
623 * Set the given MpegEncContext to common defaults (same for encoding
624 * and decoding). The changed fields will not depend upon the prior
625 * state of the MpegEncContext.
626 */
627 void ff_mpv_common_defaults(MpegEncContext *s);
628
629 int ff_mpv_common_init(MpegEncContext *s);
630 void ff_mpv_common_init_arm(MpegEncContext *s);
631 void ff_mpv_common_init_neon(MpegEncContext *s);
632 void ff_mpv_common_init_ppc(MpegEncContext *s);
633 void ff_mpv_common_init_x86(MpegEncContext *s);
634
635 int ff_mpv_common_frame_size_change(MpegEncContext *s);
636 void ff_mpv_common_end(MpegEncContext *s);
637
638 void ff_mpv_decode_defaults(MpegEncContext *s);
639 void ff_mpv_decode_mb(MpegEncContext *s, int16_t block[12][64]);
640 void ff_mpv_report_decode_progress(MpegEncContext *s);
641
642 int ff_mpv_frame_start(MpegEncContext *s, AVCodecContext *avctx);
643 void ff_mpv_frame_end(MpegEncContext *s);
644
645 int ff_mpv_encode_init(AVCodecContext *avctx);
646 void ff_mpv_encode_init_x86(MpegEncContext *s);
647
648 int ff_mpv_encode_end(AVCodecContext *avctx);
649 int ff_mpv_encode_picture(AVCodecContext *avctx, AVPacket *pkt,
650 const AVFrame *frame, int *got_packet);
651
652 void ff_clean_intra_table_entries(MpegEncContext *s);
653 void ff_mpeg_draw_horiz_band(MpegEncContext *s, int y, int h);
654 void ff_mpeg_flush(AVCodecContext *avctx);
655 void ff_print_debug_info(MpegEncContext *s, Picture *p);
656 void ff_write_quant_matrix(PutBitContext *pb, uint16_t *matrix);
657
658 int ff_update_duplicate_context(MpegEncContext *dst, MpegEncContext *src);
659 int ff_mpeg_update_thread_context(AVCodecContext *dst, const AVCodecContext *src);
660 void ff_set_qscale(MpegEncContext * s, int qscale);
661
662 void ff_mpv_idct_init(MpegEncContext *s);
663 void ff_convert_matrix(MpegEncContext *s, int (*qmat)[64], uint16_t (*qmat16)[2][64],
664 const uint16_t *quant_matrix, int bias, int qmin, int qmax, int intra);
665 int ff_dct_quantize_c(MpegEncContext *s, int16_t *block, int n, int qscale, int *overflow);
666
667 void ff_init_block_index(MpegEncContext *s);
668
669 void ff_mpv_motion(MpegEncContext *s,
670 uint8_t *dest_y, uint8_t *dest_cb,
671 uint8_t *dest_cr, int dir,
672 uint8_t **ref_picture,
673 op_pixels_func (*pix_op)[4],
674 qpel_mc_func (*qpix_op)[16]);
675
676 static inline void ff_update_block_index(MpegEncContext *s){
677 const int block_size = 8;
678
679 s->block_index[0]+=2;
680 s->block_index[1]+=2;
681 s->block_index[2]+=2;
682 s->block_index[3]+=2;
683 s->block_index[4]++;
684 s->block_index[5]++;
685 s->dest[0]+= 2*block_size;
686 s->dest[1]+= block_size;
687 s->dest[2]+= block_size;
688 }
689
690 static inline int get_bits_diff(MpegEncContext *s){
691 const int bits= put_bits_count(&s->pb);
692 const int last= s->last_bits;
693
694 s->last_bits = bits;
695
696 return bits - last;
697 }
698
699 #endif /* AVCODEC_MPEGVIDEO_H */