b_quant_offset (unfinished, should be 0 for now)
[libav.git] / libavcodec / avcodec.h
CommitLineData
a96b68b7
FB
1#ifndef AVCODEC_H
2#define AVCODEC_H
3
de6d9b64
FB
4#include "common.h"
5
4b1f4f23
J
6#define LIBAVCODEC_VERSION_INT 0x000406
7#define LIBAVCODEC_VERSION "0.4.6"
92ba5ffb
MN
8#define LIBAVCODEC_BUILD 4610
9#define LIBAVCODEC_BUILD_STR "4610"
4b1f4f23 10
de6d9b64
FB
11enum CodecID {
12 CODEC_ID_NONE,
13 CODEC_ID_MPEG1VIDEO,
14 CODEC_ID_H263,
15 CODEC_ID_RV10,
16 CODEC_ID_MP2,
9d36bdc9 17 CODEC_ID_MP3LAME,
de6d9b64
FB
18 CODEC_ID_AC3,
19 CODEC_ID_MJPEG,
bf89e6b1 20 CODEC_ID_MPEG4,
de6d9b64 21 CODEC_ID_RAWVIDEO,
84afee34
MN
22 CODEC_ID_MSMPEG4V1,
23 CODEC_ID_MSMPEG4V2,
24 CODEC_ID_MSMPEG4V3,
e1a9dbff 25 CODEC_ID_WMV1,
de6d9b64
FB
26 CODEC_ID_H263P,
27 CODEC_ID_H263I,
a96b68b7
FB
28
29 /* various pcm "codecs" */
30 CODEC_ID_PCM_S16LE,
31 CODEC_ID_PCM_S16BE,
32 CODEC_ID_PCM_U16LE,
33 CODEC_ID_PCM_U16BE,
34 CODEC_ID_PCM_S8,
35 CODEC_ID_PCM_U8,
36 CODEC_ID_PCM_MULAW,
37 CODEC_ID_PCM_ALAW,
de6d9b64 38};
84afee34 39#define CODEC_ID_MSMPEG4 CODEC_ID_MSMPEG4V3
de6d9b64
FB
40
41enum CodecType {
4a663d80 42 CODEC_TYPE_UNKNOWN = -1,
de6d9b64
FB
43 CODEC_TYPE_VIDEO,
44 CODEC_TYPE_AUDIO,
45};
46
47enum PixelFormat {
9f862d11 48 PIX_FMT_ANY = -1,
de6d9b64
FB
49 PIX_FMT_YUV420P,
50 PIX_FMT_YUV422,
51 PIX_FMT_RGB24,
52 PIX_FMT_BGR24,
d9fea0b5
FB
53 PIX_FMT_YUV422P,
54 PIX_FMT_YUV444P,
de6d9b64
FB
55};
56
a96b68b7
FB
57/* currently unused, may be used if 24/32 bits samples ever supported */
58enum SampleFormat {
59 SAMPLE_FMT_S16 = 0, /* signed 16 bits */
60};
61
de6d9b64
FB
62/* in bytes */
63#define AVCODEC_MAX_AUDIO_FRAME_SIZE 18432
64
e4986da9
J
65/* motion estimation type, EPZS by default */
66enum Motion_Est_ID {
101bea5f 67 ME_ZERO = 1,
e4986da9
J
68 ME_FULL,
69 ME_LOG,
70 ME_PHODS,
71 ME_EPZS,
72 ME_X1
73};
74
101bea5f
J
75/* only for ME compatiblity with old apps */
76extern int motion_estimation_method;
77
e4986da9 78/* ME algos sorted by quality */
bc6caae2
J
79static const int Motion_Est_QTab[] = { ME_ZERO, ME_PHODS, ME_LOG,
80 ME_X1, ME_EPZS, ME_FULL };
de6d9b64 81
9dbcbd92
MN
82#define FF_MAX_B_FRAMES 4
83
de6d9b64 84/* encoding support */
ba6802de 85/* note not everything is supported yet */
de6d9b64
FB
86
87#define CODEC_FLAG_HQ 0x0001 /* high quality (non real time) encoding */
88#define CODEC_FLAG_QSCALE 0x0002 /* use fixed qscale */
ba6802de 89#define CODEC_FLAG_4MV 0x0004 /* 4 MV per MB allowed */
ba6802de
MN
90#define CODEC_FLAG_QPEL 0x0010 /* use qpel MC */
91#define CODEC_FLAG_GMC 0x0020 /* use GMC */
60e575ac 92#define CODEC_FLAG_TYPE 0x0040 /* fixed I/P frame type, from avctx->key_frame */
9dbcbd92
MN
93/* parent program gurantees that the input for b-frame containing streams is not written to
94 for at least s->max_b_frames+1 frames, if this is not set than the input will be copied */
8b4c7dbc
MN
95#define CODEC_FLAG_INPUT_PRESERVED 0x0100
96#define CODEC_FLAG_PASS1 0x0200 /* use internal 2pass ratecontrol in first pass mode */
97#define CODEC_FLAG_PASS2 0x0400 /* use internal 2pass ratecontrol in second pass mode */
e84c31dc 98#define CODEC_FLAG_EXTERN_HUFF 0x1000 /* use external huffman table (for mjpeg) */
de6d9b64 99
bf89e6b1
FB
100/* codec capabilities */
101
102/* decoder can use draw_horiz_band callback */
103#define CODEC_CAP_DRAW_HORIZ_BAND 0x0001
104
de6d9b64
FB
105#define FRAME_RATE_BASE 10000
106
107typedef struct AVCodecContext {
108 int bit_rate;
9cdd6a24 109 int bit_rate_tolerance; /* amount of +- bits (>0)*/
de6d9b64
FB
110 int flags;
111 int sub_id; /* some codecs needs additionnal format info. It is
112 stored there */
e4986da9
J
113
114 int me_method; /* ME algorithm used for video coding */
115
e84c31dc
A
116 /* extra data from parent application to codec, e.g. huffman table
117 for mjpeg */
118 /* the parent should allocate and free this buffer */
119 void *extradata;
120 int extradata_size;
121
de6d9b64
FB
122 /* video only */
123 int frame_rate; /* frames per sec multiplied by FRAME_RATE_BASE */
124 int width, height;
11ce8834 125 int aspect_ratio_info;
2efe097a
MN
126#define FF_ASPECT_SQUARE 1
127#define FF_ASPECT_4_3_625 2
128#define FF_ASPECT_4_3_525 3
129#define FF_ASPECT_16_9_625 4
130#define FF_ASPECT_16_9_525 5
de6d9b64 131 int gop_size; /* 0 = intra only */
bc657ac3 132 enum PixelFormat pix_fmt; /* pixel format, see PIX_FMT_xxx */
e0a3d744
J
133 int repeat_pict; /* when decoding, this signal how much the picture */
134 /* must be delayed. */
135 /* extra_delay = (repeat_pict / 2) * (1/fps) */
bf89e6b1
FB
136 /* if non NULL, 'draw_horiz_band' is called by the libavcodec
137 decoder to draw an horizontal band. It improve cache usage. Not
138 all codecs can do that. You must check the codec capabilities
139 before */
140 void (*draw_horiz_band)(struct AVCodecContext *s,
141 UINT8 **src_ptr, int linesize,
142 int y, int width, int height);
143
de6d9b64
FB
144 /* audio only */
145 int sample_rate; /* samples per sec */
146 int channels;
a96b68b7 147 int sample_fmt; /* sample format, currenly unused */
de6d9b64
FB
148
149 /* the following data should not be initialized */
9cd3766f
J
150 int frame_size; /* in samples, initialized when calling 'init' */
151 int frame_number; /* audio or video frame number */
152 int real_pict_num; /* returns the real picture number of
153 previous encoded frame */
154 int key_frame; /* true if the previous compressed frame was
155 a key frame (intra, or seekable) */
156 int pict_type; /* picture type of the previous
157 encoded frame */
158/* FIXME: these should have FF_ */
159#define I_TYPE 1 // Intra
160#define P_TYPE 2 // Predicted
161#define B_TYPE 3 // Bi-dir predicted
162#define S_TYPE 4 // S(GMC)-VOP MPEG4
163
164 int delay; /* number of frames the decoded output
165 will be delayed relative to the encoded input */
2417652e
MN
166 uint8_t *mbskip_table; /* =1 if MB didnt change, is only valid for I/P frames
167 stride= mb_width = (width+15)>>4 */
1ff662cc
MN
168
169 /* encoding parameters */
de6d9b64 170 int quality; /* quality of the previous encoded frame
8b4c7dbc
MN
171 (between 1 (good) and 31 (bad))
172 this is allso used to set the quality in vbr mode
173 and the per frame quality in CODEC_FLAG_TYPE (second pass mode) */
9cdd6a24
MN
174 float qcompress; /* amount of qscale change between easy & hard scenes (0.0-1.0)*/
175 float qblur; /* amount of qscale smoothing over time (0.0-1.0) */
176 int qmin; /* min qscale */
177 int qmax; /* max qscale */
178 int max_qdiff; /* max qscale difference between frames */
9dbcbd92 179 int max_b_frames; /* maximum b frames, the output will be delayed by max_b_frames+1 relative to the input */
8b4c7dbc
MN
180 float b_quant_factor;/* qscale factor between ips and b frames */
181 int rc_strategy;
182 int b_frame_strategy;
9dbcbd92 183
2417652e
MN
184 int hurry_up; /* when set to 1 during decoding, b frames will be skiped
185 when set to 2 idct/dequant will be skipped too */
186
de6d9b64
FB
187 struct AVCodec *codec;
188 void *priv_data;
189
644d98a4
J
190 /* The following data is for RTP friendly coding */
191 /* By now only H.263/H.263+ coder honours this */
192 int rtp_mode; /* 1 for activate RTP friendly-mode */
193 /* highers numbers represent more error-prone */
194 /* enviroments, by now just "1" exist */
195
196 int rtp_payload_size; /* The size of the RTP payload, the coder will */
197 /* do it's best to deliver a chunk with size */
198 /* below rtp_payload_size, the chunk will start */
199 /* with a start code on some codecs like H.263 */
200 /* This doesn't take account of any particular */
201 /* headers inside the transmited RTP payload */
81401c1f
J
202
203
204 /* The RTP callcack: This function is called */
205 /* every time the encoder as a packet to send */
206 /* Depends on the encoder if the data starts */
207 /* with a Start Code (it should) H.263 does */
208 void (*rtp_callback)(void *data, int size, int packet_number);
209
43f1708f
J
210 /* These are for PSNR calculation, if you set get_psnr to 1 */
211 /* after encoding you will have the PSNR on psnr_y/cb/cr */
212 int get_psnr;
213 float psnr_y;
214 float psnr_cb;
215 float psnr_cr;
098eefe1
MN
216
217 /* statistics, used for 2-pass encoding */
218 int mv_bits;
219 int header_bits;
220 int i_tex_bits;
221 int p_tex_bits;
222 int i_count;
223 int p_count;
224 int skip_count;
225 int misc_bits; // cbp, mb_type
226 int frame_bits;
644d98a4 227
de6d9b64 228 /* the following fields are ignored */
bf89e6b1 229 void *opaque; /* can be used to carry app specific stuff */
de6d9b64 230 char codec_name[32];
4a663d80
PG
231 enum CodecType codec_type; /* see CODEC_TYPE_xxx */
232 enum CodecID codec_id; /* see CODEC_ID_xxx */
de6d9b64 233 unsigned int codec_tag; /* codec tag, only used if unknown codec */
92ba5ffb
MN
234
235 int workaround_bugs; /* workaround bugs in encoders which cannot be detected automatically */
5d27890a
NK
236 /*
237 Note: Below are located reserved fields for further usage
238 It requires for ABI !!!
239 If you'll perform some changes then borrow new space from these fields
240 (void * can be safety replaced with struct * ;)
241 P L E A S E ! ! !
242 IMPORTANT: Never change order of already declared fields!!!
243 */
244 unsigned long long int
245 ull_res0,ull_res1,ull_res2,ull_res3,ull_res4,ull_res5,
246 ull_res6,ull_res7,ull_res8,ull_res9,ull_res10,ull_res11,ull_res12;
247 float
248 flt_res0,flt_res1,flt_res2,flt_res3,flt_res4,flt_res5,
249 flt_res6,flt_res7,flt_res8,flt_res9,flt_res10,flt_res11,flt_res12;
250 void
251 *ptr_res0,*ptr_res1,*ptr_res2,*ptr_res3,*ptr_res4,*ptr_res5,
252 *ptr_res6,*ptr_res7,*ptr_res8,*ptr_res9,*ptr_res10,*ptr_res11,*ptr_res12;
253 unsigned long int
254 ul_res0,ul_res1,ul_res2,ul_res3,ul_res4,ul_res5,
255 ul_res6,ul_res7,ul_res8,ul_res9,ul_res10,ul_res11,ul_res12;
256 unsigned int
257 ui_res0,ui_res1,ui_res2,ui_res3,ui_res4,ui_res5,
92ba5ffb 258 ui_res6,ui_res7,ui_res8,ui_res9,ui_res10,ui_res11;
5d27890a
NK
259 unsigned short int
260 us_res0,us_res1,us_res2,us_res3,us_res4,us_res5,
261 us_res6,us_res7,us_res8,us_res9,us_res10,us_res11,us_res12;
262 unsigned char
263 uc_res0,uc_res1,uc_res2,uc_res3,uc_res4,uc_res5,
264 uc_res6,uc_res7,uc_res8,uc_res9,uc_res10,uc_res11,uc_res12;
de6d9b64
FB
265} AVCodecContext;
266
267typedef struct AVCodec {
268 char *name;
269 int type;
270 int id;
271 int priv_data_size;
272 int (*init)(AVCodecContext *);
273 int (*encode)(AVCodecContext *, UINT8 *buf, int buf_size, void *data);
274 int (*close)(AVCodecContext *);
275 int (*decode)(AVCodecContext *, void *outdata, int *outdata_size,
276 UINT8 *buf, int buf_size);
bf89e6b1 277 int capabilities;
de6d9b64 278 struct AVCodec *next;
5d27890a
NK
279 /*
280 Note: Below are located reserved fields for further usage
281 It requires for ABI !!!
282 If you'll perform some changes then borrow new space from these fields
283 (void * can be safety replaced with struct * ;)
284 P L E A S E ! ! !
285 IMPORTANT: Never change order of already declared fields!!!
286 */
287 unsigned long long int
288 ull_res0,ull_res1,ull_res2,ull_res3,ull_res4,ull_res5,
289 ull_res6,ull_res7,ull_res8,ull_res9,ull_res10,ull_res11,ull_res12;
290 float
291 flt_res0,flt_res1,flt_res2,flt_res3,flt_res4,flt_res5,
292 flt_res6,flt_res7,flt_res8,flt_res9,flt_res10,flt_res11,flt_res12;
293 void
294 *ptr_res0,*ptr_res1,*ptr_res2,*ptr_res3,*ptr_res4,*ptr_res5,
295 *ptr_res6,*ptr_res7,*ptr_res8,*ptr_res9,*ptr_res10,*ptr_res11,*ptr_res12;
de6d9b64
FB
296} AVCodec;
297
298/* three components are given, that's all */
299typedef struct AVPicture {
300 UINT8 *data[3];
301 int linesize[3];
302} AVPicture;
303
304extern AVCodec ac3_encoder;
305extern AVCodec mp2_encoder;
9d36bdc9 306extern AVCodec mp3lame_encoder;
de6d9b64
FB
307extern AVCodec mpeg1video_encoder;
308extern AVCodec h263_encoder;
309extern AVCodec h263p_encoder;
310extern AVCodec rv10_encoder;
311extern AVCodec mjpeg_encoder;
bf89e6b1 312extern AVCodec mpeg4_encoder;
84afee34
MN
313extern AVCodec msmpeg4v1_encoder;
314extern AVCodec msmpeg4v2_encoder;
315extern AVCodec msmpeg4v3_encoder;
de6d9b64
FB
316
317extern AVCodec h263_decoder;
bf89e6b1 318extern AVCodec mpeg4_decoder;
84afee34
MN
319extern AVCodec msmpeg4v1_decoder;
320extern AVCodec msmpeg4v2_decoder;
321extern AVCodec msmpeg4v3_decoder;
e1a9dbff 322extern AVCodec wmv1_decoder;
de6d9b64
FB
323extern AVCodec mpeg_decoder;
324extern AVCodec h263i_decoder;
325extern AVCodec rv10_decoder;
4e66ab3b 326extern AVCodec mjpeg_decoder;
4b1f4f23 327extern AVCodec mp2_decoder;
a96b68b7 328extern AVCodec mp3_decoder;
de6d9b64 329
a96b68b7
FB
330/* pcm codecs */
331#define PCM_CODEC(id, name) \
332extern AVCodec name ## _decoder; \
333extern AVCodec name ## _encoder;
334
335PCM_CODEC(CODEC_ID_PCM_S16LE, pcm_s16le);
336PCM_CODEC(CODEC_ID_PCM_S16BE, pcm_s16be);
337PCM_CODEC(CODEC_ID_PCM_U16LE, pcm_u16le);
338PCM_CODEC(CODEC_ID_PCM_U16BE, pcm_u16be);
339PCM_CODEC(CODEC_ID_PCM_S8, pcm_s8);
340PCM_CODEC(CODEC_ID_PCM_U8, pcm_u8);
341PCM_CODEC(CODEC_ID_PCM_ALAW, pcm_alaw);
342PCM_CODEC(CODEC_ID_PCM_MULAW, pcm_mulaw);
343
344#undef PCM_CODEC
345
346/* dummy raw video codec */
de6d9b64
FB
347extern AVCodec rawvideo_codec;
348
349/* the following codecs use external GPL libs */
de6d9b64
FB
350extern AVCodec ac3_decoder;
351
352/* resample.c */
353
354struct ReSampleContext;
355
356typedef struct ReSampleContext ReSampleContext;
357
358ReSampleContext *audio_resample_init(int output_channels, int input_channels,
359 int output_rate, int input_rate);
360int audio_resample(ReSampleContext *s, short *output, short *input, int nb_samples);
361void audio_resample_close(ReSampleContext *s);
362
363/* YUV420 format is assumed ! */
364
365struct ImgReSampleContext;
366
367typedef struct ImgReSampleContext ImgReSampleContext;
368
369ImgReSampleContext *img_resample_init(int output_width, int output_height,
370 int input_width, int input_height);
371void img_resample(ImgReSampleContext *s,
372 AVPicture *output, AVPicture *input);
373
374void img_resample_close(ImgReSampleContext *s);
375
219b06c6
FB
376void avpicture_fill(AVPicture *picture, UINT8 *ptr,
377 int pix_fmt, int width, int height);
378int avpicture_get_size(int pix_fmt, int width, int height);
379
380/* convert among pixel formats */
381int img_convert(AVPicture *dst, int dst_pix_fmt,
382 AVPicture *src, int pix_fmt,
383 int width, int height);
384
385/* deinterlace a picture */
386int avpicture_deinterlace(AVPicture *dst, AVPicture *src,
de6d9b64
FB
387 int pix_fmt, int width, int height);
388
389/* external high level API */
390
391extern AVCodec *first_avcodec;
392
156e5023 393/* returns LIBAVCODEC_VERSION_INT constant */
51a49663 394unsigned avcodec_version(void);
8bceb6af 395/* returns LIBAVCODEC_BUILD constant */
51a49663 396unsigned avcodec_build(void);
de6d9b64
FB
397void avcodec_init(void);
398
51a49663
FB
399void avcodec_set_bit_exact(void);
400
de6d9b64
FB
401void register_avcodec(AVCodec *format);
402AVCodec *avcodec_find_encoder(enum CodecID id);
98f3b098 403AVCodec *avcodec_find_encoder_by_name(const char *name);
de6d9b64
FB
404AVCodec *avcodec_find_decoder(enum CodecID id);
405AVCodec *avcodec_find_decoder_by_name(const char *name);
406void avcodec_string(char *buf, int buf_size, AVCodecContext *enc, int encode);
407
408int avcodec_open(AVCodecContext *avctx, AVCodec *codec);
409int avcodec_decode_audio(AVCodecContext *avctx, INT16 *samples,
410 int *frame_size_ptr,
411 UINT8 *buf, int buf_size);
412int avcodec_decode_video(AVCodecContext *avctx, AVPicture *picture,
413 int *got_picture_ptr,
414 UINT8 *buf, int buf_size);
415int avcodec_encode_audio(AVCodecContext *avctx, UINT8 *buf, int buf_size,
416 const short *samples);
417int avcodec_encode_video(AVCodecContext *avctx, UINT8 *buf, int buf_size,
418 const AVPicture *pict);
419
420int avcodec_close(AVCodecContext *avctx);
421
422void avcodec_register_all(void);
a96b68b7 423
1c2a8c7f
MN
424void avcodec_flush_buffers(AVCodecContext *avctx);
425
79e7b305
A
426#ifdef FF_POSTPROCESS
427#ifndef MBC
2c029cd1
A
428#define MBC 128
429#define MBR 96
79e7b305
A
430#endif
431extern int quant_store[MBR+1][MBC+1]; // [Review]
432#endif
433
e8f14793
ZK
434
435/**
436 * Interface for 0.5.0 version
437 *
438 * do not even think about it's usage for this moment
439 */
440
441typedef struct {
442 // compressed size used from given memory buffer
443 int size;
444 /// I/P/B frame type
445 int frame_type;
446} avc_enc_result_t;
447
448/**
449 * Commands
450 * order can't be changed - once it was defined
451 */
452typedef enum {
453 // general commands
454 AVC_OPEN_BY_NAME = 0xACA000,
455 AVC_OPEN_BY_CODEC_ID,
456 AVC_OPEN_BY_FOURCC,
457 AVC_CLOSE,
458
459 AVC_FLUSH,
460 // pin - struct { uint8_t* src, uint_t src_size }
461 // pout - struct { AVPicture* img, consumed_bytes,
462 AVC_DECODE,
463 // pin - struct { AVPicture* img, uint8_t* dest, uint_t dest_size }
464 // pout - uint_t used_from_dest_size
465 AVC_ENCODE,
466
467 // query/get video commands
468 AVC_GET_VERSION = 0xACB000,
469 AVC_GET_WIDTH,
470 AVC_GET_HEIGHT,
471 AVC_GET_DELAY,
472 AVC_GET_QUANT_TABLE,
473 // ...
474
475 // query/get audio commands
476 AVC_GET_FRAME_SIZE = 0xABC000,
477
478 // maybe define some simple structure which
479 // might be passed to the user - but they can't
480 // contain any codec specific parts and these
481 // calls are usualy necessary only few times
482
483 // set video commands
484 AVC_SET_WIDTH = 0xACD000,
485 AVC_SET_HEIGHT,
486
487 // set video encoding commands
488 AVC_SET_FRAME_RATE = 0xACD800,
489 AVC_SET_QUALITY,
490 AVC_SET_HURRY_UP,
491
492 // set audio commands
493 AVC_SET_SAMPLE_RATE = 0xACE000,
494 AVC_SET_CHANNELS,
495
496} avc_cmd_t;
497
498/**
499 * \param handle allocated private structure by libavcodec
500 * for initialization pass NULL - will be returned pout
501 * user is supposed to know nothing about its structure
502 * \param cmd type of operation to be performed
503 * \param pint input parameter
504 * \param pout output parameter
505 *
506 * \returns command status - eventually for query command it might return
507 * integer resulting value
508 */
509int avcodec(void* handle, avc_cmd_t cmd, void* pin, void* pout);
510
a96b68b7 511#endif /* AVCODEC_H */