2 * Generic DCT based hybrid video encoder
3 * Copyright (c) 2000, 2001, 2002 Fabrice Bellard.
5 * This library is free software; you can redistribute it and/or
6 * modify it under the terms of the GNU Lesser General Public
7 * License as published by the Free Software Foundation; either
8 * version 2 of the License, or (at your option) any later version.
10 * This library is distributed in the hope that it will be useful,
11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13 * Lesser General Public License for more details.
15 * You should have received a copy of the GNU Lesser General Public
16 * License along with this library; if not, write to the Free Software
17 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
20 #ifndef AVCODEC_MPEGVIDEO_H
21 #define AVCODEC_MPEGVIDEO_H
23 #define FRAME_SKIPED 100 // return value for header parsers if frame is not coded
31 #define MPEG_BUF_SIZE (16 * 1024)
33 #define QMAT_SHIFT_MMX 19
38 #define REORDER_BUFFER_SIZE (FF_MAX_B_FRAMES+2)
40 #define ME_MAP_SIZE 64
41 #define ME_MAP_SHIFT 3
42 #define ME_MAP_MV_BITS 11
44 /* run length table */
48 typedef struct Predictor{
54 typedef struct RateControlEntry{
66 typedef struct RateControlContext{
69 RateControlEntry *entry;
72 typedef struct ReorderBuffer{
78 int picture_in_gop_number;
81 typedef struct MpegEncContext {
82 struct AVCodecContext *avctx;
83 /* the following parameters must be initialized before encoding */
84 int width, height; /* picture size. must be a multiple of 16 */
86 int frame_rate; /* number of frames per second */
87 int intra_only; /* if true, only intra pictures are generated */
88 int bit_rate; /* wanted bit rate */
89 int bit_rate_tolerance; /* amount of +- bits (>0)*/
90 enum OutputFormat out_format; /* output format */
91 int h263_pred; /* use mpeg4/h263 ac/dc predictions */
93 /* the following codec id fields are deprecated in favor of codec_id */
94 int h263_plus; /* h263 plus headers */
95 int h263_rv10; /* use RV10 variation for H263 */
96 int h263_msmpeg4; /* generate MSMPEG4 compatible stream (deprecated, use msmpeg4_version instead)*/
97 int h263_intel; /* use I263 intel h263 header */
99 int codec_id; /* see CODEC_ID_xxx */
100 int fixed_qscale; /* fixed qscale if non zero */
101 float qcompress; /* amount of qscale change between easy & hard scenes (0.0-1.0) */
102 float qblur; /* amount of qscale smoothing over time (0.0-1.0) */
103 int qmin; /* min qscale */
104 int qmax; /* max qscale */
105 int max_qdiff; /* max qscale difference between frames */
106 int encoding; /* true if we are encoding (vs decoding) */
107 int flags; /* AVCodecContext.flags (HQ, MV4, ...) */
108 int force_input_type;/* 0= no force, otherwise I_TYPE, P_TYPE, ... */
109 int max_b_frames; /* max number of b-frames for encoding */
110 float b_quant_factor;/* qscale factor between ips and b frames */
111 float b_quant_offset;/* qscale offset between ips and b frames */
113 int b_frame_strategy;
114 int luma_elim_threshold;
115 int chroma_elim_threshold;
116 int strict_std_compliance; /* strictly follow the std (MPEG4, ...) */
117 int workaround_bugs; /* workaround bugs in encoders which cannot be detected automatically */
118 /* the following fields are managed internally by the encoder */
123 /* sequence parameters */
124 int context_initialized;
125 int input_picture_number;
126 int input_picture_in_gop_number; /* 0-> first pic in gop, ... */
128 int fake_picture_number; /* picture number at the bitstream frame rate */
129 int gop_picture_number; /* index of the first picture of a GOP based on fake_pic_num & mpeg1 specific */
130 int picture_in_gop_number; /* 0-> first pic in gop, ... */
131 int b_frames_since_non_b; /* used for encoding, relative to not yet reordered input */
132 int mb_width, mb_height; /* number of MBs horizontally & vertically */
133 int mb_num; /* number of MBs of a picture */
134 int linesize; /* line size, in bytes, may be different from width */
135 int uvlinesize; /* line size, for chroma in bytes, may be different from width */
136 UINT8 *new_picture[3]; /* picture to be compressed */
137 UINT8 *picture_buffer[REORDER_BUFFER_SIZE][3]; /* internal buffers used for reordering of input pictures */
138 int picture_buffer_index;
139 ReorderBuffer coded_order[REORDER_BUFFER_SIZE];
140 UINT8 *last_picture[3]; /* previous picture */
141 UINT8 *last_picture_base[3]; /* real start of the picture */
142 UINT8 *next_picture[3]; /* previous picture (for bidir pred) */
143 UINT8 *next_picture_base[3]; /* real start of the picture */
144 UINT8 *aux_picture[3]; /* aux picture (for B frames only) */
145 UINT8 *aux_picture_base[3]; /* real start of the picture */
146 UINT8 *current_picture[3]; /* buffer to store the decompressed current picture */
147 void *last_dr_opaque;
148 void *next_dr_opaque;
149 int ip_buffer_count; /* number of buffers, currently only >2 if dr1 is used */
150 int num_available_buffers; /* is 0 at the start & after seeking, after the first I frame its 1 after next I/P 2 */
151 int last_dc[3]; /* last DC values for MPEG1 */
152 INT16 *dc_val[3]; /* used for mpeg4 DC prediction, all 3 arrays must be continuous */
153 int y_dc_scale, c_dc_scale;
154 UINT8 *y_dc_scale_table; /* qscale -> y_dc_scale table */
155 UINT8 *c_dc_scale_table; /* qscale -> c_dc_scale table */
156 UINT8 *coded_block; /* used for coded block pattern prediction (msmpeg4v3, wmv1)*/
157 INT16 (*ac_val[3])[16]; /* used for for mpeg4 AC prediction, all 3 arrays must be continuous */
159 int mb_skiped; /* MUST BE SET only during DECODING */
160 UINT8 *mbskip_table; /* used to avoid copy if macroblock skipped (for black regions for example)
161 and used for b-frame encoding & decoding (contains skip table of next P Frame) */
162 UINT8 *mbintra_table; /* used to avoid setting {ac, dc, cbp}-pred stuff to zero on inter MB decoding */
163 UINT8 *cbp_table; /* used to store cbp, ac_pred for partitioned decoding */
164 UINT8 *pred_dir_table; /* used to store pred_dir for partitioned decoding */
165 INT8 *qscale_table; /* used to store qscale for partitioned decoding (& postprocessing FIXME export) */
166 UINT8 *edge_emu_buffer;
168 int input_qscale; /* qscale prior to reordering of frames */
169 int input_pict_type; /* pict_type prior to reordering of frames */
170 int force_type; /* 0= no force, otherwise I_TYPE, P_TYPE, ... */
172 int last_non_b_qscale; /* QP of last non b frame used for b frame qscale*/
173 int pict_type; /* I_TYPE, P_TYPE, B_TYPE, ... */
174 int last_non_b_pict_type; /* used for mpeg4 gmc b-frames & ratecontrol */
175 int frame_rate_index;
176 /* motion compensation */
178 int h263_long_vectors; /* use horrible h263v1 long vector mode */
180 int f_code; /* forward MV resolution */
181 int b_code; /* backward MV resolution for B Frames (mpeg4) */
182 INT16 (*motion_val)[2]; /* used for MV prediction (4MV per MB) */
183 INT16 (*p_mv_table)[2]; /* MV table (1MV per MB) p-frame encoding */
184 INT16 (*b_forw_mv_table)[2]; /* MV table (1MV per MB) forward mode b-frame encoding */
185 INT16 (*b_back_mv_table)[2]; /* MV table (1MV per MB) backward mode b-frame encoding */
186 INT16 (*b_bidir_forw_mv_table)[2]; /* MV table (1MV per MB) bidir mode b-frame encoding */
187 INT16 (*b_bidir_back_mv_table)[2]; /* MV table (1MV per MB) bidir mode b-frame encoding */
188 INT16 (*b_direct_forw_mv_table)[2];/* MV table (1MV per MB) direct mode b-frame encoding */
189 INT16 (*b_direct_back_mv_table)[2];/* MV table (1MV per MB) direct mode b-frame encoding */
190 INT16 (*b_direct_mv_table)[2]; /* MV table (1MV per MB) direct mode b-frame encoding */
191 int me_method; /* ME algorithm */
192 uint8_t *me_scratchpad; /* data area for the me algo, so that the ME doesnt need to malloc/free */
193 uint32_t *me_map; /* map to avoid duplicate evaluations */
194 uint16_t *me_score_map; /* map to store the SADs */
195 int me_map_generation;
196 int skip_me; /* set if ME is skiped for the current MB */
198 #define MV_DIR_BACKWARD 1
199 #define MV_DIR_FORWARD 2
200 #define MV_DIRECT 4 // bidirectional mode where the difference equals the MV of the last P/S/I-Frame (mpeg4)
202 #define MV_TYPE_16X16 0 /* 1 vector for the whole mb */
203 #define MV_TYPE_8X8 1 /* 4 vectors (h263, mpeg4 4MV) */
204 #define MV_TYPE_16X8 2 /* 2 vectors, one per 16x8 block */
205 #define MV_TYPE_FIELD 3 /* 2 vectors, one per field */
206 #define MV_TYPE_DMV 4 /* 2 vectors, special mpeg2 Dual Prime Vectors */
207 /* motion vectors for a macroblock
208 first coordinate : 0 = forward 1 = backward
209 second " : depend on type
210 third " : 0 = x, 1 = y
213 int field_select[2][2];
214 int last_mv[2][2][2]; /* last MV, used for MV prediction in MPEG1 & B-frame MPEG4 */
215 UINT16 (*mv_penalty)[MAX_MV*2+1]; /* amount of bits needed to encode a MV, used for ME */
216 UINT8 *fcode_tab; /* smallest fcode needed for each MV */
219 int no_rounding; /* apply no rounding to motion compensation (MPEG4, msmpeg4, ...)
220 for b-frames rounding mode is allways 0 */
222 int hurry_up; /* when set to 1 during decoding, b frames will be skiped
223 when set to 2 idct/dequant will be skipped too */
225 /* macroblock layer */
229 UINT16 *mb_var; /* Table for MB variances */
230 UINT16 *mc_mb_var; /* Table for motion compensated MB variances */
231 UINT8 *mb_type; /* Table for MB type */
232 #define MB_TYPE_INTRA 0x01
233 #define MB_TYPE_INTER 0x02
234 #define MB_TYPE_INTER4V 0x04
235 #define MB_TYPE_SKIPED 0x08
236 #define MB_TYPE_GMC 0x10
238 #define MB_TYPE_DIRECT 0x10
239 #define MB_TYPE_FORWARD 0x20
240 #define MB_TYPE_BACKWARD 0x40
241 #define MB_TYPE_BIDIR 0x80
243 int block_index[6]; /* index to current MB in block based arrays with edges*/
246 /* matrix transmitted in the bitstream */
247 UINT16 intra_matrix[64];
248 UINT16 chroma_intra_matrix[64];
249 UINT16 inter_matrix[64];
250 UINT16 chroma_inter_matrix[64];
251 #define QUANT_BIAS_SHIFT 4
252 int intra_quant_bias; /* bias for the quantizer */
253 int inter_quant_bias; /* bias for the quantizer */
254 int min_qcoeff; /* minimum encodable coefficient */
255 int max_qcoeff; /* maximum encodable coefficient */
256 /* precomputed matrix (combine qscale and DCT renorm) */
257 int q_intra_matrix[32][64];
258 int q_inter_matrix[32][64];
259 /* identical to the above but for MMX & these are not permutated */
260 UINT16 __align8 q_intra_matrix16[32][64];
261 UINT16 __align8 q_inter_matrix16[32][64];
262 UINT16 __align8 q_intra_matrix16_bias[32][64];
263 UINT16 __align8 q_inter_matrix16_bias[32][64];
264 int block_last_index[6]; /* last non zero coefficient in block */
266 void *opaque; /* private data for the user */
268 /* bit rate control */
269 int I_frame_bits; //FIXME used in mpeg12 ...
270 int mb_var_sum; /* sum of MB variance for current frame */
271 int mc_mb_var_sum; /* motion compensated MB variance for current frame */
272 int last_non_b_mc_mb_var;/* motion compensated MB variance for last non b frame */
275 int frame_bits; /* bits used for the current frame */
276 int pb_frame_bits; /* bits of the last b...bp group */
279 double qsum; /* sum of qscales */
280 double qcount; /* count of qscales */
281 double short_term_qsum; /* sum of recent qscales */
282 double short_term_qcount; /* count of recent qscales */
283 RateControlContext rc_context;
285 /* statistics, used for 2-pass encoding */
293 int misc_bits; // cbp, mb_type
294 int last_bits; //temp var used for calculating the above vars
296 /* error concealment / resync */
297 int resync_mb_x; /* x position of last resync marker */
298 int resync_mb_y; /* y position of last resync marker */
299 int mb_num_left; /* number of MBs left in this video packet */
300 GetBitContext next_resync_gb; /* starts at the next resync marker */
301 int next_resync_qscale; /* qscale of next resync marker */
302 int next_resync_pos; /* bitstream position of next resync marker */
303 #define DECODING_AC_LOST -1
304 #define DECODING_ACDC_LOST -2
305 #define DECODING_DESYNC -3
307 int next_p_frame_damaged; /* set if the next p frame is damaged, to avoid showing trashed b frames */
308 int error_resilience;
314 /* H.263+ specific */
317 int h263_aic; /* Advanded INTRA Coding (AIC) */
318 int h263_aic_dir; /* AIC direction: 0 = left, 1 = top */
321 int time_increment_resolution;
322 int time_increment_bits; /* number of bits to represent the fractional part of time */
324 int time_base; /* time in seconds of last I,P,S Frame */
325 INT64 time; /* time of current frame */
326 INT64 last_non_b_time;
327 UINT16 pp_time; /* time distance between the last 2 p,s,i frames */
328 UINT16 bp_time; /* time distance between the last b and p,s,i frame */
330 int vol_sprite_usage;
335 int sprite_brightness_change;
336 int num_sprite_warping_points;
337 int real_sprite_warping_points;
338 int sprite_offset[2][2];
339 int sprite_delta[2][2][2];
340 int sprite_shift[2][2];
343 int quarter_sample; /* 1->qpel, 0->half pel ME/MC */
346 int enhancement_type;
349 int aspect_ratio_info;
350 int sprite_warping_accuracy;
351 int low_latency_sprite;
352 int data_partitioning;
353 int rvlc; /* reversible vlc */
354 int resync_marker; /* could this stream contain resync markers*/
355 int low_delay; /* no reordering needed / has no b-frames */
357 int vol_control_parameters; /* does the stream contain the low_delay flag, used to workaround buggy encoders */
358 PutBitContext tex_pb; /* used for data partitioned VOPs */
359 PutBitContext pb2; /* used for data partitioned VOPs */
360 #define PB_BUFFER_SIZE 1024*256
361 uint8_t *tex_pb_buffer;
364 /* divx specific, used to workaround (many) bugs in divx5 */
367 #define BITSTREAM_BUFFER_SIZE 1024*256
368 UINT8 *bitstream_buffer; //Divx 5.01 puts several frames in a single one, this is used to reorder them
369 int bitstream_buffer_size;
372 int rv10_version; /* RV10 version: 0 or 3 */
373 int rv10_first_dc_coded[3];
376 struct MJpegContext *mjpeg_ctx;
377 int mjpeg_vsample[3]; /* vertical sampling factors, default = {2, 1, 1} */
378 int mjpeg_hsample[3]; /* horizontal sampling factors, default = {2, 1, 1} */
379 int mjpeg_write_tables; /* do we want to have quantisation- and
380 huffmantables in the jpeg file ? */
381 int mjpeg_data_only_frames; /* frames only with SOI, SOS and EOI markers */
383 /* MSMPEG4 specific */
386 int rl_chroma_table_index;
388 int use_skip_mb_code;
389 int slice_height; /* in macroblocks */
390 int first_slice_line; /* used in mpeg4 too to handle resync markers */
391 int flipflop_rounding;
392 int msmpeg4_version; /* 0=not msmpeg4, 1=mp41, 2=mp42, 3=mp43/divx3 4=wmv1/7 5=wmv2/8*/
394 int esc3_level_length;
396 UINT8 *inter_scantable;
397 UINT8 *intra_scantable;
398 UINT8 *intra_v_scantable;
399 UINT8 *intra_h_scantable;
400 /* [mb_intra][isChroma][level][run][last] */
401 int ac_stats[2][2][MAX_LEVEL+1][MAX_RUN+1][2];
402 int inter_intra_pred;
405 /* decompression specific */
408 /* MPEG2 specific - I wish I had not to support this mess. */
409 int progressive_sequence;
410 int mpeg_f_code[2][2];
411 int picture_structure;
413 #define PICT_TOP_FIELD 1
414 #define PICT_BOTTOM_FIELD 2
417 int intra_dc_precision;
418 int frame_pred_frame_dct;
420 int concealment_motion_vectors;
422 int intra_vlc_format;
424 int repeat_first_field;
426 int progressive_frame;
434 /* These are explained on avcodec.h */
436 int rtp_payload_size;
437 void (*rtp_callback)(void *data, int size, int packet_number);
439 UINT8 *ptr_last_mb_line;
440 UINT32 mb_line_avgsize;
442 DCTELEM (*block)[64]; /* points to one of the following blocks */
443 DCTELEM blocks[2][6][64] __align8; // for HQ mode we need to keep the best block
444 void (*dct_unquantize_mpeg1)(struct MpegEncContext *s,
445 DCTELEM *block, int n, int qscale);
446 void (*dct_unquantize_mpeg2)(struct MpegEncContext *s,
447 DCTELEM *block, int n, int qscale);
448 void (*dct_unquantize_h263)(struct MpegEncContext *s,
449 DCTELEM *block, int n, int qscale);
450 void (*dct_unquantize)(struct MpegEncContext *s, // unquantizer to use (mpeg4 can use both)
451 DCTELEM *block, int n, int qscale);
454 int MPV_common_init(MpegEncContext *s);
455 void MPV_common_end(MpegEncContext *s);
456 void MPV_decode_mb(MpegEncContext *s, DCTELEM block[6][64]);
457 void MPV_frame_start(MpegEncContext *s, AVCodecContext *avctx);
458 void MPV_frame_end(MpegEncContext *s);
460 void MPV_common_init_mmx(MpegEncContext *s);
463 void MPV_common_init_axp(MpegEncContext *s);
465 extern int (*dct_quantize)(MpegEncContext *s, DCTELEM *block, int n, int qscale, int *overflow);
466 extern void (*draw_edges)(UINT8 *buf, int wrap, int width, int height, int w);
467 void ff_conceal_past_errors(MpegEncContext *s, int conceal_all);
468 void ff_copy_bits(PutBitContext *pb, UINT8 *src, int length);
469 void ff_clean_intra_table_entries(MpegEncContext *s);
472 void ff_estimate_p_frame_motion(MpegEncContext * s,
474 void ff_estimate_b_frame_motion(MpegEncContext * s,
476 int ff_get_best_fcode(MpegEncContext * s, int16_t (*mv_table)[2], int type);
477 void ff_fix_long_p_mvs(MpegEncContext * s);
478 void ff_fix_long_b_mvs(MpegEncContext * s, int16_t (*mv_table)[2], int f_code, int type);
481 extern INT16 ff_mpeg1_default_intra_matrix[64];
482 extern INT16 ff_mpeg1_default_non_intra_matrix[64];
483 extern UINT8 ff_mpeg1_dc_scale_table[128];
485 void mpeg1_encode_picture_header(MpegEncContext *s, int picture_number);
486 void mpeg1_encode_mb(MpegEncContext *s,
487 DCTELEM block[6][64],
488 int motion_x, int motion_y);
489 void ff_mpeg1_encode_init(MpegEncContext *s);
492 typedef struct RLTable {
493 int n; /* number of entries of table_vlc minus 1 */
494 int last; /* number of values for last = 0 */
495 const UINT16 (*table_vlc)[2];
496 const INT8 *table_run;
497 const INT8 *table_level;
498 UINT8 *index_run[2]; /* encoding only */
499 INT8 *max_level[2]; /* encoding & decoding */
500 INT8 *max_run[2]; /* encoding & decoding */
501 VLC vlc; /* decoding only deprected FIXME remove*/
502 RL_VLC_ELEM *rl_vlc[32]; /* decoding only */
505 void init_rl(RLTable *rl);
506 void init_vlc_rl(RLTable *rl);
508 static inline int get_rl_index(const RLTable *rl, int last, int run, int level)
511 index = rl->index_run[last][run];
514 if (level > rl->max_level[last][run])
516 return index + level - 1;
519 extern UINT8 ff_mpeg4_y_dc_scale_table[32];
520 extern UINT8 ff_mpeg4_c_dc_scale_table[32];
521 void h263_encode_mb(MpegEncContext *s,
522 DCTELEM block[6][64],
523 int motion_x, int motion_y);
524 void mpeg4_encode_mb(MpegEncContext *s,
525 DCTELEM block[6][64],
526 int motion_x, int motion_y);
527 void h263_encode_picture_header(MpegEncContext *s, int picture_number);
528 int h263_encode_gob_header(MpegEncContext * s, int mb_line);
529 INT16 *h263_pred_motion(MpegEncContext * s, int block,
531 void mpeg4_pred_ac(MpegEncContext * s, INT16 *block, int n,
533 void ff_set_mpeg4_time(MpegEncContext * s, int picture_number);
534 void mpeg4_encode_picture_header(MpegEncContext *s, int picture_number);
535 void h263_encode_init(MpegEncContext *s);
537 void h263_decode_init_vlc(MpegEncContext *s);
538 int h263_decode_picture_header(MpegEncContext *s);
539 int h263_decode_gob_header(MpegEncContext *s);
540 int mpeg4_decode_picture_header(MpegEncContext * s);
541 int intel_h263_decode_picture_header(MpegEncContext *s);
542 int h263_decode_mb(MpegEncContext *s,
543 DCTELEM block[6][64]);
544 int h263_get_picture_format(int width, int height);
545 int ff_mpeg4_decode_video_packet_header(MpegEncContext *s);
546 int ff_mpeg4_resync(MpegEncContext *s);
547 void ff_mpeg4_encode_video_packet_header(MpegEncContext *s);
548 void ff_mpeg4_clean_buffers(MpegEncContext *s);
549 void ff_mpeg4_stuffing(PutBitContext * pbc);
550 void ff_mpeg4_init_partitions(MpegEncContext *s);
551 void ff_mpeg4_merge_partitions(MpegEncContext *s);
552 extern inline int ff_mpeg4_pred_dc(MpegEncContext * s, int n, UINT16 **dc_val_ptr, int *dir_ptr);
555 void rv10_encode_picture_header(MpegEncContext *s, int picture_number);
556 int rv_decode_dc(MpegEncContext *s, int n);
559 void msmpeg4_encode_picture_header(MpegEncContext * s, int picture_number);
560 void msmpeg4_encode_ext_header(MpegEncContext * s);
561 void msmpeg4_encode_mb(MpegEncContext * s,
562 DCTELEM block[6][64],
563 int motion_x, int motion_y);
564 int msmpeg4_decode_picture_header(MpegEncContext * s);
565 int msmpeg4_decode_ext_header(MpegEncContext * s, int buf_size);
566 int msmpeg4_decode_mb(MpegEncContext *s,
567 DCTELEM block[6][64]);
568 int ff_msmpeg4_decode_init(MpegEncContext *s);
569 void ff_msmpeg4_encode_init(MpegEncContext *s);
573 int mjpeg_init(MpegEncContext *s);
574 void mjpeg_close(MpegEncContext *s);
575 void mjpeg_encode_mb(MpegEncContext *s,
576 DCTELEM block[6][64]);
577 void mjpeg_picture_header(MpegEncContext *s);
578 void mjpeg_picture_trailer(MpegEncContext *s);
581 int ff_rate_control_init(MpegEncContext *s);
582 int ff_rate_estimate_qscale(MpegEncContext *s);
583 int ff_rate_estimate_qscale_pass2(MpegEncContext *s);
584 void ff_write_pass1_stats(MpegEncContext *s);
585 void ff_rate_control_uninit(MpegEncContext *s);
587 #endif /* AVCODEC_MPEGVIDEO_H */