| 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 | * |
| 171 | * @see av_opt_set_int() |
| 172 | * @see av_opt_set_dict() |
| 173 | * |
| 174 | * @param avr audio resample context |
| 175 | * @return 0 on success, negative AVERROR code on failure |
| 176 | */ |
| 177 | int avresample_open(AVAudioResampleContext *avr); |
| 178 | |
| 179 | /** |
| 180 | * Check whether an AVAudioResampleContext is open or closed. |
| 181 | * |
| 182 | * @param avr AVAudioResampleContext to check |
| 183 | * @return 1 if avr is open, 0 if avr is closed. |
| 184 | */ |
| 185 | int avresample_is_open(AVAudioResampleContext *avr); |
| 186 | |
| 187 | /** |
| 188 | * Close AVAudioResampleContext. |
| 189 | * |
| 190 | * This closes the context, but it does not change the parameters. The context |
| 191 | * can be reopened with avresample_open(). It does, however, clear the output |
| 192 | * FIFO and any remaining leftover samples in the resampling delay buffer. If |
| 193 | * there was a custom matrix being used, that is also cleared. |
| 194 | * |
| 195 | * @see avresample_convert() |
| 196 | * @see avresample_set_matrix() |
| 197 | * |
| 198 | * @param avr audio resample context |
| 199 | */ |
| 200 | void avresample_close(AVAudioResampleContext *avr); |
| 201 | |
| 202 | /** |
| 203 | * Free AVAudioResampleContext and associated AVOption values. |
| 204 | * |
| 205 | * This also calls avresample_close() before freeing. |
| 206 | * |
| 207 | * @param avr audio resample context |
| 208 | */ |
| 209 | void avresample_free(AVAudioResampleContext **avr); |
| 210 | |
| 211 | /** |
| 212 | * Generate a channel mixing matrix. |
| 213 | * |
| 214 | * This function is the one used internally by libavresample for building the |
| 215 | * default mixing matrix. It is made public just as a utility function for |
| 216 | * building custom matrices. |
| 217 | * |
| 218 | * @param in_layout input channel layout |
| 219 | * @param out_layout output channel layout |
| 220 | * @param center_mix_level mix level for the center channel |
| 221 | * @param surround_mix_level mix level for the surround channel(s) |
| 222 | * @param lfe_mix_level mix level for the low-frequency effects channel |
| 223 | * @param normalize if 1, coefficients will be normalized to prevent |
| 224 | * overflow. if 0, coefficients will not be |
| 225 | * normalized. |
| 226 | * @param[out] matrix mixing coefficients; matrix[i + stride * o] is |
| 227 | * the weight of input channel i in output channel o. |
| 228 | * @param stride distance between adjacent input channels in the |
| 229 | * matrix array |
| 230 | * @param matrix_encoding matrixed stereo downmix mode (e.g. dplii) |
| 231 | * @return 0 on success, negative AVERROR code on failure |
| 232 | */ |
| 233 | int avresample_build_matrix(uint64_t in_layout, uint64_t out_layout, |
| 234 | double center_mix_level, double surround_mix_level, |
| 235 | double lfe_mix_level, int normalize, double *matrix, |
| 236 | int stride, enum AVMatrixEncoding matrix_encoding); |
| 237 | |
| 238 | /** |
| 239 | * Get the current channel mixing matrix. |
| 240 | * |
| 241 | * If no custom matrix has been previously set or the AVAudioResampleContext is |
| 242 | * not open, an error is returned. |
| 243 | * |
| 244 | * @param avr audio resample context |
| 245 | * @param matrix mixing coefficients; matrix[i + stride * o] is the weight of |
| 246 | * input channel i in output channel o. |
| 247 | * @param stride distance between adjacent input channels in the matrix array |
| 248 | * @return 0 on success, negative AVERROR code on failure |
| 249 | */ |
| 250 | int avresample_get_matrix(AVAudioResampleContext *avr, double *matrix, |
| 251 | int stride); |
| 252 | |
| 253 | /** |
| 254 | * Set channel mixing matrix. |
| 255 | * |
| 256 | * Allows for setting a custom mixing matrix, overriding the default matrix |
| 257 | * generated internally during avresample_open(). This function can be called |
| 258 | * anytime on an allocated context, either before or after calling |
| 259 | * avresample_open(), as long as the channel layouts have been set. |
| 260 | * avresample_convert() always uses the current matrix. |
| 261 | * Calling avresample_close() on the context will clear the current matrix. |
| 262 | * |
| 263 | * @see avresample_close() |
| 264 | * |
| 265 | * @param avr audio resample context |
| 266 | * @param matrix mixing coefficients; matrix[i + stride * o] is the weight of |
| 267 | * input channel i in output channel o. |
| 268 | * @param stride distance between adjacent input channels in the matrix array |
| 269 | * @return 0 on success, negative AVERROR code on failure |
| 270 | */ |
| 271 | int avresample_set_matrix(AVAudioResampleContext *avr, const double *matrix, |
| 272 | int stride); |
| 273 | |
| 274 | /** |
| 275 | * Set a customized input channel mapping. |
| 276 | * |
| 277 | * This function can only be called when the allocated context is not open. |
| 278 | * Also, the input channel layout must have already been set. |
| 279 | * |
| 280 | * Calling avresample_close() on the context will clear the channel mapping. |
| 281 | * |
| 282 | * The map for each input channel specifies the channel index in the source to |
| 283 | * use for that particular channel, or -1 to mute the channel. Source channels |
| 284 | * can be duplicated by using the same index for multiple input channels. |
| 285 | * |
| 286 | * Examples: |
| 287 | * |
| 288 | * Reordering 5.1 AAC order (C,L,R,Ls,Rs,LFE) to FFmpeg order (L,R,C,LFE,Ls,Rs): |
| 289 | * { 1, 2, 0, 5, 3, 4 } |
| 290 | * |
| 291 | * Muting the 3rd channel in 4-channel input: |
| 292 | * { 0, 1, -1, 3 } |
| 293 | * |
| 294 | * Duplicating the left channel of stereo input: |
| 295 | * { 0, 0 } |
| 296 | * |
| 297 | * @param avr audio resample context |
| 298 | * @param channel_map customized input channel mapping |
| 299 | * @return 0 on success, negative AVERROR code on failure |
| 300 | */ |
| 301 | int avresample_set_channel_mapping(AVAudioResampleContext *avr, |
| 302 | const int *channel_map); |
| 303 | |
| 304 | /** |
| 305 | * Set compensation for resampling. |
| 306 | * |
| 307 | * This can be called anytime after avresample_open(). If resampling is not |
| 308 | * automatically enabled because of a sample rate conversion, the |
| 309 | * "force_resampling" option must have been set to 1 when opening the context |
| 310 | * in order to use resampling compensation. |
| 311 | * |
| 312 | * @param avr audio resample context |
| 313 | * @param sample_delta compensation delta, in samples |
| 314 | * @param compensation_distance compensation distance, in samples |
| 315 | * @return 0 on success, negative AVERROR code on failure |
| 316 | */ |
| 317 | int avresample_set_compensation(AVAudioResampleContext *avr, int sample_delta, |
| 318 | int compensation_distance); |
| 319 | |
| 320 | /** |
| 321 | * Provide the upper bound on the number of samples the configured |
| 322 | * conversion would output. |
| 323 | * |
| 324 | * @param avr audio resample context |
| 325 | * @param in_nb_samples number of input samples |
| 326 | * |
| 327 | * @return number of samples or AVERROR(EINVAL) if the value |
| 328 | * would exceed INT_MAX |
| 329 | */ |
| 330 | |
| 331 | int avresample_get_out_samples(AVAudioResampleContext *avr, int in_nb_samples); |
| 332 | |
| 333 | /** |
| 334 | * Convert input samples and write them to the output FIFO. |
| 335 | * |
| 336 | * The upper bound on the number of output samples can be obtained through |
| 337 | * avresample_get_out_samples(). |
| 338 | * |
| 339 | * The output data can be NULL or have fewer allocated samples than required. |
| 340 | * In this case, any remaining samples not written to the output will be added |
| 341 | * to an internal FIFO buffer, to be returned at the next call to this function |
| 342 | * or to avresample_read(). |
| 343 | * |
| 344 | * If converting sample rate, there may be data remaining in the internal |
| 345 | * resampling delay buffer. avresample_get_delay() tells the number of remaining |
| 346 | * samples. To get this data as output, call avresample_convert() with NULL |
| 347 | * input. |
| 348 | * |
| 349 | * At the end of the conversion process, there may be data remaining in the |
| 350 | * internal FIFO buffer. avresample_available() tells the number of remaining |
| 351 | * samples. To get this data as output, either call avresample_convert() with |
| 352 | * NULL input or call avresample_read(). |
| 353 | * |
| 354 | * @see avresample_get_out_samples() |
| 355 | * @see avresample_read() |
| 356 | * @see avresample_get_delay() |
| 357 | * |
| 358 | * @param avr audio resample context |
| 359 | * @param output output data pointers |
| 360 | * @param out_plane_size output plane size, in bytes. |
| 361 | * This can be 0 if unknown, but that will lead to |
| 362 | * optimized functions not being used directly on the |
| 363 | * output, which could slow down some conversions. |
| 364 | * @param out_samples maximum number of samples that the output buffer can hold |
| 365 | * @param input input data pointers |
| 366 | * @param in_plane_size input plane size, in bytes |
| 367 | * This can be 0 if unknown, but that will lead to |
| 368 | * optimized functions not being used directly on the |
| 369 | * input, which could slow down some conversions. |
| 370 | * @param in_samples number of input samples to convert |
| 371 | * @return number of samples written to the output buffer, |
| 372 | * not including converted samples added to the internal |
| 373 | * output FIFO |
| 374 | */ |
| 375 | int avresample_convert(AVAudioResampleContext *avr, uint8_t **output, |
| 376 | int out_plane_size, int out_samples, uint8_t **input, |
| 377 | int in_plane_size, int in_samples); |
| 378 | |
| 379 | /** |
| 380 | * Return the number of samples currently in the resampling delay buffer. |
| 381 | * |
| 382 | * When resampling, there may be a delay between the input and output. Any |
| 383 | * unconverted samples in each call are stored internally in a delay buffer. |
| 384 | * This function allows the user to determine the current number of samples in |
| 385 | * the delay buffer, which can be useful for synchronization. |
| 386 | * |
| 387 | * @see avresample_convert() |
| 388 | * |
| 389 | * @param avr audio resample context |
| 390 | * @return number of samples currently in the resampling delay buffer |
| 391 | */ |
| 392 | int avresample_get_delay(AVAudioResampleContext *avr); |
| 393 | |
| 394 | /** |
| 395 | * Return the number of available samples in the output FIFO. |
| 396 | * |
| 397 | * During conversion, if the user does not specify an output buffer or |
| 398 | * specifies an output buffer that is smaller than what is needed, remaining |
| 399 | * samples that are not written to the output are stored to an internal FIFO |
| 400 | * buffer. The samples in the FIFO can be read with avresample_read() or |
| 401 | * avresample_convert(). |
| 402 | * |
| 403 | * @see avresample_read() |
| 404 | * @see avresample_convert() |
| 405 | * |
| 406 | * @param avr audio resample context |
| 407 | * @return number of samples available for reading |
| 408 | */ |
| 409 | int avresample_available(AVAudioResampleContext *avr); |
| 410 | |
| 411 | /** |
| 412 | * Read samples from the output FIFO. |
| 413 | * |
| 414 | * During conversion, if the user does not specify an output buffer or |
| 415 | * specifies an output buffer that is smaller than what is needed, remaining |
| 416 | * samples that are not written to the output are stored to an internal FIFO |
| 417 | * buffer. This function can be used to read samples from that internal FIFO. |
| 418 | * |
| 419 | * @see avresample_available() |
| 420 | * @see avresample_convert() |
| 421 | * |
| 422 | * @param avr audio resample context |
| 423 | * @param output output data pointers. May be NULL, in which case |
| 424 | * nb_samples of data is discarded from output FIFO. |
| 425 | * @param nb_samples number of samples to read from the FIFO |
| 426 | * @return the number of samples written to output |
| 427 | */ |
| 428 | int avresample_read(AVAudioResampleContext *avr, uint8_t **output, int nb_samples); |
| 429 | |
| 430 | /** |
| 431 | * Convert the samples in the input AVFrame and write them to the output AVFrame. |
| 432 | * |
| 433 | * Input and output AVFrames must have channel_layout, sample_rate and format set. |
| 434 | * |
| 435 | * The upper bound on the number of output samples is obtained through |
| 436 | * avresample_get_out_samples(). |
| 437 | * |
| 438 | * If the output AVFrame does not have the data pointers allocated the nb_samples |
| 439 | * field will be set using avresample_get_out_samples() and av_frame_get_buffer() |
| 440 | * is called to allocate the frame. |
| 441 | * |
| 442 | * The output AVFrame can be NULL or have fewer allocated samples than required. |
| 443 | * In this case, any remaining samples not written to the output will be added |
| 444 | * to an internal FIFO buffer, to be returned at the next call to this function |
| 445 | * or to avresample_convert() or to avresample_read(). |
| 446 | * |
| 447 | * If converting sample rate, there may be data remaining in the internal |
| 448 | * resampling delay buffer. avresample_get_delay() tells the number of |
| 449 | * remaining samples. To get this data as output, call this function or |
| 450 | * avresample_convert() with NULL input. |
| 451 | * |
| 452 | * At the end of the conversion process, there may be data remaining in the |
| 453 | * internal FIFO buffer. avresample_available() tells the number of remaining |
| 454 | * samples. To get this data as output, either call this function or |
| 455 | * avresample_convert() with NULL input or call avresample_read(). |
| 456 | * |
| 457 | * If the AVAudioResampleContext configuration does not match the output and |
| 458 | * input AVFrame settings the conversion does not take place and depending on |
| 459 | * which AVFrame is not matching AVERROR_OUTPUT_CHANGED, AVERROR_INPUT_CHANGED |
| 460 | * or AVERROR_OUTPUT_CHANGED|AVERROR_INPUT_CHANGED is returned. |
| 461 | * |
| 462 | * @see avresample_get_out_samples() |
| 463 | * @see avresample_available() |
| 464 | * @see avresample_convert() |
| 465 | * @see avresample_read() |
| 466 | * @see avresample_get_delay() |
| 467 | * |
| 468 | * @param avr audio resample context |
| 469 | * @param output output AVFrame |
| 470 | * @param input input AVFrame |
| 471 | * @return 0 on success, AVERROR on failure or nonmatching |
| 472 | * configuration. |
| 473 | */ |
| 474 | int avresample_convert_frame(AVAudioResampleContext *avr, |
| 475 | AVFrame *output, AVFrame *input); |
| 476 | |
| 477 | /** |
| 478 | * Configure or reconfigure the AVAudioResampleContext using the information |
| 479 | * provided by the AVFrames. |
| 480 | * |
| 481 | * The original resampling context is reset even on failure. |
| 482 | * The function calls avresample_close() internally if the context is open. |
| 483 | * |
| 484 | * @see avresample_open(); |
| 485 | * @see avresample_close(); |
| 486 | * |
| 487 | * @param avr audio resample context |
| 488 | * @param output output AVFrame |
| 489 | * @param input input AVFrame |
| 490 | * @return 0 on success, AVERROR on failure. |
| 491 | */ |
| 492 | int avresample_config(AVAudioResampleContext *avr, AVFrame *out, AVFrame *in); |
| 493 | |
| 494 | /** |
| 495 | * @} |
| 496 | */ |
| 497 | |
| 498 | #endif /* AVRESAMPLE_AVRESAMPLE_H */ |