Libav 0.7.1
|
00001 /* 00002 * samplerate conversion for both audio and video 00003 * Copyright (c) 2000 Fabrice Bellard 00004 * 00005 * This file is part of Libav. 00006 * 00007 * Libav is free software; you can redistribute it and/or 00008 * modify it under the terms of the GNU Lesser General Public 00009 * License as published by the Free Software Foundation; either 00010 * version 2.1 of the License, or (at your option) any later version. 00011 * 00012 * Libav is distributed in the hope that it will be useful, 00013 * but WITHOUT ANY WARRANTY; without even the implied warranty of 00014 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU 00015 * Lesser General Public License for more details. 00016 * 00017 * You should have received a copy of the GNU Lesser General Public 00018 * License along with Libav; if not, write to the Free Software 00019 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA 00020 */ 00021 00027 #include "avcodec.h" 00028 #include "audioconvert.h" 00029 #include "libavutil/opt.h" 00030 #include "libavutil/samplefmt.h" 00031 00032 #define MAX_CHANNELS 8 00033 00034 struct AVResampleContext; 00035 00036 static const char *context_to_name(void *ptr) 00037 { 00038 return "audioresample"; 00039 } 00040 00041 static const AVOption options[] = {{NULL}}; 00042 static const AVClass audioresample_context_class = { 00043 "ReSampleContext", context_to_name, options, LIBAVUTIL_VERSION_INT 00044 }; 00045 00046 struct ReSampleContext { 00047 struct AVResampleContext *resample_context; 00048 short *temp[MAX_CHANNELS]; 00049 int temp_len; 00050 float ratio; 00051 /* channel convert */ 00052 int input_channels, output_channels, filter_channels; 00053 AVAudioConvert *convert_ctx[2]; 00054 enum AVSampleFormat sample_fmt[2]; 00055 unsigned sample_size[2]; 00056 short *buffer[2]; 00057 unsigned buffer_size[2]; 00058 }; 00059 00060 /* n1: number of samples */ 00061 static void stereo_to_mono(short *output, short *input, int n1) 00062 { 00063 short *p, *q; 00064 int n = n1; 00065 00066 p = input; 00067 q = output; 00068 while (n >= 4) { 00069 q[0] = (p[0] + p[1]) >> 1; 00070 q[1] = (p[2] + p[3]) >> 1; 00071 q[2] = (p[4] + p[5]) >> 1; 00072 q[3] = (p[6] + p[7]) >> 1; 00073 q += 4; 00074 p += 8; 00075 n -= 4; 00076 } 00077 while (n > 0) { 00078 q[0] = (p[0] + p[1]) >> 1; 00079 q++; 00080 p += 2; 00081 n--; 00082 } 00083 } 00084 00085 /* n1: number of samples */ 00086 static void mono_to_stereo(short *output, short *input, int n1) 00087 { 00088 short *p, *q; 00089 int n = n1; 00090 int v; 00091 00092 p = input; 00093 q = output; 00094 while (n >= 4) { 00095 v = p[0]; q[0] = v; q[1] = v; 00096 v = p[1]; q[2] = v; q[3] = v; 00097 v = p[2]; q[4] = v; q[5] = v; 00098 v = p[3]; q[6] = v; q[7] = v; 00099 q += 8; 00100 p += 4; 00101 n -= 4; 00102 } 00103 while (n > 0) { 00104 v = p[0]; q[0] = v; q[1] = v; 00105 q += 2; 00106 p += 1; 00107 n--; 00108 } 00109 } 00110 00111 static void deinterleave(short **output, short *input, int channels, int samples) 00112 { 00113 int i, j; 00114 00115 for (i = 0; i < samples; i++) { 00116 for (j = 0; j < channels; j++) { 00117 *output[j]++ = *input++; 00118 } 00119 } 00120 } 00121 00122 static void interleave(short *output, short **input, int channels, int samples) 00123 { 00124 int i, j; 00125 00126 for (i = 0; i < samples; i++) { 00127 for (j = 0; j < channels; j++) { 00128 *output++ = *input[j]++; 00129 } 00130 } 00131 } 00132 00133 static void ac3_5p1_mux(short *output, short *input1, short *input2, int n) 00134 { 00135 int i; 00136 short l, r; 00137 00138 for (i = 0; i < n; i++) { 00139 l = *input1++; 00140 r = *input2++; 00141 *output++ = l; /* left */ 00142 *output++ = (l / 2) + (r / 2); /* center */ 00143 *output++ = r; /* right */ 00144 *output++ = 0; /* left surround */ 00145 *output++ = 0; /* right surroud */ 00146 *output++ = 0; /* low freq */ 00147 } 00148 } 00149 00150 ReSampleContext *av_audio_resample_init(int output_channels, int input_channels, 00151 int output_rate, int input_rate, 00152 enum AVSampleFormat sample_fmt_out, 00153 enum AVSampleFormat sample_fmt_in, 00154 int filter_length, int log2_phase_count, 00155 int linear, double cutoff) 00156 { 00157 ReSampleContext *s; 00158 00159 if (input_channels > MAX_CHANNELS) { 00160 av_log(NULL, AV_LOG_ERROR, 00161 "Resampling with input channels greater than %d is unsupported.\n", 00162 MAX_CHANNELS); 00163 return NULL; 00164 } 00165 if (output_channels > 2 && 00166 !(output_channels == 6 && input_channels == 2) && 00167 output_channels != input_channels) { 00168 av_log(NULL, AV_LOG_ERROR, 00169 "Resampling output channel count must be 1 or 2 for mono input; 1, 2 or 6 for stereo input; or N for N channel input.\n"); 00170 return NULL; 00171 } 00172 00173 s = av_mallocz(sizeof(ReSampleContext)); 00174 if (!s) { 00175 av_log(NULL, AV_LOG_ERROR, "Can't allocate memory for resample context.\n"); 00176 return NULL; 00177 } 00178 00179 s->ratio = (float)output_rate / (float)input_rate; 00180 00181 s->input_channels = input_channels; 00182 s->output_channels = output_channels; 00183 00184 s->filter_channels = s->input_channels; 00185 if (s->output_channels < s->filter_channels) 00186 s->filter_channels = s->output_channels; 00187 00188 s->sample_fmt[0] = sample_fmt_in; 00189 s->sample_fmt[1] = sample_fmt_out; 00190 s->sample_size[0] = av_get_bits_per_sample_fmt(s->sample_fmt[0]) >> 3; 00191 s->sample_size[1] = av_get_bits_per_sample_fmt(s->sample_fmt[1]) >> 3; 00192 00193 if (s->sample_fmt[0] != AV_SAMPLE_FMT_S16) { 00194 if (!(s->convert_ctx[0] = av_audio_convert_alloc(AV_SAMPLE_FMT_S16, 1, 00195 s->sample_fmt[0], 1, NULL, 0))) { 00196 av_log(s, AV_LOG_ERROR, 00197 "Cannot convert %s sample format to s16 sample format\n", 00198 av_get_sample_fmt_name(s->sample_fmt[0])); 00199 av_free(s); 00200 return NULL; 00201 } 00202 } 00203 00204 if (s->sample_fmt[1] != AV_SAMPLE_FMT_S16) { 00205 if (!(s->convert_ctx[1] = av_audio_convert_alloc(s->sample_fmt[1], 1, 00206 AV_SAMPLE_FMT_S16, 1, NULL, 0))) { 00207 av_log(s, AV_LOG_ERROR, 00208 "Cannot convert s16 sample format to %s sample format\n", 00209 av_get_sample_fmt_name(s->sample_fmt[1])); 00210 av_audio_convert_free(s->convert_ctx[0]); 00211 av_free(s); 00212 return NULL; 00213 } 00214 } 00215 00216 #define TAPS 16 00217 s->resample_context = av_resample_init(output_rate, input_rate, 00218 filter_length, log2_phase_count, 00219 linear, cutoff); 00220 00221 *(const AVClass**)s->resample_context = &audioresample_context_class; 00222 00223 return s; 00224 } 00225 00226 /* resample audio. 'nb_samples' is the number of input samples */ 00227 /* XXX: optimize it ! */ 00228 int audio_resample(ReSampleContext *s, short *output, short *input, int nb_samples) 00229 { 00230 int i, nb_samples1; 00231 short *bufin[MAX_CHANNELS]; 00232 short *bufout[MAX_CHANNELS]; 00233 short *buftmp2[MAX_CHANNELS], *buftmp3[MAX_CHANNELS]; 00234 short *output_bak = NULL; 00235 int lenout; 00236 00237 if (s->input_channels == s->output_channels && s->ratio == 1.0 && 0) { 00238 /* nothing to do */ 00239 memcpy(output, input, nb_samples * s->input_channels * sizeof(short)); 00240 return nb_samples; 00241 } 00242 00243 if (s->sample_fmt[0] != AV_SAMPLE_FMT_S16) { 00244 int istride[1] = { s->sample_size[0] }; 00245 int ostride[1] = { 2 }; 00246 const void *ibuf[1] = { input }; 00247 void *obuf[1]; 00248 unsigned input_size = nb_samples * s->input_channels * 2; 00249 00250 if (!s->buffer_size[0] || s->buffer_size[0] < input_size) { 00251 av_free(s->buffer[0]); 00252 s->buffer_size[0] = input_size; 00253 s->buffer[0] = av_malloc(s->buffer_size[0]); 00254 if (!s->buffer[0]) { 00255 av_log(s->resample_context, AV_LOG_ERROR, "Could not allocate buffer\n"); 00256 return 0; 00257 } 00258 } 00259 00260 obuf[0] = s->buffer[0]; 00261 00262 if (av_audio_convert(s->convert_ctx[0], obuf, ostride, 00263 ibuf, istride, nb_samples * s->input_channels) < 0) { 00264 av_log(s->resample_context, AV_LOG_ERROR, 00265 "Audio sample format conversion failed\n"); 00266 return 0; 00267 } 00268 00269 input = s->buffer[0]; 00270 } 00271 00272 lenout = 4 * nb_samples * s->ratio + 16; 00273 00274 if (s->sample_fmt[1] != AV_SAMPLE_FMT_S16) { 00275 output_bak = output; 00276 00277 if (!s->buffer_size[1] || s->buffer_size[1] < lenout) { 00278 av_free(s->buffer[1]); 00279 s->buffer_size[1] = lenout; 00280 s->buffer[1] = av_malloc(s->buffer_size[1]); 00281 if (!s->buffer[1]) { 00282 av_log(s->resample_context, AV_LOG_ERROR, "Could not allocate buffer\n"); 00283 return 0; 00284 } 00285 } 00286 00287 output = s->buffer[1]; 00288 } 00289 00290 /* XXX: move those malloc to resample init code */ 00291 for (i = 0; i < s->filter_channels; i++) { 00292 bufin[i] = av_malloc((nb_samples + s->temp_len) * sizeof(short)); 00293 memcpy(bufin[i], s->temp[i], s->temp_len * sizeof(short)); 00294 buftmp2[i] = bufin[i] + s->temp_len; 00295 bufout[i] = av_malloc(lenout * sizeof(short)); 00296 } 00297 00298 if (s->input_channels == 2 && s->output_channels == 1) { 00299 buftmp3[0] = output; 00300 stereo_to_mono(buftmp2[0], input, nb_samples); 00301 } else if (s->output_channels >= 2 && s->input_channels == 1) { 00302 buftmp3[0] = bufout[0]; 00303 memcpy(buftmp2[0], input, nb_samples * sizeof(short)); 00304 } else if (s->output_channels >= s->input_channels && s->input_channels >= 2) { 00305 for (i = 0; i < s->input_channels; i++) { 00306 buftmp3[i] = bufout[i]; 00307 } 00308 deinterleave(buftmp2, input, s->input_channels, nb_samples); 00309 } else { 00310 buftmp3[0] = output; 00311 memcpy(buftmp2[0], input, nb_samples * sizeof(short)); 00312 } 00313 00314 nb_samples += s->temp_len; 00315 00316 /* resample each channel */ 00317 nb_samples1 = 0; /* avoid warning */ 00318 for (i = 0; i < s->filter_channels; i++) { 00319 int consumed; 00320 int is_last = i + 1 == s->filter_channels; 00321 00322 nb_samples1 = av_resample(s->resample_context, buftmp3[i], bufin[i], 00323 &consumed, nb_samples, lenout, is_last); 00324 s->temp_len = nb_samples - consumed; 00325 s->temp[i] = av_realloc(s->temp[i], s->temp_len * sizeof(short)); 00326 memcpy(s->temp[i], bufin[i] + consumed, s->temp_len * sizeof(short)); 00327 } 00328 00329 if (s->output_channels == 2 && s->input_channels == 1) { 00330 mono_to_stereo(output, buftmp3[0], nb_samples1); 00331 } else if (s->output_channels == 6 && s->input_channels == 2) { 00332 ac3_5p1_mux(output, buftmp3[0], buftmp3[1], nb_samples1); 00333 } else if (s->output_channels == s->input_channels && s->input_channels >= 2) { 00334 interleave(output, buftmp3, s->output_channels, nb_samples1); 00335 } 00336 00337 if (s->sample_fmt[1] != AV_SAMPLE_FMT_S16) { 00338 int istride[1] = { 2 }; 00339 int ostride[1] = { s->sample_size[1] }; 00340 const void *ibuf[1] = { output }; 00341 void *obuf[1] = { output_bak }; 00342 00343 if (av_audio_convert(s->convert_ctx[1], obuf, ostride, 00344 ibuf, istride, nb_samples1 * s->output_channels) < 0) { 00345 av_log(s->resample_context, AV_LOG_ERROR, 00346 "Audio sample format convertion failed\n"); 00347 return 0; 00348 } 00349 } 00350 00351 for (i = 0; i < s->filter_channels; i++) { 00352 av_free(bufin[i]); 00353 av_free(bufout[i]); 00354 } 00355 00356 return nb_samples1; 00357 } 00358 00359 void audio_resample_close(ReSampleContext *s) 00360 { 00361 int i; 00362 av_resample_close(s->resample_context); 00363 for (i = 0; i < s->filter_channels; i++) 00364 av_freep(&s->temp[i]); 00365 av_freep(&s->buffer[0]); 00366 av_freep(&s->buffer[1]); 00367 av_audio_convert_free(s->convert_ctx[0]); 00368 av_audio_convert_free(s->convert_ctx[1]); 00369 av_free(s); 00370 }