blob: 4cfbdfcf8a9ff86d9a3daf606f4921a028fa77c3
1 | /* |
2 | * Copyright (c) 2012 Justin Ruggles <justin.ruggles@gmail.com> |
3 | * |
4 | * This file is part of FFmpeg. |
5 | * |
6 | * FFmpeg 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 2.1 of the License, or (at your option) any later version. |
10 | * |
11 | * FFmpeg 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 FFmpeg; if not, write to the Free Software |
18 | * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA |
19 | */ |
20 | |
21 | #ifndef AVRESAMPLE_AVRESAMPLE_H |
22 | #define AVRESAMPLE_AVRESAMPLE_H |
23 | |
24 | /** |
25 | * @file |
26 | * @ingroup lavr |
27 | * external API header |
28 | */ |
29 | |
30 | /** |
31 | * @defgroup lavr libavresample |
32 | * @{ |
33 | * |
34 | * Libavresample (lavr) is a library that handles audio resampling, sample |
35 | * format conversion and mixing. |
36 | * |
37 | * Interaction with lavr is done through AVAudioResampleContext, which is |
38 | * allocated with avresample_alloc_context(). It is opaque, so all parameters |
39 | * must be set with the @ref avoptions API. |
40 | * |
41 | * For example the following code will setup conversion from planar float sample |
42 | * format to interleaved signed 16-bit integer, downsampling from 48kHz to |
43 | * 44.1kHz and downmixing from 5.1 channels to stereo (using the default mixing |
44 | * matrix): |
45 | * @code |
46 | * AVAudioResampleContext *avr = avresample_alloc_context(); |
47 | * av_opt_set_int(avr, "in_channel_layout", AV_CH_LAYOUT_5POINT1, 0); |
48 | * av_opt_set_int(avr, "out_channel_layout", AV_CH_LAYOUT_STEREO, 0); |
49 | * av_opt_set_int(avr, "in_sample_rate", 48000, 0); |
50 | * av_opt_set_int(avr, "out_sample_rate", 44100, 0); |
51 | * av_opt_set_int(avr, "in_sample_fmt", AV_SAMPLE_FMT_FLTP, 0); |
52 | * av_opt_set_int(avr, "out_sample_fmt", AV_SAMPLE_FMT_S16, 0); |
53 | * @endcode |
54 | * |
55 | * Once the context is initialized, it must be opened with avresample_open(). If |
56 | * you need to change the conversion parameters, you must close the context with |
57 | * avresample_close(), change the parameters as described above, then reopen it |
58 | * again. |
59 | * |
60 | * The conversion itself is done by repeatedly calling avresample_convert(). |
61 | * Note that the samples may get buffered in two places in lavr. The first one |
62 | * is the output FIFO, where the samples end up if the output buffer is not |
63 | * large enough. The data stored in there may be retrieved at any time with |
64 | * avresample_read(). The second place is the resampling delay buffer, |
65 | * applicable only when resampling is done. The samples in it require more input |
66 | * before they can be processed. Their current amount is returned by |
67 | * avresample_get_delay(). At the end of conversion the resampling buffer can be |
68 | * flushed by calling avresample_convert() with NULL input. |
69 | * |
70 | * The following code demonstrates the conversion loop assuming the parameters |
71 | * from above and caller-defined functions get_input() and handle_output(): |
72 | * @code |
73 | * uint8_t **input; |
74 | * int in_linesize, in_samples; |
75 | * |
76 | * while (get_input(&input, &in_linesize, &in_samples)) { |
77 | * uint8_t *output |
78 | * int out_linesize; |
79 | * int out_samples = avresample_get_out_samples(avr, in_samples); |
80 | * |
81 | * av_samples_alloc(&output, &out_linesize, 2, out_samples, |
82 | * AV_SAMPLE_FMT_S16, 0); |
83 | * out_samples = avresample_convert(avr, &output, out_linesize, out_samples, |
84 | * input, in_linesize, in_samples); |
85 | * handle_output(output, out_linesize, out_samples); |
86 | * av_freep(&output); |
87 | * } |
88 | * @endcode |
89 | * |
90 | * When the conversion is finished and the FIFOs are flushed if required, the |
91 | * conversion context and everything associated with it must be freed with |
92 | * avresample_free(). |
93 | */ |
94 | |
95 | #include "libavutil/avutil.h" |
96 | #include "libavutil/channel_layout.h" |
97 | #include "libavutil/dict.h" |
98 | #include "libavutil/frame.h" |
99 | #include "libavutil/log.h" |
100 | #include "libavutil/mathematics.h" |
101 | |
102 | #include "libavresample/version.h" |
103 | |
104 | #define AVRESAMPLE_MAX_CHANNELS 32 |
105 | |
106 | typedef struct AVAudioResampleContext AVAudioResampleContext; |
107 | |
108 | /** Mixing Coefficient Types */ |
109 | enum AVMixCoeffType { |
110 | AV_MIX_COEFF_TYPE_Q8, /** 16-bit 8.8 fixed-point */ |
111 | AV_MIX_COEFF_TYPE_Q15, /** 32-bit 17.15 fixed-point */ |
112 | AV_MIX_COEFF_TYPE_FLT, /** floating-point */ |
113 | AV_MIX_COEFF_TYPE_NB, /** Number of coeff types. Not part of ABI */ |
114 | }; |
115 | |
116 | /** Resampling Filter Types */ |
117 | enum AVResampleFilterType { |
118 | AV_RESAMPLE_FILTER_TYPE_CUBIC, /**< Cubic */ |
119 | AV_RESAMPLE_FILTER_TYPE_BLACKMAN_NUTTALL, /**< Blackman Nuttall Windowed Sinc */ |
120 | AV_RESAMPLE_FILTER_TYPE_KAISER, /**< Kaiser Windowed Sinc */ |
121 | }; |
122 | |
123 | enum AVResampleDitherMethod { |
124 | AV_RESAMPLE_DITHER_NONE, /**< Do not use dithering */ |
125 | AV_RESAMPLE_DITHER_RECTANGULAR, /**< Rectangular Dither */ |
126 | AV_RESAMPLE_DITHER_TRIANGULAR, /**< Triangular Dither*/ |
127 | AV_RESAMPLE_DITHER_TRIANGULAR_HP, /**< Triangular Dither with High Pass */ |
128 | AV_RESAMPLE_DITHER_TRIANGULAR_NS, /**< Triangular Dither with Noise Shaping */ |
129 | AV_RESAMPLE_DITHER_NB, /**< Number of dither types. Not part of ABI. */ |
130 | }; |
131 | |
132 | /** |
133 | * Return the LIBAVRESAMPLE_VERSION_INT constant. |
134 | */ |
135 | unsigned avresample_version(void); |
136 | |
137 | /** |
138 | * Return the libavresample build-time configuration. |
139 | * @return configure string |
140 | */ |
141 | const char *avresample_configuration(void); |
142 | |
143 | /** |
144 | * Return the libavresample license. |
145 | */ |
146 | const char *avresample_license(void); |
147 | |
148 | /** |
149 | * Get the AVClass for AVAudioResampleContext. |
150 | * |
151 | * Can be used in combination with AV_OPT_SEARCH_FAKE_OBJ for examining options |
152 | * without allocating a context. |
153 | * |
154 | * @see av_opt_find(). |
155 | * |
156 | * @return AVClass for AVAudioResampleContext |
157 | */ |
158 | const AVClass *avresample_get_class(void); |
159 | |
160 | /** |
161 | * Allocate AVAudioResampleContext and set options. |
162 | * |
163 | * @return allocated audio resample context, or NULL on failure |
164 | */ |
165 | AVAudioResampleContext *avresample_alloc_context(void); |
166 | |
167 | /** |
168 | * Initialize AVAudioResampleContext. |
169 | * @note The context must be configured using the AVOption API. |
170 | * @note The fields "in_channel_layout", "out_channel_layout", |
171 | * "in_sample_rate", "out_sample_rate", "in_sample_fmt", |
172 | * "out_sample_fmt" must be set. |
173 | * |
174 | * @see av_opt_set_int() |
175 | * @see av_opt_set_dict() |
176 | * @see av_get_default_channel_layout() |
177 | * |
178 | * @param avr audio resample context |
179 | * @return 0 on success, negative AVERROR code on failure |
180 | */ |
181 | int avresample_open(AVAudioResampleContext *avr); |
182 | |
183 | /** |
184 | * Check whether an AVAudioResampleContext is open or closed. |
185 | * |
186 | * @param avr AVAudioResampleContext to check |
187 | * @return 1 if avr is open, 0 if avr is closed. |
188 | */ |
189 | int avresample_is_open(AVAudioResampleContext *avr); |
190 | |
191 | /** |
192 | * Close AVAudioResampleContext. |
193 | * |
194 | * This closes the context, but it does not change the parameters. The context |
195 | * can be reopened with avresample_open(). It does, however, clear the output |
196 | * FIFO and any remaining leftover samples in the resampling delay buffer. If |
197 | * there was a custom matrix being used, that is also cleared. |
198 | * |
199 | * @see avresample_convert() |
200 | * @see avresample_set_matrix() |
201 | * |
202 | * @param avr audio resample context |
203 | */ |
204 | void avresample_close(AVAudioResampleContext *avr); |
205 | |
206 | /** |
207 | * Free AVAudioResampleContext and associated AVOption values. |
208 | * |
209 | * This also calls avresample_close() before freeing. |
210 | * |
211 | * @param avr audio resample context |
212 | */ |
213 | void avresample_free(AVAudioResampleContext **avr); |
214 | |
215 | /** |
216 | * Generate a channel mixing matrix. |
217 | * |
218 | * This function is the one used internally by libavresample for building the |
219 | * default mixing matrix. It is made public just as a utility function for |
220 | * building custom matrices. |
221 | * |
222 | * @param in_layout input channel layout |
223 | * @param out_layout output channel layout |
224 | * @param center_mix_level mix level for the center channel |
225 | * @param surround_mix_level mix level for the surround channel(s) |
226 | * @param lfe_mix_level mix level for the low-frequency effects channel |
227 | * @param normalize if 1, coefficients will be normalized to prevent |
228 | * overflow. if 0, coefficients will not be |
229 | * normalized. |
230 | * @param[out] matrix mixing coefficients; matrix[i + stride * o] is |
231 | * the weight of input channel i in output channel o. |
232 | * @param stride distance between adjacent input channels in the |
233 | * matrix array |
234 | * @param matrix_encoding matrixed stereo downmix mode (e.g. dplii) |
235 | * @return 0 on success, negative AVERROR code on failure |
236 | */ |
237 | int avresample_build_matrix(uint64_t in_layout, uint64_t out_layout, |
238 | double center_mix_level, double surround_mix_level, |
239 | double lfe_mix_level, int normalize, double *matrix, |
240 | int stride, enum AVMatrixEncoding matrix_encoding); |
241 | |
242 | /** |
243 | * Get the current channel mixing matrix. |
244 | * |
245 | * If no custom matrix has been previously set or the AVAudioResampleContext is |
246 | * not open, an error is returned. |
247 | * |
248 | * @param avr audio resample context |
249 | * @param matrix mixing coefficients; matrix[i + stride * o] is the weight of |
250 | * input channel i in output channel o. |
251 | * @param stride distance between adjacent input channels in the matrix array |
252 | * @return 0 on success, negative AVERROR code on failure |
253 | */ |
254 | int avresample_get_matrix(AVAudioResampleContext *avr, double *matrix, |
255 | int stride); |
256 | |
257 | /** |
258 | * Set channel mixing matrix. |
259 | * |
260 | * Allows for setting a custom mixing matrix, overriding the default matrix |
261 | * generated internally during avresample_open(). This function can be called |
262 | * anytime on an allocated context, either before or after calling |
263 | * avresample_open(), as long as the channel layouts have been set. |
264 | * avresample_convert() always uses the current matrix. |
265 | * Calling avresample_close() on the context will clear the current matrix. |
266 | * |
267 | * @see avresample_close() |
268 | * |
269 | * @param avr audio resample context |
270 | * @param matrix mixing coefficients; matrix[i + stride * o] is the weight of |
271 | * input channel i in output channel o. |
272 | * @param stride distance between adjacent input channels in the matrix array |
273 | * @return 0 on success, negative AVERROR code on failure |
274 | */ |
275 | int avresample_set_matrix(AVAudioResampleContext *avr, const double *matrix, |
276 | int stride); |
277 | |
278 | /** |
279 | * Set a customized input channel mapping. |
280 | * |
281 | * This function can only be called when the allocated context is not open. |
282 | * Also, the input channel layout must have already been set. |
283 | * |
284 | * Calling avresample_close() on the context will clear the channel mapping. |
285 | * |
286 | * The map for each input channel specifies the channel index in the source to |
287 | * use for that particular channel, or -1 to mute the channel. Source channels |
288 | * can be duplicated by using the same index for multiple input channels. |
289 | * |
290 | * Examples: |
291 | * |
292 | * Reordering 5.1 AAC order (C,L,R,Ls,Rs,LFE) to FFmpeg order (L,R,C,LFE,Ls,Rs): |
293 | * { 1, 2, 0, 5, 3, 4 } |
294 | * |
295 | * Muting the 3rd channel in 4-channel input: |
296 | * { 0, 1, -1, 3 } |
297 | * |
298 | * Duplicating the left channel of stereo input: |
299 | * { 0, 0 } |
300 | * |
301 | * @param avr audio resample context |
302 | * @param channel_map customized input channel mapping |
303 | * @return 0 on success, negative AVERROR code on failure |
304 | */ |
305 | int avresample_set_channel_mapping(AVAudioResampleContext *avr, |
306 | const int *channel_map); |
307 | |
308 | /** |
309 | * Set compensation for resampling. |
310 | * |
311 | * This can be called anytime after avresample_open(). If resampling is not |
312 | * automatically enabled because of a sample rate conversion, the |
313 | * "force_resampling" option must have been set to 1 when opening the context |
314 | * in order to use resampling compensation. |
315 | * |
316 | * @param avr audio resample context |
317 | * @param sample_delta compensation delta, in samples |
318 | * @param compensation_distance compensation distance, in samples |
319 | * @return 0 on success, negative AVERROR code on failure |
320 | */ |
321 | int avresample_set_compensation(AVAudioResampleContext *avr, int sample_delta, |
322 | int compensation_distance); |
323 | |
324 | /** |
325 | * Provide the upper bound on the number of samples the configured |
326 | * conversion would output. |
327 | * |
328 | * @param avr audio resample context |
329 | * @param in_nb_samples number of input samples |
330 | * |
331 | * @return number of samples or AVERROR(EINVAL) if the value |
332 | * would exceed INT_MAX |
333 | */ |
334 | |
335 | int avresample_get_out_samples(AVAudioResampleContext *avr, int in_nb_samples); |
336 | |
337 | /** |
338 | * Convert input samples and write them to the output FIFO. |
339 | * |
340 | * The upper bound on the number of output samples can be obtained through |
341 | * avresample_get_out_samples(). |
342 | * |
343 | * The output data can be NULL or have fewer allocated samples than required. |
344 | * In this case, any remaining samples not written to the output will be added |
345 | * to an internal FIFO buffer, to be returned at the next call to this function |
346 | * or to avresample_read(). |
347 | * |
348 | * If converting sample rate, there may be data remaining in the internal |
349 | * resampling delay buffer. avresample_get_delay() tells the number of remaining |
350 | * samples. To get this data as output, call avresample_convert() with NULL |
351 | * input. |
352 | * |
353 | * At the end of the conversion process, there may be data remaining in the |
354 | * internal FIFO buffer. avresample_available() tells the number of remaining |
355 | * samples. To get this data as output, either call avresample_convert() with |
356 | * NULL input or call avresample_read(). |
357 | * |
358 | * @see avresample_get_out_samples() |
359 | * @see avresample_read() |
360 | * @see avresample_get_delay() |
361 | * |
362 | * @param avr audio resample context |
363 | * @param output output data pointers |
364 | * @param out_plane_size output plane size, in bytes. |
365 | * This can be 0 if unknown, but that will lead to |
366 | * optimized functions not being used directly on the |
367 | * output, which could slow down some conversions. |
368 | * @param out_samples maximum number of samples that the output buffer can hold |
369 | * @param input input data pointers |
370 | * @param in_plane_size input plane size, in bytes |
371 | * This can be 0 if unknown, but that will lead to |
372 | * optimized functions not being used directly on the |
373 | * input, which could slow down some conversions. |
374 | * @param in_samples number of input samples to convert |
375 | * @return number of samples written to the output buffer, |
376 | * not including converted samples added to the internal |
377 | * output FIFO |
378 | */ |
379 | int avresample_convert(AVAudioResampleContext *avr, uint8_t **output, |
380 | int out_plane_size, int out_samples, |
381 | uint8_t * const *input, int in_plane_size, |
382 | int in_samples); |
383 | |
384 | /** |
385 | * Return the number of samples currently in the resampling delay buffer. |
386 | * |
387 | * When resampling, there may be a delay between the input and output. Any |
388 | * unconverted samples in each call are stored internally in a delay buffer. |
389 | * This function allows the user to determine the current number of samples in |
390 | * the delay buffer, which can be useful for synchronization. |
391 | * |
392 | * @see avresample_convert() |
393 | * |
394 | * @param avr audio resample context |
395 | * @return number of samples currently in the resampling delay buffer |
396 | */ |
397 | int avresample_get_delay(AVAudioResampleContext *avr); |
398 | |
399 | /** |
400 | * Return the number of available samples in the output FIFO. |
401 | * |
402 | * During conversion, if the user does not specify an output buffer or |
403 | * specifies an output buffer that is smaller than what is needed, remaining |
404 | * samples that are not written to the output are stored to an internal FIFO |
405 | * buffer. The samples in the FIFO can be read with avresample_read() or |
406 | * avresample_convert(). |
407 | * |
408 | * @see avresample_read() |
409 | * @see avresample_convert() |
410 | * |
411 | * @param avr audio resample context |
412 | * @return number of samples available for reading |
413 | */ |
414 | int avresample_available(AVAudioResampleContext *avr); |
415 | |
416 | /** |
417 | * Read samples from the output FIFO. |
418 | * |
419 | * During conversion, if the user does not specify an output buffer or |
420 | * specifies an output buffer that is smaller than what is needed, remaining |
421 | * samples that are not written to the output are stored to an internal FIFO |
422 | * buffer. This function can be used to read samples from that internal FIFO. |
423 | * |
424 | * @see avresample_available() |
425 | * @see avresample_convert() |
426 | * |
427 | * @param avr audio resample context |
428 | * @param output output data pointers. May be NULL, in which case |
429 | * nb_samples of data is discarded from output FIFO. |
430 | * @param nb_samples number of samples to read from the FIFO |
431 | * @return the number of samples written to output |
432 | */ |
433 | int avresample_read(AVAudioResampleContext *avr, uint8_t **output, int nb_samples); |
434 | |
435 | /** |
436 | * Convert the samples in the input AVFrame and write them to the output AVFrame. |
437 | * |
438 | * Input and output AVFrames must have channel_layout, sample_rate and format set. |
439 | * |
440 | * The upper bound on the number of output samples is obtained through |
441 | * avresample_get_out_samples(). |
442 | * |
443 | * If the output AVFrame does not have the data pointers allocated the nb_samples |
444 | * field will be set using avresample_get_out_samples() and av_frame_get_buffer() |
445 | * is called to allocate the frame. |
446 | * |
447 | * The output AVFrame can be NULL or have fewer allocated samples than required. |
448 | * In this case, any remaining samples not written to the output will be added |
449 | * to an internal FIFO buffer, to be returned at the next call to this function |
450 | * or to avresample_convert() or to avresample_read(). |
451 | * |
452 | * If converting sample rate, there may be data remaining in the internal |
453 | * resampling delay buffer. avresample_get_delay() tells the number of |
454 | * remaining samples. To get this data as output, call this function or |
455 | * avresample_convert() with NULL input. |
456 | * |
457 | * At the end of the conversion process, there may be data remaining in the |
458 | * internal FIFO buffer. avresample_available() tells the number of remaining |
459 | * samples. To get this data as output, either call this function or |
460 | * avresample_convert() with NULL input or call avresample_read(). |
461 | * |
462 | * If the AVAudioResampleContext configuration does not match the output and |
463 | * input AVFrame settings the conversion does not take place and depending on |
464 | * which AVFrame is not matching AVERROR_OUTPUT_CHANGED, AVERROR_INPUT_CHANGED |
465 | * or AVERROR_OUTPUT_CHANGED|AVERROR_INPUT_CHANGED is returned. |
466 | * |
467 | * @see avresample_get_out_samples() |
468 | * @see avresample_available() |
469 | * @see avresample_convert() |
470 | * @see avresample_read() |
471 | * @see avresample_get_delay() |
472 | * |
473 | * @param avr audio resample context |
474 | * @param output output AVFrame |
475 | * @param input input AVFrame |
476 | * @return 0 on success, AVERROR on failure or nonmatching |
477 | * configuration. |
478 | */ |
479 | int avresample_convert_frame(AVAudioResampleContext *avr, |
480 | AVFrame *output, AVFrame *input); |
481 | |
482 | /** |
483 | * Configure or reconfigure the AVAudioResampleContext using the information |
484 | * provided by the AVFrames. |
485 | * |
486 | * The original resampling context is reset even on failure. |
487 | * The function calls avresample_close() internally if the context is open. |
488 | * |
489 | * @see avresample_open(); |
490 | * @see avresample_close(); |
491 | * |
492 | * @param avr audio resample context |
493 | * @param output output AVFrame |
494 | * @param input input AVFrame |
495 | * @return 0 on success, AVERROR on failure. |
496 | */ |
497 | int avresample_config(AVAudioResampleContext *avr, AVFrame *out, AVFrame *in); |
498 | |
499 | /** |
500 | * @} |
501 | */ |
502 | |
503 | #endif /* AVRESAMPLE_AVRESAMPLE_H */ |
504 |