- ffmpeg relies on libavcodec to repeat frames from 3:2 sources.
[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"
c80340a4
J
8#define LIBAVCODEC_BUILD 4607
9#define LIBAVCODEC_BUILD_STR "4607"
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 */
c80340a4
J
133 int repeat_pict; /* set this to 1 if you want the decoder
134 to repeat frames for 3:2 pulldown (MPEG-2) */
bf89e6b1
FB
135 /* if non NULL, 'draw_horiz_band' is called by the libavcodec
136 decoder to draw an horizontal band. It improve cache usage. Not
137 all codecs can do that. You must check the codec capabilities
138 before */
139 void (*draw_horiz_band)(struct AVCodecContext *s,
140 UINT8 **src_ptr, int linesize,
141 int y, int width, int height);
142
de6d9b64
FB
143 /* audio only */
144 int sample_rate; /* samples per sec */
145 int channels;
a96b68b7 146 int sample_fmt; /* sample format, currenly unused */
de6d9b64
FB
147
148 /* the following data should not be initialized */
9cd3766f
J
149 int frame_size; /* in samples, initialized when calling 'init' */
150 int frame_number; /* audio or video frame number */
151 int real_pict_num; /* returns the real picture number of
152 previous encoded frame */
153 int key_frame; /* true if the previous compressed frame was
154 a key frame (intra, or seekable) */
155 int pict_type; /* picture type of the previous
156 encoded frame */
157/* FIXME: these should have FF_ */
158#define I_TYPE 1 // Intra
159#define P_TYPE 2 // Predicted
160#define B_TYPE 3 // Bi-dir predicted
161#define S_TYPE 4 // S(GMC)-VOP MPEG4
162
163 int delay; /* number of frames the decoded output
164 will be delayed relative to the encoded input */
2417652e
MN
165 uint8_t *mbskip_table; /* =1 if MB didnt change, is only valid for I/P frames
166 stride= mb_width = (width+15)>>4 */
1ff662cc
MN
167
168 /* encoding parameters */
de6d9b64 169 int quality; /* quality of the previous encoded frame
8b4c7dbc
MN
170 (between 1 (good) and 31 (bad))
171 this is allso used to set the quality in vbr mode
172 and the per frame quality in CODEC_FLAG_TYPE (second pass mode) */
9cdd6a24
MN
173 float qcompress; /* amount of qscale change between easy & hard scenes (0.0-1.0)*/
174 float qblur; /* amount of qscale smoothing over time (0.0-1.0) */
175 int qmin; /* min qscale */
176 int qmax; /* max qscale */
177 int max_qdiff; /* max qscale difference between frames */
9dbcbd92 178 int max_b_frames; /* maximum b frames, the output will be delayed by max_b_frames+1 relative to the input */
8b4c7dbc
MN
179 float b_quant_factor;/* qscale factor between ips and b frames */
180 int rc_strategy;
181 int b_frame_strategy;
9dbcbd92 182
2417652e
MN
183 int hurry_up; /* when set to 1 during decoding, b frames will be skiped
184 when set to 2 idct/dequant will be skipped too */
185
de6d9b64
FB
186 struct AVCodec *codec;
187 void *priv_data;
188
644d98a4
J
189 /* The following data is for RTP friendly coding */
190 /* By now only H.263/H.263+ coder honours this */
191 int rtp_mode; /* 1 for activate RTP friendly-mode */
192 /* highers numbers represent more error-prone */
193 /* enviroments, by now just "1" exist */
194
195 int rtp_payload_size; /* The size of the RTP payload, the coder will */
196 /* do it's best to deliver a chunk with size */
197 /* below rtp_payload_size, the chunk will start */
198 /* with a start code on some codecs like H.263 */
199 /* This doesn't take account of any particular */
200 /* headers inside the transmited RTP payload */
81401c1f
J
201
202
203 /* The RTP callcack: This function is called */
204 /* every time the encoder as a packet to send */
205 /* Depends on the encoder if the data starts */
206 /* with a Start Code (it should) H.263 does */
207 void (*rtp_callback)(void *data, int size, int packet_number);
208
43f1708f
J
209 /* These are for PSNR calculation, if you set get_psnr to 1 */
210 /* after encoding you will have the PSNR on psnr_y/cb/cr */
211 int get_psnr;
212 float psnr_y;
213 float psnr_cb;
214 float psnr_cr;
098eefe1
MN
215
216 /* statistics, used for 2-pass encoding */
217 int mv_bits;
218 int header_bits;
219 int i_tex_bits;
220 int p_tex_bits;
221 int i_count;
222 int p_count;
223 int skip_count;
224 int misc_bits; // cbp, mb_type
225 int frame_bits;
644d98a4 226
de6d9b64 227 /* the following fields are ignored */
bf89e6b1 228 void *opaque; /* can be used to carry app specific stuff */
de6d9b64 229 char codec_name[32];
4a663d80
PG
230 enum CodecType codec_type; /* see CODEC_TYPE_xxx */
231 enum CodecID codec_id; /* see CODEC_ID_xxx */
de6d9b64
FB
232 unsigned int codec_tag; /* codec tag, only used if unknown codec */
233} AVCodecContext;
234
235typedef struct AVCodec {
236 char *name;
237 int type;
238 int id;
239 int priv_data_size;
240 int (*init)(AVCodecContext *);
241 int (*encode)(AVCodecContext *, UINT8 *buf, int buf_size, void *data);
242 int (*close)(AVCodecContext *);
243 int (*decode)(AVCodecContext *, void *outdata, int *outdata_size,
244 UINT8 *buf, int buf_size);
bf89e6b1 245 int capabilities;
de6d9b64
FB
246 struct AVCodec *next;
247} AVCodec;
248
249/* three components are given, that's all */
250typedef struct AVPicture {
251 UINT8 *data[3];
252 int linesize[3];
253} AVPicture;
254
255extern AVCodec ac3_encoder;
256extern AVCodec mp2_encoder;
9d36bdc9 257extern AVCodec mp3lame_encoder;
de6d9b64
FB
258extern AVCodec mpeg1video_encoder;
259extern AVCodec h263_encoder;
260extern AVCodec h263p_encoder;
261extern AVCodec rv10_encoder;
262extern AVCodec mjpeg_encoder;
bf89e6b1 263extern AVCodec mpeg4_encoder;
84afee34
MN
264extern AVCodec msmpeg4v1_encoder;
265extern AVCodec msmpeg4v2_encoder;
266extern AVCodec msmpeg4v3_encoder;
de6d9b64
FB
267
268extern AVCodec h263_decoder;
bf89e6b1 269extern AVCodec mpeg4_decoder;
84afee34
MN
270extern AVCodec msmpeg4v1_decoder;
271extern AVCodec msmpeg4v2_decoder;
272extern AVCodec msmpeg4v3_decoder;
e1a9dbff 273extern AVCodec wmv1_decoder;
de6d9b64
FB
274extern AVCodec mpeg_decoder;
275extern AVCodec h263i_decoder;
276extern AVCodec rv10_decoder;
4e66ab3b 277extern AVCodec mjpeg_decoder;
4b1f4f23 278extern AVCodec mp2_decoder;
a96b68b7 279extern AVCodec mp3_decoder;
de6d9b64 280
a96b68b7
FB
281/* pcm codecs */
282#define PCM_CODEC(id, name) \
283extern AVCodec name ## _decoder; \
284extern AVCodec name ## _encoder;
285
286PCM_CODEC(CODEC_ID_PCM_S16LE, pcm_s16le);
287PCM_CODEC(CODEC_ID_PCM_S16BE, pcm_s16be);
288PCM_CODEC(CODEC_ID_PCM_U16LE, pcm_u16le);
289PCM_CODEC(CODEC_ID_PCM_U16BE, pcm_u16be);
290PCM_CODEC(CODEC_ID_PCM_S8, pcm_s8);
291PCM_CODEC(CODEC_ID_PCM_U8, pcm_u8);
292PCM_CODEC(CODEC_ID_PCM_ALAW, pcm_alaw);
293PCM_CODEC(CODEC_ID_PCM_MULAW, pcm_mulaw);
294
295#undef PCM_CODEC
296
297/* dummy raw video codec */
de6d9b64
FB
298extern AVCodec rawvideo_codec;
299
300/* the following codecs use external GPL libs */
de6d9b64
FB
301extern AVCodec ac3_decoder;
302
303/* resample.c */
304
305struct ReSampleContext;
306
307typedef struct ReSampleContext ReSampleContext;
308
309ReSampleContext *audio_resample_init(int output_channels, int input_channels,
310 int output_rate, int input_rate);
311int audio_resample(ReSampleContext *s, short *output, short *input, int nb_samples);
312void audio_resample_close(ReSampleContext *s);
313
314/* YUV420 format is assumed ! */
315
316struct ImgReSampleContext;
317
318typedef struct ImgReSampleContext ImgReSampleContext;
319
320ImgReSampleContext *img_resample_init(int output_width, int output_height,
321 int input_width, int input_height);
322void img_resample(ImgReSampleContext *s,
323 AVPicture *output, AVPicture *input);
324
325void img_resample_close(ImgReSampleContext *s);
326
219b06c6
FB
327void avpicture_fill(AVPicture *picture, UINT8 *ptr,
328 int pix_fmt, int width, int height);
329int avpicture_get_size(int pix_fmt, int width, int height);
330
331/* convert among pixel formats */
332int img_convert(AVPicture *dst, int dst_pix_fmt,
333 AVPicture *src, int pix_fmt,
334 int width, int height);
335
336/* deinterlace a picture */
337int avpicture_deinterlace(AVPicture *dst, AVPicture *src,
de6d9b64
FB
338 int pix_fmt, int width, int height);
339
340/* external high level API */
341
342extern AVCodec *first_avcodec;
343
156e5023
NK
344/* returns LIBAVCODEC_VERSION_INT constant */
345unsigned avcodec_version( void );
de6d9b64
FB
346void avcodec_init(void);
347
348void register_avcodec(AVCodec *format);
349AVCodec *avcodec_find_encoder(enum CodecID id);
98f3b098 350AVCodec *avcodec_find_encoder_by_name(const char *name);
de6d9b64
FB
351AVCodec *avcodec_find_decoder(enum CodecID id);
352AVCodec *avcodec_find_decoder_by_name(const char *name);
353void avcodec_string(char *buf, int buf_size, AVCodecContext *enc, int encode);
354
355int avcodec_open(AVCodecContext *avctx, AVCodec *codec);
356int avcodec_decode_audio(AVCodecContext *avctx, INT16 *samples,
357 int *frame_size_ptr,
358 UINT8 *buf, int buf_size);
359int avcodec_decode_video(AVCodecContext *avctx, AVPicture *picture,
360 int *got_picture_ptr,
361 UINT8 *buf, int buf_size);
362int avcodec_encode_audio(AVCodecContext *avctx, UINT8 *buf, int buf_size,
363 const short *samples);
364int avcodec_encode_video(AVCodecContext *avctx, UINT8 *buf, int buf_size,
365 const AVPicture *pict);
366
367int avcodec_close(AVCodecContext *avctx);
368
369void avcodec_register_all(void);
a96b68b7 370
1c2a8c7f
MN
371void avcodec_flush_buffers(AVCodecContext *avctx);
372
79e7b305
A
373#ifdef FF_POSTPROCESS
374#ifndef MBC
2c029cd1
A
375#define MBC 128
376#define MBR 96
79e7b305
A
377#endif
378extern int quant_store[MBR+1][MBC+1]; // [Review]
379#endif
380
a96b68b7 381#endif /* AVCODEC_H */