FFmpeg
 All Data Structures Namespaces Files Functions Variables Typedefs Enumerations Enumerator Macros Groups Pages
af_dynaudnorm.c
Go to the documentation of this file.
1 /*
2  * Dynamic Audio Normalizer
3  * Copyright (c) 2015 LoRd_MuldeR <mulder2@gmx.de>. Some rights reserved.
4  *
5  * This file is part of FFmpeg.
6  *
7  * FFmpeg is free software; you can redistribute it and/or
8  * modify it under the terms of the GNU Lesser General Public
9  * License as published by the Free Software Foundation; either
10  * version 2.1 of the License, or (at your option) any later version.
11  *
12  * FFmpeg is distributed in the hope that it will be useful,
13  * but WITHOUT ANY WARRANTY; without even the implied warranty of
14  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15  * Lesser General Public License for more details.
16  *
17  * You should have received a copy of the GNU Lesser General Public
18  * License along with FFmpeg; if not, write to the Free Software
19  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20  */
21 
22 /**
23  * @file
24  * Dynamic Audio Normalizer
25  */
26 
27 #include <float.h>
28 
29 #include "libavutil/avassert.h"
30 #include "libavutil/opt.h"
31 
32 #define FF_BUFQUEUE_SIZE 302
34 
35 #include "audio.h"
36 #include "avfilter.h"
37 #include "internal.h"
38 
39 typedef struct cqueue {
40  double *elements;
41  int size;
43  int first;
44 } cqueue;
45 
47  const AVClass *class;
48 
49  struct FFBufQueue queue;
50 
51  int frame_len;
57 
58  double peak_value;
60  double target_rms;
65  double *fade_factors[2];
66  double *weights;
67 
68  int channels;
69  int delay;
70 
75 
76 #define OFFSET(x) offsetof(DynamicAudioNormalizerContext, x)
77 #define FLAGS AV_OPT_FLAG_AUDIO_PARAM|AV_OPT_FLAG_FILTERING_PARAM
78 
79 static const AVOption dynaudnorm_options[] = {
80  { "f", "set the frame length in msec", OFFSET(frame_len_msec), AV_OPT_TYPE_INT, {.i64 = 500}, 10, 8000, FLAGS },
81  { "g", "set the filter size", OFFSET(filter_size), AV_OPT_TYPE_INT, {.i64 = 31}, 3, 301, FLAGS },
82  { "p", "set the peak value", OFFSET(peak_value), AV_OPT_TYPE_DOUBLE, {.dbl = 0.95}, 0.0, 1.0, FLAGS },
83  { "m", "set the max amplification", OFFSET(max_amplification), AV_OPT_TYPE_DOUBLE, {.dbl = 10.0}, 1.0, 100.0, FLAGS },
84  { "r", "set the target RMS", OFFSET(target_rms), AV_OPT_TYPE_DOUBLE, {.dbl = 0.0}, 0.0, 1.0, FLAGS },
85  { "n", "set channel coupling", OFFSET(channels_coupled), AV_OPT_TYPE_BOOL, {.i64 = 1}, 0, 1, FLAGS },
86  { "c", "set DC correction", OFFSET(dc_correction), AV_OPT_TYPE_BOOL, {.i64 = 0}, 0, 1, FLAGS },
87  { "b", "set alternative boundary mode", OFFSET(alt_boundary_mode), AV_OPT_TYPE_BOOL, {.i64 = 0}, 0, 1, FLAGS },
88  { "s", "set the compress factor", OFFSET(compress_factor), AV_OPT_TYPE_DOUBLE, {.dbl = 0.0}, 0.0, 30.0, FLAGS },
89  { NULL }
90 };
91 
92 AVFILTER_DEFINE_CLASS(dynaudnorm);
93 
95 {
97 
98  if (!(s->filter_size & 1)) {
99  av_log(ctx, AV_LOG_ERROR, "filter size %d is invalid. Must be an odd value.\n", s->filter_size);
100  return AVERROR(EINVAL);
101  }
102 
103  return 0;
104 }
105 
107 {
110  static const enum AVSampleFormat sample_fmts[] = {
113  };
114  int ret;
115 
116  layouts = ff_all_channel_counts();
117  if (!layouts)
118  return AVERROR(ENOMEM);
119  ret = ff_set_common_channel_layouts(ctx, layouts);
120  if (ret < 0)
121  return ret;
122 
123  formats = ff_make_format_list(sample_fmts);
124  if (!formats)
125  return AVERROR(ENOMEM);
126  ret = ff_set_common_formats(ctx, formats);
127  if (ret < 0)
128  return ret;
129 
130  formats = ff_all_samplerates();
131  if (!formats)
132  return AVERROR(ENOMEM);
133  return ff_set_common_samplerates(ctx, formats);
134 }
135 
136 static inline int frame_size(int sample_rate, int frame_len_msec)
137 {
138  const int frame_size = lrint((double)sample_rate * (frame_len_msec / 1000.0));
139  return frame_size + (frame_size % 2);
140 }
141 
142 static void precalculate_fade_factors(double *fade_factors[2], int frame_len)
143 {
144  const double step_size = 1.0 / frame_len;
145  int pos;
146 
147  for (pos = 0; pos < frame_len; pos++) {
148  fade_factors[0][pos] = 1.0 - (step_size * (pos + 1.0));
149  fade_factors[1][pos] = 1.0 - fade_factors[0][pos];
150  }
151 }
152 
154 {
155  cqueue *q;
156 
157  q = av_malloc(sizeof(cqueue));
158  if (!q)
159  return NULL;
160 
161  q->size = size;
162  q->nb_elements = 0;
163  q->first = 0;
164 
165  q->elements = av_malloc_array(size, sizeof(double));
166  if (!q->elements) {
167  av_free(q);
168  return NULL;
169  }
170 
171  return q;
172 }
173 
174 static void cqueue_free(cqueue *q)
175 {
176  if (q)
177  av_free(q->elements);
178  av_free(q);
179 }
180 
181 static int cqueue_size(cqueue *q)
182 {
183  return q->nb_elements;
184 }
185 
186 static int cqueue_empty(cqueue *q)
187 {
188  return !q->nb_elements;
189 }
190 
191 static int cqueue_enqueue(cqueue *q, double element)
192 {
193  int i;
194 
195  av_assert2(q->nb_elements != q->size);
196 
197  i = (q->first + q->nb_elements) % q->size;
198  q->elements[i] = element;
199  q->nb_elements++;
200 
201  return 0;
202 }
203 
204 static double cqueue_peek(cqueue *q, int index)
205 {
206  av_assert2(index < q->nb_elements);
207  return q->elements[(q->first + index) % q->size];
208 }
209 
210 static int cqueue_dequeue(cqueue *q, double *element)
211 {
213 
214  *element = q->elements[q->first];
215  q->first = (q->first + 1) % q->size;
216  q->nb_elements--;
217 
218  return 0;
219 }
220 
221 static int cqueue_pop(cqueue *q)
222 {
224 
225  q->first = (q->first + 1) % q->size;
226  q->nb_elements--;
227 
228  return 0;
229 }
230 
232 {
233  double total_weight = 0.0;
234  const double sigma = (((s->filter_size / 2.0) - 1.0) / 3.0) + (1.0 / 3.0);
235  double adjust;
236  int i;
237 
238  // Pre-compute constants
239  const int offset = s->filter_size / 2;
240  const double c1 = 1.0 / (sigma * sqrt(2.0 * M_PI));
241  const double c2 = 2.0 * sigma * sigma;
242 
243  // Compute weights
244  for (i = 0; i < s->filter_size; i++) {
245  const int x = i - offset;
246 
247  s->weights[i] = c1 * exp(-x * x / c2);
248  total_weight += s->weights[i];
249  }
250 
251  // Adjust weights
252  adjust = 1.0 / total_weight;
253  for (i = 0; i < s->filter_size; i++) {
254  s->weights[i] *= adjust;
255  }
256 }
257 
259 {
261  int c;
262 
266  av_freep(&s->fade_factors[0]);
267  av_freep(&s->fade_factors[1]);
268 
269  for (c = 0; c < s->channels; c++) {
270  if (s->gain_history_original)
272  if (s->gain_history_minimum)
274  if (s->gain_history_smoothed)
276  }
277 
281 
282  av_freep(&s->weights);
283 
285 }
286 
287 static int config_input(AVFilterLink *inlink)
288 {
289  AVFilterContext *ctx = inlink->dst;
291  int c;
292 
293  uninit(ctx);
294 
295  s->frame_len =
296  inlink->min_samples =
297  inlink->max_samples =
298  inlink->partial_buf_size = frame_size(inlink->sample_rate, s->frame_len_msec);
299  av_log(ctx, AV_LOG_DEBUG, "frame len %d\n", s->frame_len);
300 
301  s->fade_factors[0] = av_malloc_array(s->frame_len, sizeof(*s->fade_factors[0]));
302  s->fade_factors[1] = av_malloc_array(s->frame_len, sizeof(*s->fade_factors[1]));
303 
305  s->dc_correction_value = av_calloc(inlink->channels, sizeof(*s->dc_correction_value));
306  s->compress_threshold = av_calloc(inlink->channels, sizeof(*s->compress_threshold));
308  s->gain_history_minimum = av_calloc(inlink->channels, sizeof(*s->gain_history_minimum));
310  s->weights = av_malloc_array(s->filter_size, sizeof(*s->weights));
312  !s->compress_threshold || !s->fade_factors[0] || !s->fade_factors[1] ||
314  !s->gain_history_smoothed || !s->weights)
315  return AVERROR(ENOMEM);
316 
317  for (c = 0; c < inlink->channels; c++) {
318  s->prev_amplification_factor[c] = 1.0;
319 
323 
324  if (!s->gain_history_original[c] || !s->gain_history_minimum[c] ||
325  !s->gain_history_smoothed[c])
326  return AVERROR(ENOMEM);
327  }
328 
331 
332  s->channels = inlink->channels;
333  s->delay = s->filter_size;
334 
335  return 0;
336 }
337 
338 static inline double fade(double prev, double next, int pos,
339  double *fade_factors[2])
340 {
341  return fade_factors[0][pos] * prev + fade_factors[1][pos] * next;
342 }
343 
344 static inline double pow_2(const double value)
345 {
346  return value * value;
347 }
348 
349 static inline double bound(const double threshold, const double val)
350 {
351  const double CONST = 0.8862269254527580136490837416705725913987747280611935; //sqrt(PI) / 2.0
352  return erf(CONST * (val / threshold)) * threshold;
353 }
354 
356 {
357  double max = DBL_EPSILON;
358  int c, i;
359 
360  if (channel == -1) {
361  for (c = 0; c < frame->channels; c++) {
362  double *data_ptr = (double *)frame->extended_data[c];
363 
364  for (i = 0; i < frame->nb_samples; i++)
365  max = FFMAX(max, fabs(data_ptr[i]));
366  }
367  } else {
368  double *data_ptr = (double *)frame->extended_data[channel];
369 
370  for (i = 0; i < frame->nb_samples; i++)
371  max = FFMAX(max, fabs(data_ptr[i]));
372  }
373 
374  return max;
375 }
376 
378 {
379  double rms_value = 0.0;
380  int c, i;
381 
382  if (channel == -1) {
383  for (c = 0; c < frame->channels; c++) {
384  const double *data_ptr = (double *)frame->extended_data[c];
385 
386  for (i = 0; i < frame->nb_samples; i++) {
387  rms_value += pow_2(data_ptr[i]);
388  }
389  }
390 
391  rms_value /= frame->nb_samples * frame->channels;
392  } else {
393  const double *data_ptr = (double *)frame->extended_data[channel];
394  for (i = 0; i < frame->nb_samples; i++) {
395  rms_value += pow_2(data_ptr[i]);
396  }
397 
398  rms_value /= frame->nb_samples;
399  }
400 
401  return FFMAX(sqrt(rms_value), DBL_EPSILON);
402 }
403 
405  int channel)
406 {
407  const double maximum_gain = s->peak_value / find_peak_magnitude(frame, channel);
408  const double rms_gain = s->target_rms > DBL_EPSILON ? (s->target_rms / compute_frame_rms(frame, channel)) : DBL_MAX;
409  return bound(s->max_amplification, FFMIN(maximum_gain, rms_gain));
410 }
411 
412 static double minimum_filter(cqueue *q)
413 {
414  double min = DBL_MAX;
415  int i;
416 
417  for (i = 0; i < cqueue_size(q); i++) {
418  min = FFMIN(min, cqueue_peek(q, i));
419  }
420 
421  return min;
422 }
423 
425 {
426  double result = 0.0;
427  int i;
428 
429  for (i = 0; i < cqueue_size(q); i++) {
430  result += cqueue_peek(q, i) * s->weights[i];
431  }
432 
433  return result;
434 }
435 
437  double current_gain_factor)
438 {
439  if (cqueue_empty(s->gain_history_original[channel]) ||
440  cqueue_empty(s->gain_history_minimum[channel])) {
441  const int pre_fill_size = s->filter_size / 2;
442  const double initial_value = s->alt_boundary_mode ? current_gain_factor : 1.0;
443 
444  s->prev_amplification_factor[channel] = initial_value;
445 
446  while (cqueue_size(s->gain_history_original[channel]) < pre_fill_size) {
447  cqueue_enqueue(s->gain_history_original[channel], initial_value);
448  }
449  }
450 
451  cqueue_enqueue(s->gain_history_original[channel], current_gain_factor);
452 
453  while (cqueue_size(s->gain_history_original[channel]) >= s->filter_size) {
454  double minimum;
456 
457  if (cqueue_empty(s->gain_history_minimum[channel])) {
458  const int pre_fill_size = s->filter_size / 2;
459  double initial_value = s->alt_boundary_mode ? cqueue_peek(s->gain_history_original[channel], 0) : 1.0;
460  int input = pre_fill_size;
461 
462  while (cqueue_size(s->gain_history_minimum[channel]) < pre_fill_size) {
463  input++;
464  initial_value = FFMIN(initial_value, cqueue_peek(s->gain_history_original[channel], input));
465  cqueue_enqueue(s->gain_history_minimum[channel], initial_value);
466  }
467  }
468 
469  minimum = minimum_filter(s->gain_history_original[channel]);
470 
471  cqueue_enqueue(s->gain_history_minimum[channel], minimum);
472 
473  cqueue_pop(s->gain_history_original[channel]);
474  }
475 
476  while (cqueue_size(s->gain_history_minimum[channel]) >= s->filter_size) {
477  double smoothed;
479  smoothed = gaussian_filter(s, s->gain_history_minimum[channel]);
480 
481  cqueue_enqueue(s->gain_history_smoothed[channel], smoothed);
482 
483  cqueue_pop(s->gain_history_minimum[channel]);
484  }
485 }
486 
487 static inline double update_value(double new, double old, double aggressiveness)
488 {
489  av_assert0((aggressiveness >= 0.0) && (aggressiveness <= 1.0));
490  return aggressiveness * new + (1.0 - aggressiveness) * old;
491 }
492 
494 {
495  const double diff = 1.0 / frame->nb_samples;
496  int is_first_frame = cqueue_empty(s->gain_history_original[0]);
497  int c, i;
498 
499  for (c = 0; c < s->channels; c++) {
500  double *dst_ptr = (double *)frame->extended_data[c];
501  double current_average_value = 0.0;
502  double prev_value;
503 
504  for (i = 0; i < frame->nb_samples; i++)
505  current_average_value += dst_ptr[i] * diff;
506 
507  prev_value = is_first_frame ? current_average_value : s->dc_correction_value[c];
508  s->dc_correction_value[c] = is_first_frame ? current_average_value : update_value(current_average_value, s->dc_correction_value[c], 0.1);
509 
510  for (i = 0; i < frame->nb_samples; i++) {
511  dst_ptr[i] -= fade(prev_value, s->dc_correction_value[c], i, s->fade_factors);
512  }
513  }
514 }
515 
516 static double setup_compress_thresh(double threshold)
517 {
518  if ((threshold > DBL_EPSILON) && (threshold < (1.0 - DBL_EPSILON))) {
519  double current_threshold = threshold;
520  double step_size = 1.0;
521 
522  while (step_size > DBL_EPSILON) {
523  while ((llrint((current_threshold + step_size) * (UINT64_C(1) << 63)) >
524  llrint(current_threshold * (UINT64_C(1) << 63))) &&
525  (bound(current_threshold + step_size, 1.0) <= threshold)) {
526  current_threshold += step_size;
527  }
528 
529  step_size /= 2.0;
530  }
531 
532  return current_threshold;
533  } else {
534  return threshold;
535  }
536 }
537 
539  AVFrame *frame, int channel)
540 {
541  double variance = 0.0;
542  int i, c;
543 
544  if (channel == -1) {
545  for (c = 0; c < s->channels; c++) {
546  const double *data_ptr = (double *)frame->extended_data[c];
547 
548  for (i = 0; i < frame->nb_samples; i++) {
549  variance += pow_2(data_ptr[i]); // Assume that MEAN is *zero*
550  }
551  }
552  variance /= (s->channels * frame->nb_samples) - 1;
553  } else {
554  const double *data_ptr = (double *)frame->extended_data[channel];
555 
556  for (i = 0; i < frame->nb_samples; i++) {
557  variance += pow_2(data_ptr[i]); // Assume that MEAN is *zero*
558  }
559  variance /= frame->nb_samples - 1;
560  }
561 
562  return FFMAX(sqrt(variance), DBL_EPSILON);
563 }
564 
566 {
567  int is_first_frame = cqueue_empty(s->gain_history_original[0]);
568  int c, i;
569 
570  if (s->channels_coupled) {
571  const double standard_deviation = compute_frame_std_dev(s, frame, -1);
572  const double current_threshold = FFMIN(1.0, s->compress_factor * standard_deviation);
573 
574  const double prev_value = is_first_frame ? current_threshold : s->compress_threshold[0];
575  double prev_actual_thresh, curr_actual_thresh;
576  s->compress_threshold[0] = is_first_frame ? current_threshold : update_value(current_threshold, s->compress_threshold[0], (1.0/3.0));
577 
578  prev_actual_thresh = setup_compress_thresh(prev_value);
579  curr_actual_thresh = setup_compress_thresh(s->compress_threshold[0]);
580 
581  for (c = 0; c < s->channels; c++) {
582  double *const dst_ptr = (double *)frame->extended_data[c];
583  for (i = 0; i < frame->nb_samples; i++) {
584  const double localThresh = fade(prev_actual_thresh, curr_actual_thresh, i, s->fade_factors);
585  dst_ptr[i] = copysign(bound(localThresh, fabs(dst_ptr[i])), dst_ptr[i]);
586  }
587  }
588  } else {
589  for (c = 0; c < s->channels; c++) {
590  const double standard_deviation = compute_frame_std_dev(s, frame, c);
591  const double current_threshold = setup_compress_thresh(FFMIN(1.0, s->compress_factor * standard_deviation));
592 
593  const double prev_value = is_first_frame ? current_threshold : s->compress_threshold[c];
594  double prev_actual_thresh, curr_actual_thresh;
595  double *dst_ptr;
596  s->compress_threshold[c] = is_first_frame ? current_threshold : update_value(current_threshold, s->compress_threshold[c], 1.0/3.0);
597 
598  prev_actual_thresh = setup_compress_thresh(prev_value);
599  curr_actual_thresh = setup_compress_thresh(s->compress_threshold[c]);
600 
601  dst_ptr = (double *)frame->extended_data[c];
602  for (i = 0; i < frame->nb_samples; i++) {
603  const double localThresh = fade(prev_actual_thresh, curr_actual_thresh, i, s->fade_factors);
604  dst_ptr[i] = copysign(bound(localThresh, fabs(dst_ptr[i])), dst_ptr[i]);
605  }
606  }
607  }
608 }
609 
611 {
612  if (s->dc_correction) {
613  perform_dc_correction(s, frame);
614  }
615 
616  if (s->compress_factor > DBL_EPSILON) {
617  perform_compression(s, frame);
618  }
619 
620  if (s->channels_coupled) {
621  const double current_gain_factor = get_max_local_gain(s, frame, -1);
622  int c;
623 
624  for (c = 0; c < s->channels; c++)
625  update_gain_history(s, c, current_gain_factor);
626  } else {
627  int c;
628 
629  for (c = 0; c < s->channels; c++)
630  update_gain_history(s, c, get_max_local_gain(s, frame, c));
631  }
632 }
633 
635 {
636  int c, i;
637 
638  for (c = 0; c < s->channels; c++) {
639  double *dst_ptr = (double *)frame->extended_data[c];
640  double current_amplification_factor;
641 
642  cqueue_dequeue(s->gain_history_smoothed[c], &current_amplification_factor);
643 
644  for (i = 0; i < frame->nb_samples; i++) {
645  const double amplification_factor = fade(s->prev_amplification_factor[c],
646  current_amplification_factor, i,
647  s->fade_factors);
648 
649  dst_ptr[i] *= amplification_factor;
650 
651  if (fabs(dst_ptr[i]) > s->peak_value)
652  dst_ptr[i] = copysign(s->peak_value, dst_ptr[i]);
653  }
654 
655  s->prev_amplification_factor[c] = current_amplification_factor;
656  }
657 }
658 
659 static int filter_frame(AVFilterLink *inlink, AVFrame *in)
660 {
661  AVFilterContext *ctx = inlink->dst;
663  AVFilterLink *outlink = inlink->dst->outputs[0];
664  int ret = 0;
665 
666  if (!cqueue_empty(s->gain_history_smoothed[0])) {
668 
669  amplify_frame(s, out);
670  ret = ff_filter_frame(outlink, out);
671  }
672 
673  analyze_frame(s, in);
674  ff_bufqueue_add(ctx, &s->queue, in);
675 
676  return ret;
677 }
678 
680  AVFilterLink *outlink)
681 {
682  AVFrame *out = ff_get_audio_buffer(outlink, s->frame_len);
683  int c, i;
684 
685  if (!out)
686  return AVERROR(ENOMEM);
687 
688  for (c = 0; c < s->channels; c++) {
689  double *dst_ptr = (double *)out->extended_data[c];
690 
691  for (i = 0; i < out->nb_samples; i++) {
692  dst_ptr[i] = s->alt_boundary_mode ? DBL_EPSILON : ((s->target_rms > DBL_EPSILON) ? FFMIN(s->peak_value, s->target_rms) : s->peak_value);
693  if (s->dc_correction) {
694  dst_ptr[i] *= ((i % 2) == 1) ? -1 : 1;
695  dst_ptr[i] += s->dc_correction_value[c];
696  }
697  }
698  }
699 
700  s->delay--;
701  return filter_frame(inlink, out);
702 }
703 
704 static int request_frame(AVFilterLink *outlink)
705 {
706  AVFilterContext *ctx = outlink->src;
708  int ret = 0;
709 
710  ret = ff_request_frame(ctx->inputs[0]);
711 
712  if (ret == AVERROR_EOF && !ctx->is_disabled && s->delay) {
713  if (!cqueue_empty(s->gain_history_smoothed[0])) {
714  ret = flush_buffer(s, ctx->inputs[0], outlink);
715  } else if (s->queue.available) {
717 
718  ret = ff_filter_frame(outlink, out);
719  }
720  }
721 
722  return ret;
723 }
724 
726  {
727  .name = "default",
728  .type = AVMEDIA_TYPE_AUDIO,
729  .filter_frame = filter_frame,
730  .config_props = config_input,
731  .needs_writable = 1,
732  },
733  { NULL }
734 };
735 
737  {
738  .name = "default",
739  .type = AVMEDIA_TYPE_AUDIO,
740  .request_frame = request_frame,
741  },
742  { NULL }
743 };
744 
746  .name = "dynaudnorm",
747  .description = NULL_IF_CONFIG_SMALL("Dynamic Audio Normalizer."),
748  .query_formats = query_formats,
749  .priv_size = sizeof(DynamicAudioNormalizerContext),
750  .init = init,
751  .uninit = uninit,
752  .inputs = avfilter_af_dynaudnorm_inputs,
753  .outputs = avfilter_af_dynaudnorm_outputs,
754  .priv_class = &dynaudnorm_class,
755 };
static AVFrame * ff_bufqueue_get(struct FFBufQueue *queue)
Get the first buffer from the queue and remove it.
Definition: bufferqueue.h:98
static const AVFilterPad avfilter_af_dynaudnorm_inputs[]
#define FLAGS
Definition: af_dynaudnorm.c:77
static double bound(const double threshold, const double val)
#define NULL
Definition: coverity.c:32
int ff_set_common_channel_layouts(AVFilterContext *ctx, AVFilterChannelLayouts *layouts)
A helper for query_formats() which sets all links to the same list of channel layouts/sample rates...
Definition: formats.c:549
const char const char void * val
Definition: avisynth_c.h:771
static double compute_frame_rms(AVFrame *frame, int channel)
This structure describes decoded (raw) audio or video data.
Definition: frame.h:226
AVOption.
Definition: opt.h:246
#define CONST(name, help, val, unit)
Definition: vf_bwdif.c:534
static int cqueue_empty(cqueue *q)
static const AVFilterPad avfilter_af_dynaudnorm_outputs[]
static double pow_2(const double value)
static double erf(double z)
erf function Algorithm taken from the Boost project, source: http://www.boost.org/doc/libs/1_46_1/boo...
Definition: libm.h:121
Main libavfilter public API header.
static int cqueue_size(cqueue *q)
int first
Definition: af_dynaudnorm.c:43
double, planar
Definition: samplefmt.h:70
static int filter_frame(AVFilterLink *inlink, AVFrame *in)
static double get_max_local_gain(DynamicAudioNormalizerContext *s, AVFrame *frame, int channel)
static void analyze_frame(DynamicAudioNormalizerContext *s, AVFrame *frame)
static void precalculate_fade_factors(double *fade_factors[2], int frame_len)
int is_disabled
the enabled state from the last expression evaluation
Definition: avfilter.h:385
static int request_frame(AVFilterLink *outlink)
void * av_calloc(size_t nmemb, size_t size)
Non-inlined equivalent of av_mallocz_array().
Definition: mem.c:244
static int config_input(AVFilterLink *inlink)
AVFilterFormats * ff_make_format_list(const int *fmts)
Create a list of supported formats.
Definition: formats.c:283
Structure holding the queue.
Definition: bufferqueue.h:49
const char * name
Pad name.
Definition: internal.h:60
AVFilterLink ** inputs
array of pointers to input links
Definition: avfilter.h:346
#define av_assert0(cond)
assert() equivalent, that is always enabled.
Definition: avassert.h:37
int ff_filter_frame(AVFilterLink *link, AVFrame *frame)
Send a frame of data to the next filter.
Definition: avfilter.c:1080
#define av_cold
Definition: attributes.h:82
#define av_malloc(s)
#define av_assert2(cond)
assert() equivalent, that does lie in speed critical code.
Definition: avassert.h:64
AVOptions.
static double gaussian_filter(DynamicAudioNormalizerContext *s, cqueue *q)
double * elements
Definition: af_dynaudnorm.c:40
static AVFrame * frame
static const uint64_t c1
Definition: murmur3.c:49
#define AVERROR_EOF
End of file.
Definition: error.h:55
ptrdiff_t size
Definition: opengl_enc.c:101
static av_cold void uninit(AVFilterContext *ctx)
static void cqueue_free(cqueue *q)
#define av_log(a,...)
A filter pad used for either input or output.
Definition: internal.h:54
static int flush_buffer(DynamicAudioNormalizerContext *s, AVFilterLink *inlink, AVFilterLink *outlink)
static int query_formats(AVFilterContext *ctx)
static double cqueue_peek(cqueue *q, int index)
#define AV_LOG_ERROR
Something went wrong and cannot losslessly be recovered.
Definition: log.h:176
int ff_set_common_formats(AVFilterContext *ctx, AVFilterFormats *formats)
A helper for query_formats() which sets all links to the same list of formats.
Definition: formats.c:568
static void init_gaussian_filter(DynamicAudioNormalizerContext *s)
AVFILTER_DEFINE_CLASS(dynaudnorm)
AVFrame * ff_get_audio_buffer(AVFilterLink *link, int nb_samples)
Request an audio samples buffer with a specific set of permissions.
Definition: audio.c:86
#define AVERROR(e)
Definition: error.h:43
#define NULL_IF_CONFIG_SMALL(x)
Return NULL if CONFIG_SMALL is true, otherwise the argument without modification. ...
Definition: internal.h:186
void * priv
private data for use by the filter
Definition: avfilter.h:353
#define AV_LOG_DEBUG
Stuff which is only useful for libav* developers.
Definition: log.h:197
simple assert() macros that are a bit more flexible than ISO C assert().
tuple adjust
Definition: normalize.py:25
#define OFFSET(x)
Definition: af_dynaudnorm.c:76
static const uint8_t offset[127][2]
Definition: vf_spp.c:92
#define FFMAX(a, b)
Definition: common.h:94
int8_t exp
Definition: eval.c:72
int channels
number of audio channels, only used for audio.
Definition: frame.h:531
#define FFMIN(a, b)
Definition: common.h:96
GLsizei GLboolean const GLfloat * value
Definition: opengl_enc.c:109
static void perform_dc_correction(DynamicAudioNormalizerContext *s, AVFrame *frame)
AVFormatContext * ctx
Definition: movenc.c:48
#define s(width, name)
Definition: cbs_vp9.c:257
int size
Definition: af_dynaudnorm.c:41
static void ff_bufqueue_discard_all(struct FFBufQueue *queue)
Unref and remove all buffers from the queue.
Definition: bufferqueue.h:111
static const AVFilterPad inputs[]
Definition: af_acontrast.c:193
static const AVFilterPad outputs[]
Definition: af_acontrast.c:203
A list of supported channel layouts.
Definition: formats.h:85
sample_rate
int nb_elements
Definition: af_dynaudnorm.c:42
AVFilter ff_af_dynaudnorm
static void update_gain_history(DynamicAudioNormalizerContext *s, int channel, double current_gain_factor)
AVSampleFormat
Audio sample formats.
Definition: samplefmt.h:58
unsigned short available
number of available buffers
Definition: bufferqueue.h:52
uint8_t pi<< 24) CONV_FUNC_GROUP(AV_SAMPLE_FMT_FLT, float, AV_SAMPLE_FMT_U8, uint8_t,(*(constuint8_t *) pi-0x80)*(1.0f/(1<< 7))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_DBL, double, AV_SAMPLE_FMT_U8, uint8_t,(*(constuint8_t *) pi-0x80)*(1.0/(1<< 7))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_U8, uint8_t, AV_SAMPLE_FMT_S16, int16_t,(*(constint16_t *) pi >>8)+0x80) CONV_FUNC_GROUP(AV_SAMPLE_FMT_FLT, float, AV_SAMPLE_FMT_S16, int16_t,*(constint16_t *) pi *(1.0f/(1<< 15))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_DBL, double, AV_SAMPLE_FMT_S16, int16_t,*(constint16_t *) pi *(1.0/(1<< 15))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_U8, uint8_t, AV_SAMPLE_FMT_S32, int32_t,(*(constint32_t *) pi >>24)+0x80) CONV_FUNC_GROUP(AV_SAMPLE_FMT_FLT, float, AV_SAMPLE_FMT_S32, int32_t,*(constint32_t *) pi *(1.0f/(1U<< 31))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_DBL, double, AV_SAMPLE_FMT_S32, int32_t,*(constint32_t *) pi *(1.0/(1U<< 31))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_U8, uint8_t, AV_SAMPLE_FMT_FLT, float, av_clip_uint8(lrintf(*(constfloat *) pi *(1<< 7))+0x80)) CONV_FUNC_GROUP(AV_SAMPLE_FMT_S16, int16_t, AV_SAMPLE_FMT_FLT, float, av_clip_int16(lrintf(*(constfloat *) pi *(1<< 15)))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_S32, int32_t, AV_SAMPLE_FMT_FLT, float, av_clipl_int32(llrintf(*(constfloat *) pi *(1U<< 31)))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_U8, uint8_t, AV_SAMPLE_FMT_DBL, double, av_clip_uint8(lrint(*(constdouble *) pi *(1<< 7))+0x80)) CONV_FUNC_GROUP(AV_SAMPLE_FMT_S16, int16_t, AV_SAMPLE_FMT_DBL, double, av_clip_int16(lrint(*(constdouble *) pi *(1<< 15)))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_S32, int32_t, AV_SAMPLE_FMT_DBL, double, av_clipl_int32(llrint(*(constdouble *) pi *(1U<< 31))))#defineSET_CONV_FUNC_GROUP(ofmt, ifmt) staticvoidset_generic_function(AudioConvert *ac){}voidff_audio_convert_free(AudioConvert **ac){if(!*ac) return;ff_dither_free(&(*ac) ->dc);av_freep(ac);}AudioConvert *ff_audio_convert_alloc(AVAudioResampleContext *avr, enumAVSampleFormatout_fmt, enumAVSampleFormatin_fmt, intchannels, intsample_rate, intapply_map){AudioConvert *ac;intin_planar, out_planar;ac=av_mallocz(sizeof(*ac));if(!ac) returnNULL;ac->avr=avr;ac->out_fmt=out_fmt;ac->in_fmt=in_fmt;ac->channels=channels;ac->apply_map=apply_map;if(avr->dither_method!=AV_RESAMPLE_DITHER_NONE &&av_get_packed_sample_fmt(out_fmt)==AV_SAMPLE_FMT_S16 &&av_get_bytes_per_sample(in_fmt)>2){ac->dc=ff_dither_alloc(avr, out_fmt, in_fmt, channels, sample_rate, apply_map);if(!ac->dc){av_free(ac);returnNULL;}returnac;}in_planar=ff_sample_fmt_is_planar(in_fmt, channels);out_planar=ff_sample_fmt_is_planar(out_fmt, channels);if(in_planar==out_planar){ac->func_type=CONV_FUNC_TYPE_FLAT;ac->planes=in_planar?ac->channels:1;}elseif(in_planar) ac->func_type=CONV_FUNC_TYPE_INTERLEAVE;elseac->func_type=CONV_FUNC_TYPE_DEINTERLEAVE;set_generic_function(ac);if(ARCH_AARCH64) ff_audio_convert_init_aarch64(ac);if(ARCH_ARM) ff_audio_convert_init_arm(ac);if(ARCH_X86) ff_audio_convert_init_x86(ac);returnac;}intff_audio_convert(AudioConvert *ac, AudioData *out, AudioData *in){intuse_generic=1;intlen=in->nb_samples;intp;if(ac->dc){av_log(ac->avr, AV_LOG_TRACE,"%dsamples-audio_convert:%sto%s(dithered)\n", len, av_get_sample_fmt_name(ac->in_fmt), av_get_sample_fmt_name(ac->out_fmt));returnff_convert_dither(ac-> in
static double compute_frame_std_dev(DynamicAudioNormalizerContext *s, AVFrame *frame, int channel)
#define llrint(x)
Definition: libm.h:394
static av_cold int init(AVFilterContext *ctx)
Definition: af_dynaudnorm.c:94
Describe the class of an AVClass context structure.
Definition: log.h:67
Filter definition.
Definition: avfilter.h:144
int index
Definition: gxfenc.c:89
const char * name
Filter name.
Definition: avfilter.h:148
static av_always_inline double copysign(double x, double y)
Definition: libm.h:68
static double setup_compress_thresh(double threshold)
AVFilterLink ** outputs
array of pointers to output links
Definition: avfilter.h:350
enum MovChannelLayoutTag * layouts
Definition: mov_chan.c:434
AVFilterFormats * ff_all_samplerates(void)
Definition: formats.c:395
static double find_peak_magnitude(AVFrame *frame, int channel)
static int cqueue_pop(cqueue *q)
static double c[64]
static double minimum_filter(cqueue *q)
channel
Use these values when setting the channel map with ebur128_set_channel().
Definition: ebur128.h:39
static const uint64_t c2
Definition: murmur3.c:50
static int cqueue_enqueue(cqueue *q, double element)
static double fade(double prev, double next, int pos, double *fade_factors[2])
static av_always_inline int diff(const uint32_t a, const uint32_t b)
#define av_free(p)
static double update_value(double new, double old, double aggressiveness)
A list of supported formats for one end of a filter link.
Definition: formats.h:64
#define lrint
Definition: tablegen.h:53
An instance of a filter.
Definition: avfilter.h:338
static enum AVSampleFormat sample_fmts[]
Definition: adpcmenc.c:701
FILE * out
Definition: movenc.c:54
#define av_freep(p)
#define M_PI
Definition: mathematics.h:52
static void ff_bufqueue_add(void *log, struct FFBufQueue *queue, AVFrame *buf)
Add a buffer to the queue.
Definition: bufferqueue.h:71
#define av_malloc_array(a, b)
int ff_request_frame(AVFilterLink *link)
Request an input frame from the filter at the other end of the link.
Definition: avfilter.c:407
formats
Definition: signature.h:48
static cqueue * cqueue_create(int size)
static const AVOption dynaudnorm_options[]
Definition: af_dynaudnorm.c:79
internal API functions
AVFilterChannelLayouts * ff_all_channel_counts(void)
Construct an AVFilterChannelLayouts coding for any channel layout, with known or unknown disposition...
Definition: formats.c:410
static int cqueue_dequeue(cqueue *q, double *element)
uint8_t ** extended_data
pointers to the data planes/channels.
Definition: frame.h:273
float min
static int frame_size(int sample_rate, int frame_len_msec)
int nb_samples
number of audio samples (per channel) described by this frame
Definition: frame.h:292
for(j=16;j >0;--j)
int ff_set_common_samplerates(AVFilterContext *ctx, AVFilterFormats *samplerates)
Definition: formats.c:556
static void amplify_frame(DynamicAudioNormalizerContext *s, AVFrame *frame)
static void perform_compression(DynamicAudioNormalizerContext *s, AVFrame *frame)