MPEG-2 4:2:2 encoding support
[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 library is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU Lesser General Public
8 * License as published by the Free Software Foundation; either
9 * version 2 of the License, or (at your option) any later version.
10 *
11 * This library is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 * Lesser General Public License for more details.
15 *
16 * You should have received a copy of the GNU Lesser General Public
17 * License along with this library; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19 */
20
21 /**
22 * @file mpegvideo.h
23 * mpegvideo header.
24 */
25
26 #ifndef AVCODEC_MPEGVIDEO_H
27 #define AVCODEC_MPEGVIDEO_H
28
29 #include "dsputil.h"
30 #include "bitstream.h"
31
32 #define FRAME_SKIPPED 100 ///< return value for header parsers if frame is not coded
33
34 enum OutputFormat {
35 FMT_MPEG1,
36 FMT_H261,
37 FMT_H263,
38 FMT_MJPEG,
39 FMT_H264,
40 };
41
42 #define EDGE_WIDTH 16
43
44 #define MPEG_BUF_SIZE (16 * 1024)
45
46 #define QMAT_SHIFT_MMX 16
47 #define QMAT_SHIFT 22
48
49 #define MAX_FCODE 7
50 #define MAX_MV 2048
51
52 #define MAX_THREADS 8
53
54 #define MAX_PICTURE_COUNT 32
55
56 #define ME_MAP_SIZE 64
57 #define ME_MAP_SHIFT 3
58 #define ME_MAP_MV_BITS 11
59
60 /* run length table */
61 #define MAX_RUN 64
62 #define MAX_LEVEL 64
63
64 #define I_TYPE FF_I_TYPE ///< Intra
65 #define P_TYPE FF_P_TYPE ///< Predicted
66 #define B_TYPE FF_B_TYPE ///< Bi-dir predicted
67 #define S_TYPE FF_S_TYPE ///< S(GMC)-VOP MPEG4
68 #define SI_TYPE FF_SI_TYPE ///< Switching Intra
69 #define SP_TYPE FF_SP_TYPE ///< Switching Predicted
70
71 #define MAX_MB_BYTES (30*16*16*3/8 + 120)
72
73 #define INPLACE_OFFSET 16
74
75 typedef struct Predictor{
76 double coeff;
77 double count;
78 double decay;
79 } Predictor;
80
81 typedef struct RateControlEntry{
82 int pict_type;
83 float qscale;
84 int mv_bits;
85 int i_tex_bits;
86 int p_tex_bits;
87 int misc_bits;
88 int header_bits;
89 uint64_t expected_bits;
90 int new_pict_type;
91 float new_qscale;
92 int mc_mb_var_sum;
93 int mb_var_sum;
94 int i_count;
95 int skip_count;
96 int f_code;
97 int b_code;
98 }RateControlEntry;
99
100 /**
101 * rate control context.
102 */
103 typedef struct RateControlContext{
104 FILE *stats_file;
105 int num_entries; ///< number of RateControlEntries
106 RateControlEntry *entry;
107 double buffer_index; ///< amount of bits in the video/audio buffer
108 Predictor pred[5];
109 double short_term_qsum; ///< sum of recent qscales
110 double short_term_qcount; ///< count of recent qscales
111 double pass1_rc_eq_output_sum;///< sum of the output of the rc equation, this is used for normalization
112 double pass1_wanted_bits; ///< bits which should have been outputed by the pass1 code (including complexity init)
113 double last_qscale;
114 double last_qscale_for[5]; ///< last qscale for a specific pict type, used for max_diff & ipb factor stuff
115 int last_mc_mb_var_sum;
116 int last_mb_var_sum;
117 uint64_t i_cplx_sum[5];
118 uint64_t p_cplx_sum[5];
119 uint64_t mv_bits_sum[5];
120 uint64_t qscale_sum[5];
121 int frame_count[5];
122 int last_non_b_pict_type;
123
124 void *non_lavc_opaque; ///< context for non lavc rc code (for example xvid)
125 float dry_run_qscale; ///< for xvid rc
126 int last_picture_number; ///< for xvid rc
127 }RateControlContext;
128
129 /**
130 * Scantable.
131 */
132 typedef struct ScanTable{
133 const uint8_t *scantable;
134 uint8_t permutated[64];
135 uint8_t raster_end[64];
136 #ifdef ARCH_POWERPC
137 /** Used by dct_quantise_alitvec to find last-non-zero */
138 DECLARE_ALIGNED_8(uint8_t, inverse[64]);
139 #endif
140 } ScanTable;
141
142 /**
143 * Picture.
144 */
145 typedef struct Picture{
146 FF_COMMON_FRAME
147
148 /**
149 * halfpel luma planes.
150 */
151 uint8_t *interpolated[3];
152 int16_t (*motion_val_base[2])[2];
153 uint32_t *mb_type_base;
154 #define MB_TYPE_INTRA MB_TYPE_INTRA4x4 //default mb_type if theres just one type
155 #define IS_INTRA4x4(a) ((a)&MB_TYPE_INTRA4x4)
156 #define IS_INTRA16x16(a) ((a)&MB_TYPE_INTRA16x16)
157 #define IS_PCM(a) ((a)&MB_TYPE_INTRA_PCM)
158 #define IS_INTRA(a) ((a)&7)
159 #define IS_INTER(a) ((a)&(MB_TYPE_16x16|MB_TYPE_16x8|MB_TYPE_8x16|MB_TYPE_8x8))
160 #define IS_SKIP(a) ((a)&MB_TYPE_SKIP)
161 #define IS_INTRA_PCM(a) ((a)&MB_TYPE_INTRA_PCM)
162 #define IS_INTERLACED(a) ((a)&MB_TYPE_INTERLACED)
163 #define IS_DIRECT(a) ((a)&MB_TYPE_DIRECT2)
164 #define IS_GMC(a) ((a)&MB_TYPE_GMC)
165 #define IS_16X16(a) ((a)&MB_TYPE_16x16)
166 #define IS_16X8(a) ((a)&MB_TYPE_16x8)
167 #define IS_8X16(a) ((a)&MB_TYPE_8x16)
168 #define IS_8X8(a) ((a)&MB_TYPE_8x8)
169 #define IS_SUB_8X8(a) ((a)&MB_TYPE_16x16) //note reused
170 #define IS_SUB_8X4(a) ((a)&MB_TYPE_16x8) //note reused
171 #define IS_SUB_4X8(a) ((a)&MB_TYPE_8x16) //note reused
172 #define IS_SUB_4X4(a) ((a)&MB_TYPE_8x8) //note reused
173 #define IS_ACPRED(a) ((a)&MB_TYPE_ACPRED)
174 #define IS_QUANT(a) ((a)&MB_TYPE_QUANT)
175 #define IS_DIR(a, part, list) ((a) & (MB_TYPE_P0L0<<((part)+2*(list))))
176 #define USES_LIST(a, list) ((a) & ((MB_TYPE_P0L0|MB_TYPE_P1L0)<<(2*(list)))) ///< does this mb use listX, note doesnt work if subMBs
177 #define HAS_CBP(a) ((a)&MB_TYPE_CBP)
178
179 int field_poc[2]; ///< h264 top/bottom POC
180 int poc; ///< h264 frame POC
181 int frame_num; ///< h264 frame_num
182 int pic_id; ///< h264 pic_num or long_term_pic_idx
183 int long_ref; ///< 1->long term reference 0->short term reference
184 int ref_poc[2][16]; ///< h264 POCs of the frames used as reference
185 int ref_count[2]; ///< number of entries in ref_poc
186
187 int mb_var_sum; ///< sum of MB variance for current frame
188 int mc_mb_var_sum; ///< motion compensated MB variance for current frame
189 uint16_t *mb_var; ///< Table for MB variances
190 uint16_t *mc_mb_var; ///< Table for motion compensated MB variances
191 uint8_t *mb_mean; ///< Table for MB luminance
192 int32_t *mb_cmp_score; ///< Table for MB cmp scores, for mb decision FIXME remove
193 int b_frame_score; /* */
194 } Picture;
195
196 typedef struct ParseContext{
197 uint8_t *buffer;
198 int index;
199 int last_index;
200 unsigned int buffer_size;
201 uint32_t state; ///< contains the last few bytes in MSB order
202 int frame_start_found;
203 int overread; ///< the number of bytes which where irreversibly read from the next frame
204 int overread_index; ///< the index into ParseContext.buffer of the overreaded bytes
205 } ParseContext;
206
207 struct MpegEncContext;
208
209 /**
210 * Motion estimation context.
211 */
212 typedef struct MotionEstContext{
213 AVCodecContext *avctx;
214 int skip; ///< set if ME is skipped for the current MB
215 int co_located_mv[4][2]; ///< mv from last p frame for direct mode ME
216 int direct_basis_mv[4][2];
217 uint8_t *scratchpad; ///< data area for the me algo, so that the ME doesnt need to malloc/free
218 uint8_t *best_mb;
219 uint8_t *temp_mb[2];
220 uint8_t *temp;
221 int best_bits;
222 uint32_t *map; ///< map to avoid duplicate evaluations
223 uint32_t *score_map; ///< map to store the scores
224 int map_generation;
225 int pre_penalty_factor;
226 int penalty_factor;
227 int sub_penalty_factor;
228 int mb_penalty_factor;
229 int flags;
230 int sub_flags;
231 int mb_flags;
232 int pre_pass; ///< = 1 for the pre pass
233 int dia_size;
234 int xmin;
235 int xmax;
236 int ymin;
237 int ymax;
238 int pred_x;
239 int pred_y;
240 uint8_t *src[4][4];
241 uint8_t *ref[4][4];
242 int stride;
243 int uvstride;
244 /* temp variables for picture complexity calculation */
245 int mc_mb_var_sum_temp;
246 int mb_var_sum_temp;
247 int scene_change_score;
248 /* cmp, chroma_cmp;*/
249 op_pixels_func (*hpel_put)[4];
250 op_pixels_func (*hpel_avg)[4];
251 qpel_mc_func (*qpel_put)[16];
252 qpel_mc_func (*qpel_avg)[16];
253 uint8_t (*mv_penalty)[MAX_MV*2+1]; ///< amount of bits needed to encode a MV
254 uint8_t *current_mv_penalty;
255 int (*sub_motion_search)(struct MpegEncContext * s,
256 int *mx_ptr, int *my_ptr, int dmin,
257 int src_index, int ref_index,
258 int size, int h);
259 }MotionEstContext;
260
261 /**
262 * MpegEncContext.
263 */
264 typedef struct MpegEncContext {
265 struct AVCodecContext *avctx;
266 /* the following parameters must be initialized before encoding */
267 int width, height;///< picture size. must be a multiple of 16
268 int gop_size;
269 int intra_only; ///< if true, only intra pictures are generated
270 int bit_rate; ///< wanted bit rate
271 enum OutputFormat out_format; ///< output format
272 int h263_pred; ///< use mpeg4/h263 ac/dc predictions
273
274 /* the following codec id fields are deprecated in favor of codec_id */
275 int h263_plus; ///< h263 plus headers
276 int h263_msmpeg4; ///< generate MSMPEG4 compatible stream (deprecated, use msmpeg4_version instead)
277 int h263_flv; ///< use flv h263 header
278
279 enum CodecID codec_id; /* see CODEC_ID_xxx */
280 int fixed_qscale; ///< fixed qscale if non zero
281 int encoding; ///< true if we are encoding (vs decoding)
282 int flags; ///< AVCodecContext.flags (HQ, MV4, ...)
283 int flags2; ///< AVCodecContext.flags2
284 int max_b_frames; ///< max number of b-frames for encoding
285 int luma_elim_threshold;
286 int chroma_elim_threshold;
287 int strict_std_compliance; ///< strictly follow the std (MPEG4, ...)
288 int workaround_bugs; ///< workaround bugs in encoders which cannot be detected automatically
289 /* the following fields are managed internally by the encoder */
290
291 /** bit output */
292 PutBitContext pb;
293
294 /* sequence parameters */
295 int context_initialized;
296 int input_picture_number; ///< used to set pic->display_picture_number, shouldnt be used for/by anything else
297 int coded_picture_number; ///< used to set pic->coded_picture_number, shouldnt be used for/by anything else
298 int picture_number; //FIXME remove, unclear definition
299 int picture_in_gop_number; ///< 0-> first pic in gop, ...
300 int b_frames_since_non_b; ///< used for encoding, relative to not yet reordered input
301 int64_t user_specified_pts;///< last non zero pts from AVFrame which was passed into avcodec_encode_video()
302 int mb_width, mb_height; ///< number of MBs horizontally & vertically
303 int mb_stride; ///< mb_width+1 used for some arrays to allow simple addressing of left & top MBs without sig11
304 int b8_stride; ///< 2*mb_width+1 used for some 8x8 block arrays to allow simple addressing
305 int b4_stride; ///< 4*mb_width+1 used for some 4x4 block arrays to allow simple addressing
306 int h_edge_pos, v_edge_pos;///< horizontal / vertical position of the right/bottom edge (pixel replication)
307 int mb_num; ///< number of MBs of a picture
308 int linesize; ///< line size, in bytes, may be different from width
309 int uvlinesize; ///< line size, for chroma in bytes, may be different from width
310 Picture *picture; ///< main picture buffer
311 Picture **input_picture; ///< next pictures on display order for encoding
312 Picture **reordered_input_picture; ///< pointer to the next pictures in codedorder for encoding
313
314 int start_mb_y; ///< start mb_y of this thread (so current thread should process start_mb_y <= row < end_mb_y)
315 int end_mb_y; ///< end mb_y of this thread (so current thread should process start_mb_y <= row < end_mb_y)
316 struct MpegEncContext *thread_context[MAX_THREADS];
317
318 /**
319 * copy of the previous picture structure.
320 * note, linesize & data, might not match the previous picture (for field pictures)
321 */
322 Picture last_picture;
323
324 /**
325 * copy of the next picture structure.
326 * note, linesize & data, might not match the next picture (for field pictures)
327 */
328 Picture next_picture;
329
330 /**
331 * copy of the source picture structure for encoding.
332 * note, linesize & data, might not match the source picture (for field pictures)
333 */
334 Picture new_picture;
335
336 /**
337 * copy of the current picture structure.
338 * note, linesize & data, might not match the current picture (for field pictures)
339 */
340 Picture current_picture; ///< buffer to store the decompressed current picture
341
342 Picture *last_picture_ptr; ///< pointer to the previous picture.
343 Picture *next_picture_ptr; ///< pointer to the next picture (for bidir pred)
344 Picture *current_picture_ptr; ///< pointer to the current picture
345 uint8_t *visualization_buffer[3]; //< temporary buffer vor MV visualization
346 int last_dc[3]; ///< last DC values for MPEG1
347 int16_t *dc_val_base;
348 int16_t *dc_val[3]; ///< used for mpeg4 DC prediction, all 3 arrays must be continuous
349 int16_t dc_cache[4*5];
350 int y_dc_scale, c_dc_scale;
351 const uint8_t *y_dc_scale_table; ///< qscale -> y_dc_scale table
352 const uint8_t *c_dc_scale_table; ///< qscale -> c_dc_scale table
353 const uint8_t *chroma_qscale_table; ///< qscale -> chroma_qscale (h263)
354 uint8_t *coded_block_base;
355 uint8_t *coded_block; ///< used for coded block pattern prediction (msmpeg4v3, wmv1)
356 int16_t (*ac_val_base)[16];
357 int16_t (*ac_val[3])[16]; ///< used for for mpeg4 AC prediction, all 3 arrays must be continuous
358 int ac_pred;
359 uint8_t *prev_pict_types; ///< previous picture types in bitstream order, used for mb skip
360 #define PREV_PICT_TYPES_BUFFER_SIZE 256
361 int mb_skipped; ///< MUST BE SET only during DECODING
362 uint8_t *mbskip_table; /**< used to avoid copy if macroblock skipped (for black regions for example)
363 and used for b-frame encoding & decoding (contains skip table of next P Frame) */
364 uint8_t *mbintra_table; ///< used to avoid setting {ac, dc, cbp}-pred stuff to zero on inter MB decoding
365 uint8_t *cbp_table; ///< used to store cbp, ac_pred for partitioned decoding
366 uint8_t *pred_dir_table; ///< used to store pred_dir for partitioned decoding
367 uint8_t *allocated_edge_emu_buffer;
368 uint8_t *edge_emu_buffer; ///< points into the middle of allocated_edge_emu_buffer
369 uint8_t *rd_scratchpad; ///< scratchpad for rate distortion mb decision
370 uint8_t *obmc_scratchpad;
371 uint8_t *b_scratchpad; ///< scratchpad used for writing into write only buffers
372
373 int qscale; ///< QP
374 int chroma_qscale; ///< chroma QP
375 int lambda; ///< lagrange multipler used in rate distortion
376 int lambda2; ///< (lambda*lambda) >> FF_LAMBDA_SHIFT
377 int *lambda_table;
378 int adaptive_quant; ///< use adaptive quantization
379 int dquant; ///< qscale difference to prev qscale
380 int pict_type; ///< I_TYPE, P_TYPE, B_TYPE, ...
381 int last_pict_type; //FIXME removes
382 int last_non_b_pict_type; ///< used for mpeg4 gmc b-frames & ratecontrol
383 int dropable;
384 int frame_rate_index;
385 int last_lambda_for[5]; ///< last lambda for a specific pict type
386
387 /* motion compensation */
388 int unrestricted_mv; ///< mv can point outside of the coded picture
389 int h263_long_vectors; ///< use horrible h263v1 long vector mode
390 int decode; ///< if 0 then decoding will be skipped (for encoding b frames for example)
391
392 DSPContext dsp; ///< pointers for accelerated dsp functions
393 int f_code; ///< forward MV resolution
394 int b_code; ///< backward MV resolution for B Frames (mpeg4)
395 int16_t (*p_mv_table_base)[2];
396 int16_t (*b_forw_mv_table_base)[2];
397 int16_t (*b_back_mv_table_base)[2];
398 int16_t (*b_bidir_forw_mv_table_base)[2];
399 int16_t (*b_bidir_back_mv_table_base)[2];
400 int16_t (*b_direct_mv_table_base)[2];
401 int16_t (*p_field_mv_table_base[2][2])[2];
402 int16_t (*b_field_mv_table_base[2][2][2])[2];
403 int16_t (*p_mv_table)[2]; ///< MV table (1MV per MB) p-frame encoding
404 int16_t (*b_forw_mv_table)[2]; ///< MV table (1MV per MB) forward mode b-frame encoding
405 int16_t (*b_back_mv_table)[2]; ///< MV table (1MV per MB) backward mode b-frame encoding
406 int16_t (*b_bidir_forw_mv_table)[2]; ///< MV table (1MV per MB) bidir mode b-frame encoding
407 int16_t (*b_bidir_back_mv_table)[2]; ///< MV table (1MV per MB) bidir mode b-frame encoding
408 int16_t (*b_direct_mv_table)[2]; ///< MV table (1MV per MB) direct mode b-frame encoding
409 int16_t (*p_field_mv_table[2][2])[2]; ///< MV table (2MV per MB) interlaced p-frame encoding
410 int16_t (*b_field_mv_table[2][2][2])[2];///< MV table (4MV per MB) interlaced b-frame encoding
411 uint8_t (*p_field_select_table[2]);
412 uint8_t (*b_field_select_table[2][2]);
413 int me_method; ///< ME algorithm
414 int mv_dir;
415 #define MV_DIR_BACKWARD 1
416 #define MV_DIR_FORWARD 2
417 #define MV_DIRECT 4 ///< bidirectional mode where the difference equals the MV of the last P/S/I-Frame (mpeg4)
418 int mv_type;
419 #define MV_TYPE_16X16 0 ///< 1 vector for the whole mb
420 #define MV_TYPE_8X8 1 ///< 4 vectors (h263, mpeg4 4MV)
421 #define MV_TYPE_16X8 2 ///< 2 vectors, one per 16x8 block
422 #define MV_TYPE_FIELD 3 ///< 2 vectors, one per field
423 #define MV_TYPE_DMV 4 ///< 2 vectors, special mpeg2 Dual Prime Vectors
424 /**motion vectors for a macroblock
425 first coordinate : 0 = forward 1 = backward
426 second " : depend on type
427 third " : 0 = x, 1 = y
428 */
429 int mv[2][4][2];
430 int field_select[2][2];
431 int last_mv[2][2][2]; ///< last MV, used for MV prediction in MPEG1 & B-frame MPEG4
432 uint8_t *fcode_tab; ///< smallest fcode needed for each MV
433 int16_t direct_scale_mv[2][64]; ///< precomputed to avoid divisions in ff_mpeg4_set_direct_mv
434
435 MotionEstContext me;
436
437 int no_rounding; /**< apply no rounding to motion compensation (MPEG4, msmpeg4, ...)
438 for b-frames rounding mode is allways 0 */
439
440 int hurry_up; /**< when set to 1 during decoding, b frames will be skipped
441 when set to 2 idct/dequant will be skipped too */
442
443 /* macroblock layer */
444 int mb_x, mb_y;
445 int mb_skip_run;
446 int mb_intra;
447 uint16_t *mb_type; ///< Table for candidate MB types for encoding
448 #define CANDIDATE_MB_TYPE_INTRA 0x01
449 #define CANDIDATE_MB_TYPE_INTER 0x02
450 #define CANDIDATE_MB_TYPE_INTER4V 0x04
451 #define CANDIDATE_MB_TYPE_SKIPPED 0x08
452 //#define MB_TYPE_GMC 0x10
453
454 #define CANDIDATE_MB_TYPE_DIRECT 0x10
455 #define CANDIDATE_MB_TYPE_FORWARD 0x20
456 #define CANDIDATE_MB_TYPE_BACKWARD 0x40
457 #define CANDIDATE_MB_TYPE_BIDIR 0x80
458
459 #define CANDIDATE_MB_TYPE_INTER_I 0x100
460 #define CANDIDATE_MB_TYPE_FORWARD_I 0x200
461 #define CANDIDATE_MB_TYPE_BACKWARD_I 0x400
462 #define CANDIDATE_MB_TYPE_BIDIR_I 0x800
463
464 int block_index[6]; ///< index to current MB in block based arrays with edges
465 int block_wrap[6];
466 uint8_t *dest[3];
467
468 int *mb_index2xy; ///< mb_index -> mb_x + mb_y*mb_stride
469
470 /** matrix transmitted in the bitstream */
471 uint16_t intra_matrix[64];
472 uint16_t chroma_intra_matrix[64];
473 uint16_t inter_matrix[64];
474 uint16_t chroma_inter_matrix[64];
475 #define QUANT_BIAS_SHIFT 8
476 int intra_quant_bias; ///< bias for the quantizer
477 int inter_quant_bias; ///< bias for the quantizer
478 int min_qcoeff; ///< minimum encodable coefficient
479 int max_qcoeff; ///< maximum encodable coefficient
480 int ac_esc_length; ///< num of bits needed to encode the longest esc
481 uint8_t *intra_ac_vlc_length;
482 uint8_t *intra_ac_vlc_last_length;
483 uint8_t *inter_ac_vlc_length;
484 uint8_t *inter_ac_vlc_last_length;
485 uint8_t *luma_dc_vlc_length;
486 uint8_t *chroma_dc_vlc_length;
487 #define UNI_AC_ENC_INDEX(run,level) ((run)*128 + (level))
488
489 int coded_score[8];
490
491 /** precomputed matrix (combine qscale and DCT renorm) */
492 int (*q_intra_matrix)[64];
493 int (*q_inter_matrix)[64];
494 /** identical to the above but for MMX & these are not permutated, second 64 entries are bias*/
495 uint16_t (*q_intra_matrix16)[2][64];
496 uint16_t (*q_inter_matrix16)[2][64];
497 int block_last_index[12]; ///< last non zero coefficient in block
498 /* scantables */
499 DECLARE_ALIGNED_8(ScanTable, intra_scantable);
500 ScanTable intra_h_scantable;
501 ScanTable intra_v_scantable;
502 ScanTable inter_scantable; ///< if inter == intra then intra should be used to reduce tha cache usage
503
504 /* noise reduction */
505 int (*dct_error_sum)[64];
506 int dct_count[2];
507 uint16_t (*dct_offset)[64];
508
509 void *opaque; ///< private data for the user
510
511 /* bit rate control */
512 int64_t wanted_bits;
513 int64_t total_bits;
514 int frame_bits; ///< bits used for the current frame
515 RateControlContext rc_context; ///< contains stuff only accessed in ratecontrol.c
516
517 /* statistics, used for 2-pass encoding */
518 int mv_bits;
519 int header_bits;
520 int i_tex_bits;
521 int p_tex_bits;
522 int i_count;
523 int f_count;
524 int b_count;
525 int skip_count;
526 int misc_bits; ///< cbp, mb_type
527 int last_bits; ///< temp var used for calculating the above vars
528
529 /* error concealment / resync */
530 int error_count;
531 uint8_t *error_status_table; ///< table of the error status of each MB
532 #define VP_START 1 ///< current MB is the first after a resync marker
533 #define AC_ERROR 2
534 #define DC_ERROR 4
535 #define MV_ERROR 8
536 #define AC_END 16
537 #define DC_END 32
538 #define MV_END 64
539 //FIXME some prefix?
540
541 int resync_mb_x; ///< x position of last resync marker
542 int resync_mb_y; ///< y position of last resync marker
543 GetBitContext last_resync_gb; ///< used to search for the next resync marker
544 int mb_num_left; ///< number of MBs left in this video packet (for partitioned Slices only)
545 int next_p_frame_damaged; ///< set if the next p frame is damaged, to avoid showing trashed b frames
546 int error_resilience;
547
548 ParseContext parse_context;
549
550 /* H.263 specific */
551 int gob_index;
552 int obmc; ///< overlapped block motion compensation
553
554 /* H.263+ specific */
555 int umvplus; ///< == H263+ && unrestricted_mv
556 int h263_aic; ///< Advanded INTRA Coding (AIC)
557 int h263_aic_dir; ///< AIC direction: 0 = left, 1 = top
558 int h263_slice_structured;
559 int alt_inter_vlc; ///< alternative inter vlc
560 int modified_quant;
561 int loop_filter;
562 int custom_pcf;
563
564 /* mpeg4 specific */
565 int time_increment_bits; ///< number of bits to represent the fractional part of time
566 int last_time_base;
567 int time_base; ///< time in seconds of last I,P,S Frame
568 int64_t time; ///< time of current frame
569 int64_t last_non_b_time;
570 uint16_t pp_time; ///< time distance between the last 2 p,s,i frames
571 uint16_t pb_time; ///< time distance between the last b and p,s,i frame
572 uint16_t pp_field_time;
573 uint16_t pb_field_time; ///< like above, just for interlaced
574 int shape;
575 int vol_sprite_usage;
576 int sprite_width;
577 int sprite_height;
578 int sprite_left;
579 int sprite_top;
580 int sprite_brightness_change;
581 int num_sprite_warping_points;
582 int real_sprite_warping_points;
583 int sprite_offset[2][2]; ///< sprite offset[isChroma][isMVY]
584 int sprite_delta[2][2]; ///< sprite_delta [isY][isMVY]
585 int sprite_shift[2]; ///< sprite shift [isChroma]
586 int mcsel;
587 int quant_precision;
588 int quarter_sample; ///< 1->qpel, 0->half pel ME/MC
589 int scalability;
590 int hierachy_type;
591 int enhancement_type;
592 int new_pred;
593 int reduced_res_vop;
594 int aspect_ratio_info; //FIXME remove
595 int sprite_warping_accuracy;
596 int low_latency_sprite;
597 int data_partitioning; ///< data partitioning flag from header
598 int partitioned_frame; ///< is current frame partitioned
599 int rvlc; ///< reversible vlc
600 int resync_marker; ///< could this stream contain resync markers
601 int low_delay; ///< no reordering needed / has no b-frames
602 int vo_type;
603 int vol_control_parameters; ///< does the stream contain the low_delay flag, used to workaround buggy encoders
604 int intra_dc_threshold; ///< QP above whch the ac VLC should be used for intra dc
605 int use_intra_dc_vlc;
606 PutBitContext tex_pb; ///< used for data partitioned VOPs
607 PutBitContext pb2; ///< used for data partitioned VOPs
608 int mpeg_quant;
609 int t_frame; ///< time distance of first I -> B, used for interlaced b frames
610 int padding_bug_score; ///< used to detect the VERY common padding bug in MPEG4
611
612 /* divx specific, used to workaround (many) bugs in divx5 */
613 int divx_version;
614 int divx_build;
615 int divx_packed;
616 uint8_t *bitstream_buffer; //Divx 5.01 puts several frames in a single one, this is used to reorder them
617 int bitstream_buffer_size;
618 unsigned int allocated_bitstream_buffer_size;
619
620 int xvid_build;
621
622 /* lavc specific stuff, used to workaround bugs in libavcodec */
623 int lavc_build;
624
625 /* RV10 specific */
626 int rv10_version; ///< RV10 version: 0 or 3
627 int rv10_first_dc_coded[3];
628
629 /* MJPEG specific */
630 struct MJpegContext *mjpeg_ctx;
631 int mjpeg_vsample[3]; ///< vertical sampling factors, default = {2, 1, 1}
632 int mjpeg_hsample[3]; ///< horizontal sampling factors, default = {2, 1, 1}
633 int mjpeg_write_tables; ///< do we want to have quantisation- and huffmantables in the jpeg file ?
634 int mjpeg_data_only_frames; ///< frames only with SOI, SOS and EOI markers
635
636 /* MSMPEG4 specific */
637 int mv_table_index;
638 int rl_table_index;
639 int rl_chroma_table_index;
640 int dc_table_index;
641 int use_skip_mb_code;
642 int slice_height; ///< in macroblocks
643 int first_slice_line; ///< used in mpeg4 too to handle resync markers
644 int flipflop_rounding;
645 int msmpeg4_version; ///< 0=not msmpeg4, 1=mp41, 2=mp42, 3=mp43/divx3 4=wmv1/7 5=wmv2/8
646 int per_mb_rl_table;
647 int esc3_level_length;
648 int esc3_run_length;
649 /** [mb_intra][isChroma][level][run][last] */
650 int (*ac_stats)[2][MAX_LEVEL+1][MAX_RUN+1][2];
651 int inter_intra_pred;
652 int mspel;
653
654 /* decompression specific */
655 GetBitContext gb;
656
657 /* Mpeg1 specific */
658 int gop_picture_number; ///< index of the first picture of a GOP based on fake_pic_num & mpeg1 specific
659 int last_mv_dir; ///< last mv_dir, used for b frame encoding
660 int broken_link; ///< no_output_of_prior_pics_flag
661 uint8_t *vbv_delay_ptr; ///< pointer to vbv_delay in the bitstream
662
663 /* MPEG2 specific - I wish I had not to support this mess. */
664 int progressive_sequence;
665 int mpeg_f_code[2][2];
666 int picture_structure;
667 /* picture type */
668 #define PICT_TOP_FIELD 1
669 #define PICT_BOTTOM_FIELD 2
670 #define PICT_FRAME 3
671
672 int intra_dc_precision;
673 int frame_pred_frame_dct;
674 int top_field_first;
675 int concealment_motion_vectors;
676 int q_scale_type;
677 int intra_vlc_format;
678 int alternate_scan;
679 int repeat_first_field;
680 int chroma_420_type;
681 int chroma_format;
682 #define CHROMA_420 1
683 #define CHROMA_422 2
684 #define CHROMA_444 3
685 int chroma_x_shift;//depend on pix_format, that depend on chroma_format
686 int chroma_y_shift;
687
688 int progressive_frame;
689 int full_pel[2];
690 int interlaced_dct;
691 int first_slice;
692 int first_field; ///< is 1 for the first field of a field picture 0 otherwise
693
694 /* RTP specific */
695 int rtp_mode;
696
697 uint8_t *ptr_lastgob;
698 int swap_uv;//vcr2 codec is mpeg2 varint with UV swaped
699 short * pblocks[12];
700
701 DCTELEM (*block)[64]; ///< points to one of the following blocks
702 DCTELEM (*blocks)[6][64]; // for HQ mode we need to keep the best block
703 int (*decode_mb)(struct MpegEncContext *s, DCTELEM block[6][64]); // used by some codecs to avoid a switch()
704 #define SLICE_OK 0
705 #define SLICE_ERROR -1
706 #define SLICE_END -2 ///<end marker found
707 #define SLICE_NOEND -3 ///<no end marker or error found but mb count exceeded
708
709 void (*dct_unquantize_mpeg1_intra)(struct MpegEncContext *s,
710 DCTELEM *block/*align 16*/, int n, int qscale);
711 void (*dct_unquantize_mpeg1_inter)(struct MpegEncContext *s,
712 DCTELEM *block/*align 16*/, int n, int qscale);
713 void (*dct_unquantize_mpeg2_intra)(struct MpegEncContext *s,
714 DCTELEM *block/*align 16*/, int n, int qscale);
715 void (*dct_unquantize_mpeg2_inter)(struct MpegEncContext *s,
716 DCTELEM *block/*align 16*/, int n, int qscale);
717 void (*dct_unquantize_h263_intra)(struct MpegEncContext *s,
718 DCTELEM *block/*align 16*/, int n, int qscale);
719 void (*dct_unquantize_h263_inter)(struct MpegEncContext *s,
720 DCTELEM *block/*align 16*/, int n, int qscale);
721 void (*dct_unquantize_h261_intra)(struct MpegEncContext *s,
722 DCTELEM *block/*align 16*/, int n, int qscale);
723 void (*dct_unquantize_h261_inter)(struct MpegEncContext *s,
724 DCTELEM *block/*align 16*/, int n, int qscale);
725 void (*dct_unquantize_intra)(struct MpegEncContext *s, // unquantizer to use (mpeg4 can use both)
726 DCTELEM *block/*align 16*/, int n, int qscale);
727 void (*dct_unquantize_inter)(struct MpegEncContext *s, // unquantizer to use (mpeg4 can use both)
728 DCTELEM *block/*align 16*/, int n, int qscale);
729 int (*dct_quantize)(struct MpegEncContext *s, DCTELEM *block/*align 16*/, int n, int qscale, int *overflow);
730 int (*fast_dct_quantize)(struct MpegEncContext *s, DCTELEM *block/*align 16*/, int n, int qscale, int *overflow);
731 void (*denoise_dct)(struct MpegEncContext *s, DCTELEM *block);
732 } MpegEncContext;
733
734
735 int DCT_common_init(MpegEncContext *s);
736 void MPV_decode_defaults(MpegEncContext *s);
737 int MPV_common_init(MpegEncContext *s);
738 void MPV_common_end(MpegEncContext *s);
739 void MPV_decode_mb(MpegEncContext *s, DCTELEM block[12][64]);
740 int MPV_frame_start(MpegEncContext *s, AVCodecContext *avctx);
741 void MPV_frame_end(MpegEncContext *s);
742 int MPV_encode_init(AVCodecContext *avctx);
743 int MPV_encode_end(AVCodecContext *avctx);
744 int MPV_encode_picture(AVCodecContext *avctx, unsigned char *buf, int buf_size, void *data);
745 #ifdef HAVE_MMX
746 void MPV_common_init_mmx(MpegEncContext *s);
747 #endif
748 #ifdef ARCH_ALPHA
749 void MPV_common_init_axp(MpegEncContext *s);
750 #endif
751 #ifdef HAVE_MLIB
752 void MPV_common_init_mlib(MpegEncContext *s);
753 #endif
754 #ifdef HAVE_MMI
755 void MPV_common_init_mmi(MpegEncContext *s);
756 #endif
757 #ifdef ARCH_ARMV4L
758 void MPV_common_init_armv4l(MpegEncContext *s);
759 #endif
760 #ifdef ARCH_POWERPC
761 void MPV_common_init_ppc(MpegEncContext *s);
762 #endif
763 extern void (*draw_edges)(uint8_t *buf, int wrap, int width, int height, int w);
764 void ff_copy_bits(PutBitContext *pb, uint8_t *src, int length);
765 void ff_clean_intra_table_entries(MpegEncContext *s);
766 void ff_init_scantable(uint8_t *, ScanTable *st, const uint8_t *src_scantable);
767 void ff_draw_horiz_band(MpegEncContext *s, int y, int h);
768 void ff_emulated_edge_mc(uint8_t *buf, uint8_t *src, int linesize, int block_w, int block_h,
769 int src_x, int src_y, int w, int h);
770 #define END_NOT_FOUND -100
771 int ff_combine_frame(ParseContext *pc, int next, uint8_t **buf, int *buf_size);
772 void ff_parse_close(AVCodecParserContext *s);
773 void ff_mpeg_flush(AVCodecContext *avctx);
774 void ff_print_debug_info(MpegEncContext *s, AVFrame *pict);
775 void ff_write_quant_matrix(PutBitContext *pb, int16_t *matrix);
776 int ff_find_unused_picture(MpegEncContext *s, int shared);
777 void ff_denoise_dct(MpegEncContext *s, DCTELEM *block);
778 void ff_update_duplicate_context(MpegEncContext *dst, MpegEncContext *src);
779 const uint8_t *ff_find_start_code(const uint8_t *p, const uint8_t *end, uint32_t *state);
780
781 void ff_er_frame_start(MpegEncContext *s);
782 void ff_er_frame_end(MpegEncContext *s);
783 void ff_er_add_slice(MpegEncContext *s, int startx, int starty, int endx, int endy, int status);
784
785
786 extern enum PixelFormat ff_yuv420p_list[2];
787
788 void ff_init_block_index(MpegEncContext *s);
789
790 static inline void ff_update_block_index(MpegEncContext *s){
791 const int block_size= 8>>s->avctx->lowres;
792
793 s->block_index[0]+=2;
794 s->block_index[1]+=2;
795 s->block_index[2]+=2;
796 s->block_index[3]+=2;
797 s->block_index[4]++;
798 s->block_index[5]++;
799 s->dest[0]+= 2*block_size;
800 s->dest[1]+= block_size;
801 s->dest[2]+= block_size;
802 }
803
804 static inline int get_bits_diff(MpegEncContext *s){
805 const int bits= put_bits_count(&s->pb);
806 const int last= s->last_bits;
807
808 s->last_bits = bits;
809
810 return bits - last;
811 }
812
813 /* motion_est.c */
814 void ff_estimate_p_frame_motion(MpegEncContext * s,
815 int mb_x, int mb_y);
816 void ff_estimate_b_frame_motion(MpegEncContext * s,
817 int mb_x, int mb_y);
818 int ff_get_best_fcode(MpegEncContext * s, int16_t (*mv_table)[2], int type);
819 void ff_fix_long_p_mvs(MpegEncContext * s);
820 void ff_fix_long_mvs(MpegEncContext * s, uint8_t *field_select_table, int field_select,
821 int16_t (*mv_table)[2], int f_code, int type, int truncate);
822 void ff_init_me(MpegEncContext *s);
823 int ff_pre_estimate_p_frame_motion(MpegEncContext * s, int mb_x, int mb_y);
824 inline int ff_epzs_motion_search(MpegEncContext * s, int *mx_ptr, int *my_ptr,
825 int P[10][2], int src_index, int ref_index, int16_t (*last_mv)[2],
826 int ref_mv_scale, int size, int h);
827 int inline ff_get_mb_score(MpegEncContext * s, int mx, int my, int src_index,
828 int ref_index, int size, int h, int add_rate);
829
830 /* mpeg12.c */
831 extern const int16_t ff_mpeg1_default_intra_matrix[64];
832 extern const int16_t ff_mpeg1_default_non_intra_matrix[64];
833 extern const uint8_t ff_mpeg1_dc_scale_table[128];
834
835 void mpeg1_encode_picture_header(MpegEncContext *s, int picture_number);
836 void mpeg1_encode_mb(MpegEncContext *s,
837 DCTELEM block[6][64],
838 int motion_x, int motion_y);
839 void ff_mpeg1_encode_init(MpegEncContext *s);
840 void ff_mpeg1_encode_slice_header(MpegEncContext *s);
841 void ff_mpeg1_clean_buffers(MpegEncContext *s);
842 int ff_mpeg1_find_frame_end(ParseContext *pc, const uint8_t *buf, int buf_size);
843
844
845 /** RLTable. */
846 typedef struct RLTable {
847 int n; ///< number of entries of table_vlc minus 1
848 int last; ///< number of values for last = 0
849 const uint16_t (*table_vlc)[2];
850 const int8_t *table_run;
851 const int8_t *table_level;
852 uint8_t *index_run[2]; ///< encoding only
853 int8_t *max_level[2]; ///< encoding & decoding
854 int8_t *max_run[2]; ///< encoding & decoding
855 VLC vlc; ///< decoding only deprected FIXME remove
856 RL_VLC_ELEM *rl_vlc[32]; ///< decoding only
857 } RLTable;
858
859 void init_rl(RLTable *rl, int use_static);
860 void init_vlc_rl(RLTable *rl, int use_static);
861
862 static inline int get_rl_index(const RLTable *rl, int last, int run, int level)
863 {
864 int index;
865 index = rl->index_run[last][run];
866 if (index >= rl->n)
867 return rl->n;
868 if (level > rl->max_level[last][run])
869 return rl->n;
870 return index + level - 1;
871 }
872
873 extern const uint8_t ff_mpeg4_y_dc_scale_table[32];
874 extern const uint8_t ff_mpeg4_c_dc_scale_table[32];
875 extern const uint8_t ff_aic_dc_scale_table[32];
876 extern const int16_t ff_mpeg4_default_intra_matrix[64];
877 extern const int16_t ff_mpeg4_default_non_intra_matrix[64];
878 extern const uint8_t ff_h263_chroma_qscale_table[32];
879 extern const uint8_t ff_h263_loop_filter_strength[32];
880
881 /* h261.c */
882 void ff_h261_loop_filter(MpegEncContext *s);
883 void ff_h261_reorder_mb_index(MpegEncContext* s);
884 void ff_h261_encode_mb(MpegEncContext *s,
885 DCTELEM block[6][64],
886 int motion_x, int motion_y);
887 void ff_h261_encode_picture_header(MpegEncContext * s, int picture_number);
888 void ff_h261_encode_init(MpegEncContext *s);
889
890
891 /* h263.c, h263dec.c */
892 int ff_h263_decode_init(AVCodecContext *avctx);
893 int ff_h263_decode_frame(AVCodecContext *avctx,
894 void *data, int *data_size,
895 uint8_t *buf, int buf_size);
896 int ff_h263_decode_end(AVCodecContext *avctx);
897 void h263_encode_mb(MpegEncContext *s,
898 DCTELEM block[6][64],
899 int motion_x, int motion_y);
900 void mpeg4_encode_mb(MpegEncContext *s,
901 DCTELEM block[6][64],
902 int motion_x, int motion_y);
903 void h263_encode_picture_header(MpegEncContext *s, int picture_number);
904 void ff_flv_encode_picture_header(MpegEncContext *s, int picture_number);
905 void h263_encode_gob_header(MpegEncContext * s, int mb_line);
906 int16_t *h263_pred_motion(MpegEncContext * s, int block, int dir,
907 int *px, int *py);
908 void mpeg4_pred_ac(MpegEncContext * s, DCTELEM *block, int n,
909 int dir);
910 void ff_set_mpeg4_time(MpegEncContext * s, int picture_number);
911 void mpeg4_encode_picture_header(MpegEncContext *s, int picture_number);
912 void h263_encode_init(MpegEncContext *s);
913 void h263_decode_init_vlc(MpegEncContext *s);
914 int h263_decode_picture_header(MpegEncContext *s);
915 int ff_h263_decode_gob_header(MpegEncContext *s);
916 int ff_mpeg4_decode_picture_header(MpegEncContext * s, GetBitContext *gb);
917 void ff_h263_update_motion_val(MpegEncContext * s);
918 void ff_h263_loop_filter(MpegEncContext * s);
919 void ff_set_qscale(MpegEncContext * s, int qscale);
920 int ff_h263_decode_mba(MpegEncContext *s);
921 void ff_h263_encode_mba(MpegEncContext *s);
922
923 int intel_h263_decode_picture_header(MpegEncContext *s);
924 int flv_h263_decode_picture_header(MpegEncContext *s);
925 int ff_h263_decode_mb(MpegEncContext *s,
926 DCTELEM block[6][64]);
927 int ff_mpeg4_decode_mb(MpegEncContext *s,
928 DCTELEM block[6][64]);
929 int h263_get_picture_format(int width, int height);
930 void ff_mpeg4_encode_video_packet_header(MpegEncContext *s);
931 void ff_mpeg4_clean_buffers(MpegEncContext *s);
932 void ff_mpeg4_stuffing(PutBitContext * pbc);
933 void ff_mpeg4_init_partitions(MpegEncContext *s);
934 void ff_mpeg4_merge_partitions(MpegEncContext *s);
935 void ff_clean_mpeg4_qscales(MpegEncContext *s);
936 void ff_clean_h263_qscales(MpegEncContext *s);
937 int ff_mpeg4_decode_partitions(MpegEncContext *s);
938 int ff_mpeg4_get_video_packet_prefix_length(MpegEncContext *s);
939 int ff_h263_resync(MpegEncContext *s);
940 int ff_h263_get_gob_height(MpegEncContext *s);
941 int ff_mpeg4_set_direct_mv(MpegEncContext *s, int mx, int my);
942 int ff_h263_round_chroma(int x);
943 void ff_h263_encode_motion(MpegEncContext * s, int val, int f_code);
944 int ff_mpeg4_find_frame_end(ParseContext *pc, const uint8_t *buf, int buf_size);
945
946
947 /* rv10.c */
948 void rv10_encode_picture_header(MpegEncContext *s, int picture_number);
949 int rv_decode_dc(MpegEncContext *s, int n);
950 void rv20_encode_picture_header(MpegEncContext *s, int picture_number);
951
952
953 /* msmpeg4.c */
954 void msmpeg4_encode_picture_header(MpegEncContext * s, int picture_number);
955 void msmpeg4_encode_ext_header(MpegEncContext * s);
956 void msmpeg4_encode_mb(MpegEncContext * s,
957 DCTELEM block[6][64],
958 int motion_x, int motion_y);
959 int msmpeg4_decode_picture_header(MpegEncContext * s);
960 int msmpeg4_decode_ext_header(MpegEncContext * s, int buf_size);
961 int ff_msmpeg4_decode_init(MpegEncContext *s);
962 void ff_msmpeg4_encode_init(MpegEncContext *s);
963 int ff_wmv2_decode_picture_header(MpegEncContext * s);
964 int ff_wmv2_decode_secondary_picture_header(MpegEncContext * s);
965 void ff_wmv2_add_mb(MpegEncContext *s, DCTELEM block[6][64], uint8_t *dest_y, uint8_t *dest_cb, uint8_t *dest_cr);
966 void ff_mspel_motion(MpegEncContext *s,
967 uint8_t *dest_y, uint8_t *dest_cb, uint8_t *dest_cr,
968 uint8_t **ref_picture, op_pixels_func (*pix_op)[4],
969 int motion_x, int motion_y, int h);
970 int ff_wmv2_encode_picture_header(MpegEncContext * s, int picture_number);
971 void ff_wmv2_encode_mb(MpegEncContext * s,
972 DCTELEM block[6][64],
973 int motion_x, int motion_y);
974
975 /* mjpeg.c */
976 int mjpeg_init(MpegEncContext *s);
977 void mjpeg_close(MpegEncContext *s);
978 void mjpeg_encode_mb(MpegEncContext *s,
979 DCTELEM block[6][64]);
980 void mjpeg_picture_header(MpegEncContext *s);
981 void mjpeg_picture_trailer(MpegEncContext *s);
982 void ff_mjpeg_stuffing(PutBitContext * pbc);
983
984
985 /* rate control */
986 int ff_rate_control_init(MpegEncContext *s);
987 float ff_rate_estimate_qscale(MpegEncContext *s, int dry_run);
988 void ff_write_pass1_stats(MpegEncContext *s);
989 void ff_rate_control_uninit(MpegEncContext *s);
990 double ff_eval(char *s, double *const_value, const char **const_name,
991 double (**func1)(void *, double), const char **func1_name,
992 double (**func2)(void *, double, double), char **func2_name,
993 void *opaque);
994 int ff_vbv_update(MpegEncContext *s, int frame_size);
995 void ff_get_2pass_fcode(MpegEncContext *s);
996
997 int ff_xvid_rate_control_init(MpegEncContext *s);
998 void ff_xvid_rate_control_uninit(MpegEncContext *s);
999 float ff_xvid_rate_estimate_qscale(MpegEncContext *s, int dry_run);
1000
1001 #endif /* AVCODEC_MPEGVIDEO_H */