FFmpeg coverage


Directory: ../../../ffmpeg/
File: src/libavfilter/af_astats.c
Date: 2024-04-19 17:50:32
Exec Total Coverage
Lines: 0 608 0.0%
Functions: 0 15 0.0%
Branches: 0 631 0.0%

Line Branch Exec Source
1 /*
2 * Copyright (c) 2009 Rob Sykes <robs@users.sourceforge.net>
3 * Copyright (c) 2013 Paul B Mahol
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 #include <float.h>
23 #include <math.h>
24
25 #include "libavutil/mem.h"
26 #include "libavutil/opt.h"
27 #include "audio.h"
28 #include "avfilter.h"
29 #include "internal.h"
30
31 #define HISTOGRAM_SIZE 8192
32 #define HISTOGRAM_MAX (HISTOGRAM_SIZE-1)
33
34 #define MEASURE_ALL UINT_MAX
35 #define MEASURE_NONE 0
36
37 #define MEASURE_DC_OFFSET (1 << 0)
38 #define MEASURE_MIN_LEVEL (1 << 1)
39 #define MEASURE_MAX_LEVEL (1 << 2)
40 #define MEASURE_MIN_DIFFERENCE (1 << 3)
41 #define MEASURE_MAX_DIFFERENCE (1 << 4)
42 #define MEASURE_MEAN_DIFFERENCE (1 << 5)
43 #define MEASURE_RMS_DIFFERENCE (1 << 6)
44 #define MEASURE_PEAK_LEVEL (1 << 7)
45 #define MEASURE_RMS_LEVEL (1 << 8)
46 #define MEASURE_RMS_PEAK (1 << 9)
47 #define MEASURE_RMS_TROUGH (1 << 10)
48 #define MEASURE_CREST_FACTOR (1 << 11)
49 #define MEASURE_FLAT_FACTOR (1 << 12)
50 #define MEASURE_PEAK_COUNT (1 << 13)
51 #define MEASURE_BIT_DEPTH (1 << 14)
52 #define MEASURE_DYNAMIC_RANGE (1 << 15)
53 #define MEASURE_ZERO_CROSSINGS (1 << 16)
54 #define MEASURE_ZERO_CROSSINGS_RATE (1 << 17)
55 #define MEASURE_NUMBER_OF_SAMPLES (1 << 18)
56 #define MEASURE_NUMBER_OF_NANS (1 << 19)
57 #define MEASURE_NUMBER_OF_INFS (1 << 20)
58 #define MEASURE_NUMBER_OF_DENORMALS (1 << 21)
59 #define MEASURE_NOISE_FLOOR (1 << 22)
60 #define MEASURE_NOISE_FLOOR_COUNT (1 << 23)
61 #define MEASURE_ENTROPY (1 << 24)
62 #define MEASURE_ABS_PEAK_COUNT (1 << 25)
63
64 #define MEASURE_MINMAXPEAK (MEASURE_MIN_LEVEL | MEASURE_MAX_LEVEL | MEASURE_PEAK_LEVEL)
65
66 typedef struct ChannelStats {
67 double last;
68 double last_non_zero;
69 double min_non_zero;
70 double sigma_x, sigma_x2;
71 double avg_sigma_x2, min_sigma_x2, max_sigma_x2;
72 double min, max;
73 double nmin, nmax;
74 double min_run, max_run;
75 double min_runs, max_runs;
76 double min_diff, max_diff;
77 double diff1_sum;
78 double diff1_sum_x2;
79 double abs_peak;
80 uint64_t mask[4];
81 uint64_t min_count, max_count;
82 uint64_t abs_peak_count;
83 uint64_t noise_floor_count;
84 uint64_t zero_runs;
85 uint64_t nb_samples;
86 uint64_t nb_nans;
87 uint64_t nb_infs;
88 uint64_t nb_denormals;
89 double *win_samples;
90 double *sorted_samples;
91 uint64_t ehistogram[HISTOGRAM_SIZE];
92 int64_t lasti;
93 int sorted_front;
94 int sorted_back;
95 int win_pos;
96 int max_index;
97 double noise_floor;
98 double entropy;
99 } ChannelStats;
100
101 typedef struct AudioStatsContext {
102 const AVClass *class;
103 ChannelStats *chstats;
104 int nb_channels;
105 uint64_t tc_samples;
106 double time_constant;
107 double mult;
108 int metadata;
109 int used;
110 int reset_count;
111 int nb_frames;
112 int maxbitdepth;
113 int measure_perchannel;
114 int measure_overall;
115 int is_float;
116 int is_double;
117 } AudioStatsContext;
118
119 #define OFFSET(x) offsetof(AudioStatsContext, x)
120 #define FLAGS AV_OPT_FLAG_AUDIO_PARAM|AV_OPT_FLAG_FILTERING_PARAM
121
122 static const AVOption astats_options[] = {
123 { "length", "set the window length", OFFSET(time_constant), AV_OPT_TYPE_DOUBLE, {.dbl=.05}, 0, 10, FLAGS },
124 { "metadata", "inject metadata in the filtergraph", OFFSET(metadata), AV_OPT_TYPE_BOOL, {.i64=0}, 0, 1, FLAGS },
125 { "reset", "Set the number of frames over which cumulative stats are calculated before being reset", OFFSET(reset_count), AV_OPT_TYPE_INT, {.i64=0}, 0, INT_MAX, FLAGS },
126 { "measure_perchannel", "Select the parameters which are measured per channel", OFFSET(measure_perchannel), AV_OPT_TYPE_FLAGS, {.i64=MEASURE_ALL}, 0, UINT_MAX, FLAGS, .unit = "measure" },
127 { "none" , "", 0, AV_OPT_TYPE_CONST, {.i64=MEASURE_NONE }, 0, 0, FLAGS, .unit = "measure" },
128 { "all" , "", 0, AV_OPT_TYPE_CONST, {.i64=MEASURE_ALL }, 0, 0, FLAGS, .unit = "measure" },
129 { "Bit_depth" , "", 0, AV_OPT_TYPE_CONST, {.i64=MEASURE_BIT_DEPTH }, 0, 0, FLAGS, .unit = "measure" },
130 { "Crest_factor" , "", 0, AV_OPT_TYPE_CONST, {.i64=MEASURE_CREST_FACTOR }, 0, 0, FLAGS, .unit = "measure" },
131 { "DC_offset" , "", 0, AV_OPT_TYPE_CONST, {.i64=MEASURE_DC_OFFSET }, 0, 0, FLAGS, .unit = "measure" },
132 { "Dynamic_range" , "", 0, AV_OPT_TYPE_CONST, {.i64=MEASURE_DYNAMIC_RANGE }, 0, 0, FLAGS, .unit = "measure" },
133 { "Entropy" , "", 0, AV_OPT_TYPE_CONST, {.i64=MEASURE_ENTROPY }, 0, 0, FLAGS, .unit = "measure" },
134 { "Flat_factor" , "", 0, AV_OPT_TYPE_CONST, {.i64=MEASURE_FLAT_FACTOR }, 0, 0, FLAGS, .unit = "measure" },
135 { "Max_difference" , "", 0, AV_OPT_TYPE_CONST, {.i64=MEASURE_MAX_DIFFERENCE }, 0, 0, FLAGS, .unit = "measure" },
136 { "Max_level" , "", 0, AV_OPT_TYPE_CONST, {.i64=MEASURE_MAX_LEVEL }, 0, 0, FLAGS, .unit = "measure" },
137 { "Mean_difference" , "", 0, AV_OPT_TYPE_CONST, {.i64=MEASURE_MEAN_DIFFERENCE }, 0, 0, FLAGS, .unit = "measure" },
138 { "Min_difference" , "", 0, AV_OPT_TYPE_CONST, {.i64=MEASURE_MIN_DIFFERENCE }, 0, 0, FLAGS, .unit = "measure" },
139 { "Min_level" , "", 0, AV_OPT_TYPE_CONST, {.i64=MEASURE_MIN_LEVEL }, 0, 0, FLAGS, .unit = "measure" },
140 { "Noise_floor" , "", 0, AV_OPT_TYPE_CONST, {.i64=MEASURE_NOISE_FLOOR }, 0, 0, FLAGS, .unit = "measure" },
141 { "Noise_floor_count" , "", 0, AV_OPT_TYPE_CONST, {.i64=MEASURE_NOISE_FLOOR_COUNT }, 0, 0, FLAGS, .unit = "measure" },
142 { "Number_of_Infs" , "", 0, AV_OPT_TYPE_CONST, {.i64=MEASURE_NUMBER_OF_INFS }, 0, 0, FLAGS, .unit = "measure" },
143 { "Number_of_NaNs" , "", 0, AV_OPT_TYPE_CONST, {.i64=MEASURE_NUMBER_OF_NANS }, 0, 0, FLAGS, .unit = "measure" },
144 { "Number_of_denormals" , "", 0, AV_OPT_TYPE_CONST, {.i64=MEASURE_NUMBER_OF_DENORMALS }, 0, 0, FLAGS, .unit = "measure" },
145 { "Number_of_samples" , "", 0, AV_OPT_TYPE_CONST, {.i64=MEASURE_NUMBER_OF_SAMPLES }, 0, 0, FLAGS, .unit = "measure" },
146 { "Peak_count" , "", 0, AV_OPT_TYPE_CONST, {.i64=MEASURE_PEAK_COUNT }, 0, 0, FLAGS, .unit = "measure" },
147 { "Peak_level" , "", 0, AV_OPT_TYPE_CONST, {.i64=MEASURE_PEAK_LEVEL }, 0, 0, FLAGS, .unit = "measure" },
148 { "RMS_difference" , "", 0, AV_OPT_TYPE_CONST, {.i64=MEASURE_RMS_DIFFERENCE }, 0, 0, FLAGS, .unit = "measure" },
149 { "RMS_level" , "", 0, AV_OPT_TYPE_CONST, {.i64=MEASURE_RMS_LEVEL }, 0, 0, FLAGS, .unit = "measure" },
150 { "RMS_peak" , "", 0, AV_OPT_TYPE_CONST, {.i64=MEASURE_RMS_PEAK }, 0, 0, FLAGS, .unit = "measure" },
151 { "RMS_trough" , "", 0, AV_OPT_TYPE_CONST, {.i64=MEASURE_RMS_TROUGH }, 0, 0, FLAGS, .unit = "measure" },
152 { "Zero_crossings" , "", 0, AV_OPT_TYPE_CONST, {.i64=MEASURE_ZERO_CROSSINGS }, 0, 0, FLAGS, .unit = "measure" },
153 { "Zero_crossings_rate" , "", 0, AV_OPT_TYPE_CONST, {.i64=MEASURE_ZERO_CROSSINGS_RATE }, 0, 0, FLAGS, .unit = "measure" },
154 { "Abs_Peak_count" , "", 0, AV_OPT_TYPE_CONST, {.i64=MEASURE_ABS_PEAK_COUNT }, 0, 0, FLAGS, .unit = "measure" },
155 { "measure_overall", "Select the parameters which are measured overall", OFFSET(measure_overall), AV_OPT_TYPE_FLAGS, {.i64=MEASURE_ALL}, 0, UINT_MAX, FLAGS, .unit = "measure" },
156 { NULL }
157 };
158
159 AVFILTER_DEFINE_CLASS(astats);
160
161 static void reset_stats(AudioStatsContext *s)
162 {
163 int c;
164
165 for (c = 0; c < s->nb_channels; c++) {
166 ChannelStats *p = &s->chstats[c];
167
168 p->min = p->nmin = p->min_sigma_x2 = DBL_MAX;
169 p->max = p->nmax = p->max_sigma_x2 =-DBL_MAX;
170 p->abs_peak = 0;
171 p->min_non_zero = DBL_MAX;
172 p->min_diff = DBL_MAX;
173 p->max_diff = 0;
174 p->sigma_x = 0;
175 p->sigma_x2 = 0;
176 p->avg_sigma_x2 = 0;
177 p->min_run = 0;
178 p->max_run = 0;
179 p->min_runs = 0;
180 p->max_runs = 0;
181 p->diff1_sum = 0;
182 p->diff1_sum_x2 = 0;
183 p->mask[0] = 0;
184 p->mask[1] = 0;
185 p->mask[2] =~0;
186 p->mask[3] = 0;
187 p->min_count = 0;
188 p->max_count = 0;
189 p->abs_peak_count = 0;
190 p->zero_runs = 0;
191 p->nb_samples = 0;
192 p->nb_nans = 0;
193 p->nb_infs = 0;
194 p->nb_denormals = 0;
195 p->last = NAN;
196 p->noise_floor = NAN;
197 p->noise_floor_count = 0;
198 p->entropy = 0;
199 p->win_pos = 0;
200 p->sorted_front = 0;
201 p->sorted_back = 0;
202 memset(p->win_samples, 0, s->tc_samples * sizeof(*p->win_samples));
203 memset(p->ehistogram, 0, sizeof(p->ehistogram));
204 for (int n = 0; n < s->tc_samples; n++)
205 p->sorted_samples[n] = -1.0;
206 }
207 }
208
209 static int config_output(AVFilterLink *outlink)
210 {
211 AudioStatsContext *s = outlink->src->priv;
212
213 s->chstats = av_calloc(sizeof(*s->chstats), outlink->ch_layout.nb_channels);
214 if (!s->chstats)
215 return AVERROR(ENOMEM);
216
217 s->tc_samples = FFMAX(s->time_constant * outlink->sample_rate + .5, 1);
218 s->nb_channels = outlink->ch_layout.nb_channels;
219
220 for (int i = 0; i < s->nb_channels; i++) {
221 ChannelStats *p = &s->chstats[i];
222
223 p->win_samples = av_calloc(s->tc_samples, sizeof(*p->win_samples));
224 if (!p->win_samples)
225 return AVERROR(ENOMEM);
226
227 p->sorted_samples = av_calloc(s->tc_samples, sizeof(*p->sorted_samples));
228 if (!p->sorted_samples)
229 return AVERROR(ENOMEM);
230 }
231
232 s->mult = exp((-1 / s->time_constant / outlink->sample_rate));
233 s->nb_frames = 0;
234 s->maxbitdepth = av_get_bytes_per_sample(outlink->format) * 8;
235 s->is_double = outlink->format == AV_SAMPLE_FMT_DBL ||
236 outlink->format == AV_SAMPLE_FMT_DBLP;
237
238 s->is_float = outlink->format == AV_SAMPLE_FMT_FLT ||
239 outlink->format == AV_SAMPLE_FMT_FLTP;
240
241 reset_stats(s);
242
243 return 0;
244 }
245
246 static void bit_depth(AudioStatsContext *s, const uint64_t *const mask, uint8_t *depth)
247 {
248 unsigned result = s->maxbitdepth;
249 uint64_t amask = mask[1] & (~mask[2]);
250
251 depth[0] = 0;
252 for (int i = 0; i < result; i++)
253 depth[0] += !!(mask[0] & (1ULL << i));
254
255 depth[1] = 0;
256 for (int i = 0; i < result; i++)
257 depth[1] += !!(mask[1] & (1ULL << i));
258
259 depth[2] = result;
260 for (int i = 0; i < result && !(amask & 1); i++) {
261 depth[2]--;
262 amask >>= 1;
263 }
264
265 depth[3] = 0;
266 for (int i = 0; i < result; i++)
267 depth[3] += !!(mask[3] & (1ULL << i));
268 }
269
270 static double calc_entropy(AudioStatsContext *s, ChannelStats *p)
271 {
272 double entropy = 0.;
273
274 for (int i = 0; i < HISTOGRAM_SIZE; i++) {
275 double entry = p->ehistogram[i] / ((double)p->nb_samples);
276
277 if (entry > 1e-8)
278 entropy += entry * log2(entry);
279 }
280
281 return -entropy / log2(HISTOGRAM_SIZE);
282 }
283
284 static double calc_noise_floor(double *ss, double x, double px,
285 int n, int *ffront, int *bback)
286 {
287 double r, ax = fabs(x);
288 int front = *ffront;
289 int back = *bback;
290 int empty = front == back && ss[front] == -1.0;
291
292 if (!empty && fabs(px) == ss[front]) {
293 ss[front] = -1.0;
294 if (back != front) {
295 front--;
296 if (front < 0)
297 front = n - 1;
298 }
299 empty = front == back;
300 }
301
302 if (!empty && ax >= ss[front]) {
303 while (1) {
304 ss[front] = -1.0;
305 if (back == front) {
306 empty = 1;
307 break;
308 }
309 front--;
310 if (front < 0)
311 front = n - 1;
312 }
313 }
314
315 while (!empty && ax >= ss[back]) {
316 ss[back] = -1.0;
317 if (back == front) {
318 empty = 1;
319 break;
320 }
321 back++;
322 if (back >= n)
323 back = 0;
324 }
325
326 if (!empty) {
327 back--;
328 if (back < 0)
329 back = n - 1;
330 }
331
332 ss[back] = ax;
333 r = ss[front];
334
335 *ffront = front;
336 *bback = back;
337
338 return r;
339 }
340
341 static inline void update_minmax(AudioStatsContext *s, ChannelStats *p, double d)
342 {
343 if (d < p->min)
344 p->min = d;
345 if (d > p->max)
346 p->max = d;
347 }
348
349 static inline void update_stat(AudioStatsContext *s, ChannelStats *p, double d, double nd, int64_t i)
350 {
351 double abs_d = FFABS(d);
352 double drop, noise_floor;
353 int index;
354
355 if (p->abs_peak < abs_d) {
356 p->abs_peak = abs_d;
357 p->abs_peak_count = 1;
358 } else if (p->abs_peak == abs_d) {
359 p->abs_peak_count++;
360 }
361 if (d < p->min) {
362 p->min = d;
363 p->nmin = nd;
364 p->min_run = 1;
365 p->min_runs = 0;
366 p->min_count = 1;
367 } else if (d == p->min) {
368 p->min_count++;
369 p->min_run = d == p->last ? p->min_run + 1 : 1;
370 } else if (p->last == p->min) {
371 p->min_runs += p->min_run * p->min_run;
372 }
373
374 if (d != 0 && FFABS(d) < p->min_non_zero)
375 p->min_non_zero = FFABS(d);
376
377 if (d > p->max) {
378 p->max = d;
379 p->nmax = nd;
380 p->max_run = 1;
381 p->max_runs = 0;
382 p->max_count = 1;
383 } else if (d == p->max) {
384 p->max_count++;
385 p->max_run = d == p->last ? p->max_run + 1 : 1;
386 } else if (p->last == p->max) {
387 p->max_runs += p->max_run * p->max_run;
388 }
389
390 if (d != 0) {
391 p->zero_runs += FFSIGN(d) != FFSIGN(p->last_non_zero);
392 p->last_non_zero = d;
393 }
394
395 p->sigma_x += nd;
396 p->sigma_x2 += nd * nd;
397 p->avg_sigma_x2 = p->avg_sigma_x2 * s->mult + (1.0 - s->mult) * nd * nd;
398 if (!isnan(p->last)) {
399 p->min_diff = FFMIN(p->min_diff, fabs(d - p->last));
400 p->max_diff = FFMAX(p->max_diff, fabs(d - p->last));
401 p->diff1_sum += fabs(d - p->last);
402 p->diff1_sum_x2 += (d - p->last) * (d - p->last);
403 }
404 p->mask[0] |= (i < 0) ? -i : i;
405 p->mask[1] |= i;
406 p->mask[2] &= i;
407 if (!isnan(p->last))
408 p->mask[3] |= i ^ p->lasti;
409 p->lasti = i;
410 p->last = d;
411
412 drop = p->win_samples[p->win_pos];
413 p->win_samples[p->win_pos] = nd;
414 index = av_clip(lrint(av_clipd(FFABS(nd), 0.0, 1.0) * HISTOGRAM_MAX), 0, HISTOGRAM_MAX);
415 p->max_index = FFMAX(p->max_index, index);
416 p->ehistogram[index]++;
417 p->win_pos++;
418
419 if (p->win_pos >= s->tc_samples)
420 p->win_pos = 0;
421
422 if (p->nb_samples >= s->tc_samples) {
423 p->max_sigma_x2 = FFMAX(p->max_sigma_x2, p->avg_sigma_x2);
424 p->min_sigma_x2 = FFMIN(p->min_sigma_x2, p->avg_sigma_x2);
425 }
426 p->nb_samples++;
427
428 noise_floor = calc_noise_floor(p->sorted_samples, nd, drop,
429 s->tc_samples, &p->sorted_front, &p->sorted_back);
430 if (p->nb_samples >= s->tc_samples) {
431 if (isnan(p->noise_floor)) {
432 p->noise_floor = noise_floor;
433 p->noise_floor_count = 1;
434 } else {
435 if (noise_floor < p->noise_floor) {
436 p->noise_floor = noise_floor;
437 p->noise_floor_count = 1;
438 } else if (noise_floor == p->noise_floor) {
439 p->noise_floor_count++;
440 }
441 }
442 }
443 }
444
445 static inline void update_float_stat(AudioStatsContext *s, ChannelStats *p, float d)
446 {
447 int type = fpclassify(d);
448
449 p->nb_nans += type == FP_NAN;
450 p->nb_infs += type == FP_INFINITE;
451 p->nb_denormals += type == FP_SUBNORMAL;
452 }
453
454 static inline void update_double_stat(AudioStatsContext *s, ChannelStats *p, double d)
455 {
456 int type = fpclassify(d);
457
458 p->nb_nans += type == FP_NAN;
459 p->nb_infs += type == FP_INFINITE;
460 p->nb_denormals += type == FP_SUBNORMAL;
461 }
462
463 static void set_meta(AVDictionary **metadata, int chan, const char *key,
464 const char *fmt, double val)
465 {
466 uint8_t value[128];
467 uint8_t key2[128];
468
469 snprintf(value, sizeof(value), fmt, val);
470 if (chan)
471 snprintf(key2, sizeof(key2), "lavfi.astats.%d.%s", chan, key);
472 else
473 snprintf(key2, sizeof(key2), "lavfi.astats.%s", key);
474 av_dict_set(metadata, key2, value, 0);
475 }
476
477 #define LINEAR_TO_DB(x) (log10(x) * 20)
478
479 static void set_metadata(AudioStatsContext *s, AVDictionary **metadata)
480 {
481 uint64_t mask[4], min_count = 0, max_count = 0, nb_samples = 0, noise_floor_count = 0;
482 uint64_t nb_nans = 0, nb_infs = 0, nb_denormals = 0;
483 uint64_t abs_peak_count = 0;
484 double min_runs = 0, max_runs = 0,
485 min = DBL_MAX, max =-DBL_MAX, min_diff = DBL_MAX, max_diff = 0,
486 nmin = DBL_MAX, nmax =-DBL_MAX,
487 max_sigma_x = 0,
488 diff1_sum = 0,
489 diff1_sum_x2 = 0,
490 sigma_x2 = 0,
491 noise_floor = 0,
492 entropy = 0,
493 min_sigma_x2 = DBL_MAX,
494 max_sigma_x2 =-DBL_MAX;
495 uint8_t depth[4];
496 int c;
497
498 mask[0] = 0;
499 mask[1] = 0;
500 mask[2] =~0;
501 mask[3] = 0;
502
503 for (c = 0; c < s->nb_channels; c++) {
504 ChannelStats *p = &s->chstats[c];
505
506 if (p->nb_samples < s->tc_samples)
507 p->min_sigma_x2 = p->max_sigma_x2 = p->sigma_x2 / p->nb_samples;
508
509 min = FFMIN(min, p->min);
510 max = FFMAX(max, p->max);
511 nmin = FFMIN(nmin, p->nmin);
512 nmax = FFMAX(nmax, p->nmax);
513 min_diff = FFMIN(min_diff, p->min_diff);
514 max_diff = FFMAX(max_diff, p->max_diff);
515 diff1_sum += p->diff1_sum;
516 diff1_sum_x2 += p->diff1_sum_x2;
517 min_sigma_x2 = FFMIN(min_sigma_x2, p->min_sigma_x2);
518 max_sigma_x2 = FFMAX(max_sigma_x2, p->max_sigma_x2);
519 sigma_x2 += p->sigma_x2;
520 noise_floor = FFMAX(noise_floor, p->noise_floor);
521 noise_floor_count += p->noise_floor_count;
522 p->entropy = calc_entropy(s, p);
523 entropy += p->entropy;
524 min_count += p->min_count;
525 max_count += p->max_count;
526 abs_peak_count += p->abs_peak_count;
527 min_runs += p->min_runs;
528 max_runs += p->max_runs;
529 mask[0] |= p->mask[0];
530 mask[1] |= p->mask[1];
531 mask[2] &= p->mask[2];
532 mask[3] |= p->mask[3];
533 nb_samples += p->nb_samples;
534 nb_nans += p->nb_nans;
535 nb_infs += p->nb_infs;
536 nb_denormals += p->nb_denormals;
537 if (fabs(p->sigma_x) > fabs(max_sigma_x))
538 max_sigma_x = p->sigma_x;
539
540 if (s->measure_perchannel & MEASURE_DC_OFFSET)
541 set_meta(metadata, c + 1, "DC_offset", "%f", p->sigma_x / p->nb_samples);
542 if (s->measure_perchannel & MEASURE_MIN_LEVEL)
543 set_meta(metadata, c + 1, "Min_level", "%f", p->min);
544 if (s->measure_perchannel & MEASURE_MAX_LEVEL)
545 set_meta(metadata, c + 1, "Max_level", "%f", p->max);
546 if (s->measure_perchannel & MEASURE_MIN_DIFFERENCE)
547 set_meta(metadata, c + 1, "Min_difference", "%f", p->min_diff);
548 if (s->measure_perchannel & MEASURE_MAX_DIFFERENCE)
549 set_meta(metadata, c + 1, "Max_difference", "%f", p->max_diff);
550 if (s->measure_perchannel & MEASURE_MEAN_DIFFERENCE)
551 set_meta(metadata, c + 1, "Mean_difference", "%f", p->diff1_sum / (p->nb_samples - 1));
552 if (s->measure_perchannel & MEASURE_RMS_DIFFERENCE)
553 set_meta(metadata, c + 1, "RMS_difference", "%f", sqrt(p->diff1_sum_x2 / (p->nb_samples - 1)));
554 if (s->measure_perchannel & MEASURE_PEAK_LEVEL)
555 set_meta(metadata, c + 1, "Peak_level", "%f", LINEAR_TO_DB(FFMAX(-p->nmin, p->nmax)));
556 if (s->measure_perchannel & MEASURE_RMS_LEVEL)
557 set_meta(metadata, c + 1, "RMS_level", "%f", LINEAR_TO_DB(sqrt(p->sigma_x2 / p->nb_samples)));
558 if (s->measure_perchannel & MEASURE_RMS_PEAK)
559 set_meta(metadata, c + 1, "RMS_peak", "%f", LINEAR_TO_DB(sqrt(p->max_sigma_x2)));
560 if (s->measure_perchannel & MEASURE_RMS_TROUGH)
561 set_meta(metadata, c + 1, "RMS_trough", "%f", LINEAR_TO_DB(sqrt(p->min_sigma_x2)));
562 if (s->measure_perchannel & MEASURE_CREST_FACTOR)
563 set_meta(metadata, c + 1, "Crest_factor", "%f", p->sigma_x2 ? FFMAX(-p->min, p->max) / sqrt(p->sigma_x2 / p->nb_samples) : 1);
564 if (s->measure_perchannel & MEASURE_FLAT_FACTOR)
565 set_meta(metadata, c + 1, "Flat_factor", "%f", LINEAR_TO_DB((p->min_runs + p->max_runs) / (p->min_count + p->max_count)));
566 if (s->measure_perchannel & MEASURE_PEAK_COUNT)
567 set_meta(metadata, c + 1, "Peak_count", "%f", (float)(p->min_count + p->max_count));
568 if (s->measure_perchannel & MEASURE_ABS_PEAK_COUNT)
569 set_meta(metadata, c + 1, "Peak_count", "%f", p->abs_peak_count);
570 if (s->measure_perchannel & MEASURE_NOISE_FLOOR)
571 set_meta(metadata, c + 1, "Noise_floor", "%f", LINEAR_TO_DB(p->noise_floor));
572 if (s->measure_perchannel & MEASURE_NOISE_FLOOR_COUNT)
573 set_meta(metadata, c + 1, "Noise_floor_count", "%f", p->noise_floor_count);
574 if (s->measure_perchannel & MEASURE_ENTROPY)
575 set_meta(metadata, c + 1, "Entropy", "%f", p->entropy);
576 if (s->measure_perchannel & MEASURE_BIT_DEPTH) {
577 bit_depth(s, p->mask, depth);
578 set_meta(metadata, c + 1, "Bit_depth", "%f", depth[0]);
579 set_meta(metadata, c + 1, "Bit_depth2", "%f", depth[1]);
580 set_meta(metadata, c + 1, "Bit_depth3", "%f", depth[2]);
581 set_meta(metadata, c + 1, "Bit_depth4", "%f", depth[3]);
582 }
583 if (s->measure_perchannel & MEASURE_DYNAMIC_RANGE)
584 set_meta(metadata, c + 1, "Dynamic_range", "%f", LINEAR_TO_DB(2 * FFMAX(FFABS(p->min), FFABS(p->max))/ p->min_non_zero));
585 if (s->measure_perchannel & MEASURE_ZERO_CROSSINGS)
586 set_meta(metadata, c + 1, "Zero_crossings", "%f", p->zero_runs);
587 if (s->measure_perchannel & MEASURE_ZERO_CROSSINGS_RATE)
588 set_meta(metadata, c + 1, "Zero_crossings_rate", "%f", p->zero_runs/(double)p->nb_samples);
589 if ((s->is_float || s->is_double) && s->measure_perchannel & MEASURE_NUMBER_OF_NANS)
590 set_meta(metadata, c + 1, "Number of NaNs", "%f", p->nb_nans);
591 if ((s->is_float || s->is_double) && s->measure_perchannel & MEASURE_NUMBER_OF_INFS)
592 set_meta(metadata, c + 1, "Number of Infs", "%f", p->nb_infs);
593 if ((s->is_float || s->is_double) && s->measure_perchannel & MEASURE_NUMBER_OF_DENORMALS)
594 set_meta(metadata, c + 1, "Number of denormals", "%f", p->nb_denormals);
595 }
596
597 if (s->measure_overall & MEASURE_DC_OFFSET)
598 set_meta(metadata, 0, "Overall.DC_offset", "%f", max_sigma_x / (nb_samples / s->nb_channels));
599 if (s->measure_overall & MEASURE_MIN_LEVEL)
600 set_meta(metadata, 0, "Overall.Min_level", "%f", min);
601 if (s->measure_overall & MEASURE_MAX_LEVEL)
602 set_meta(metadata, 0, "Overall.Max_level", "%f", max);
603 if (s->measure_overall & MEASURE_MIN_DIFFERENCE)
604 set_meta(metadata, 0, "Overall.Min_difference", "%f", min_diff);
605 if (s->measure_overall & MEASURE_MAX_DIFFERENCE)
606 set_meta(metadata, 0, "Overall.Max_difference", "%f", max_diff);
607 if (s->measure_overall & MEASURE_MEAN_DIFFERENCE)
608 set_meta(metadata, 0, "Overall.Mean_difference", "%f", diff1_sum / (nb_samples - s->nb_channels));
609 if (s->measure_overall & MEASURE_RMS_DIFFERENCE)
610 set_meta(metadata, 0, "Overall.RMS_difference", "%f", sqrt(diff1_sum_x2 / (nb_samples - s->nb_channels)));
611 if (s->measure_overall & MEASURE_PEAK_LEVEL)
612 set_meta(metadata, 0, "Overall.Peak_level", "%f", LINEAR_TO_DB(FFMAX(-nmin, nmax)));
613 if (s->measure_overall & MEASURE_RMS_LEVEL)
614 set_meta(metadata, 0, "Overall.RMS_level", "%f", LINEAR_TO_DB(sqrt(sigma_x2 / nb_samples)));
615 if (s->measure_overall & MEASURE_RMS_PEAK)
616 set_meta(metadata, 0, "Overall.RMS_peak", "%f", LINEAR_TO_DB(sqrt(max_sigma_x2)));
617 if (s->measure_overall & MEASURE_RMS_TROUGH)
618 set_meta(metadata, 0, "Overall.RMS_trough", "%f", LINEAR_TO_DB(sqrt(min_sigma_x2)));
619 if (s->measure_overall & MEASURE_FLAT_FACTOR)
620 set_meta(metadata, 0, "Overall.Flat_factor", "%f", LINEAR_TO_DB((min_runs + max_runs) / (min_count + max_count)));
621 if (s->measure_overall & MEASURE_PEAK_COUNT)
622 set_meta(metadata, 0, "Overall.Peak_count", "%f", (float)(min_count + max_count) / (double)s->nb_channels);
623 if (s->measure_overall & MEASURE_ABS_PEAK_COUNT)
624 set_meta(metadata, 0, "Overall.Abs_Peak_count", "%f", (float)(abs_peak_count) / (double)s->nb_channels);
625 if (s->measure_overall & MEASURE_NOISE_FLOOR)
626 set_meta(metadata, 0, "Overall.Noise_floor", "%f", LINEAR_TO_DB(noise_floor));
627 if (s->measure_overall & MEASURE_NOISE_FLOOR_COUNT)
628 set_meta(metadata, 0, "Overall.Noise_floor_count", "%f", noise_floor_count / (double)s->nb_channels);
629 if (s->measure_overall & MEASURE_ENTROPY)
630 set_meta(metadata, 0, "Overall.Entropy", "%f", entropy / (double)s->nb_channels);
631 if (s->measure_overall & MEASURE_BIT_DEPTH) {
632 bit_depth(s, mask, depth);
633 set_meta(metadata, 0, "Overall.Bit_depth", "%f", depth[0]);
634 set_meta(metadata, 0, "Overall.Bit_depth2", "%f", depth[1]);
635 set_meta(metadata, 0, "Overall.Bit_depth3", "%f", depth[2]);
636 set_meta(metadata, 0, "Overall.Bit_depth4", "%f", depth[3]);
637 }
638 if (s->measure_overall & MEASURE_NUMBER_OF_SAMPLES)
639 set_meta(metadata, 0, "Overall.Number_of_samples", "%f", nb_samples / s->nb_channels);
640 if ((s->is_float || s->is_double) && s->measure_overall & MEASURE_NUMBER_OF_NANS)
641 set_meta(metadata, 0, "Number of NaNs", "%f", nb_nans / (float)s->nb_channels);
642 if ((s->is_float || s->is_double) && s->measure_overall & MEASURE_NUMBER_OF_INFS)
643 set_meta(metadata, 0, "Number of Infs", "%f", nb_infs / (float)s->nb_channels);
644 if ((s->is_float || s->is_double) && s->measure_overall & MEASURE_NUMBER_OF_DENORMALS)
645 set_meta(metadata, 0, "Number of denormals", "%f", nb_denormals / (float)s->nb_channels);
646 }
647
648 #define UPDATE_STATS_P(type, update_func, update_float, channel_func) \
649 for (int c = start; c < end; c++) { \
650 ChannelStats *p = &s->chstats[c]; \
651 const type *src = (const type *)data[c]; \
652 const type * const srcend = src + samples; \
653 for (; src < srcend; src++) { \
654 update_func; \
655 update_float; \
656 } \
657 channel_func; \
658 }
659
660 #define UPDATE_STATS_I(type, update_func, update_float, channel_func) \
661 for (int c = start; c < end; c++) { \
662 ChannelStats *p = &s->chstats[c]; \
663 const type *src = (const type *)data[0]; \
664 const type * const srcend = src + samples * channels; \
665 for (src += c; src < srcend; src += channels) { \
666 update_func; \
667 update_float; \
668 } \
669 channel_func; \
670 }
671
672 #define UPDATE_STATS(planar, type, sample, normalizer_suffix, int_sample) \
673 if ((s->measure_overall | s->measure_perchannel) & ~MEASURE_MINMAXPEAK) { \
674 UPDATE_STATS_##planar(type, update_stat(s, p, sample, sample normalizer_suffix, int_sample), s->is_float ? update_float_stat(s, p, sample) : s->is_double ? update_double_stat(s, p, sample) : (void)NULL, ); \
675 } else { \
676 UPDATE_STATS_##planar(type, update_minmax(s, p, sample), , p->nmin = p->min normalizer_suffix; p->nmax = p->max normalizer_suffix;); \
677 }
678
679 static int filter_channel(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)
680 {
681 AudioStatsContext *s = ctx->priv;
682 AVFilterLink *inlink = ctx->inputs[0];
683 AVFrame *buf = arg;
684 const uint8_t * const * const data = (const uint8_t * const *)buf->extended_data;
685 const int channels = s->nb_channels;
686 const int samples = buf->nb_samples;
687 const int start = (buf->ch_layout.nb_channels * jobnr) / nb_jobs;
688 const int end = (buf->ch_layout.nb_channels * (jobnr+1)) / nb_jobs;
689
690 switch (inlink->format) {
691 case AV_SAMPLE_FMT_DBLP:
692 UPDATE_STATS(P, double, *src, , llrint(*src * (UINT64_C(1) << 63)));
693 break;
694 case AV_SAMPLE_FMT_DBL:
695 UPDATE_STATS(I, double, *src, , llrint(*src * (UINT64_C(1) << 63)));
696 break;
697 case AV_SAMPLE_FMT_FLTP:
698 UPDATE_STATS(P, float, *src, , llrint(*src * (UINT64_C(1) << 31)));
699 break;
700 case AV_SAMPLE_FMT_FLT:
701 UPDATE_STATS(I, float, *src, , llrint(*src * (UINT64_C(1) << 31)));
702 break;
703 case AV_SAMPLE_FMT_S64P:
704 UPDATE_STATS(P, int64_t, *src, / (double)INT64_MAX, *src);
705 break;
706 case AV_SAMPLE_FMT_S64:
707 UPDATE_STATS(I, int64_t, *src, / (double)INT64_MAX, *src);
708 break;
709 case AV_SAMPLE_FMT_S32P:
710 UPDATE_STATS(P, int32_t, *src, / (double)INT32_MAX, *src);
711 break;
712 case AV_SAMPLE_FMT_S32:
713 UPDATE_STATS(I, int32_t, *src, / (double)INT32_MAX, *src);
714 break;
715 case AV_SAMPLE_FMT_S16P:
716 UPDATE_STATS(P, int16_t, *src, / (double)INT16_MAX, *src);
717 break;
718 case AV_SAMPLE_FMT_S16:
719 UPDATE_STATS(I, int16_t, *src, / (double)INT16_MAX, *src);
720 break;
721 }
722
723 return 0;
724 }
725
726 static int filter_frame(AVFilterLink *inlink, AVFrame *buf)
727 {
728 AVFilterContext *ctx = inlink->dst;
729 AudioStatsContext *s = ctx->priv;
730 AVDictionary **metadata = &buf->metadata;
731
732 if (s->reset_count > 0) {
733 if (s->nb_frames >= s->reset_count) {
734 reset_stats(s);
735 s->nb_frames = 0;
736 }
737 s->nb_frames++;
738 }
739
740 if (s->used == 0)
741 s->used = buf->nb_samples > 0;
742 ff_filter_execute(ctx, filter_channel, buf, NULL,
743 FFMIN(inlink->ch_layout.nb_channels, ff_filter_get_nb_threads(ctx)));
744
745 if (s->metadata)
746 set_metadata(s, metadata);
747
748 return ff_filter_frame(inlink->dst->outputs[0], buf);
749 }
750
751 static void print_stats(AVFilterContext *ctx)
752 {
753 AudioStatsContext *s = ctx->priv;
754 uint64_t mask[4], min_count = 0, max_count = 0, nb_samples = 0, noise_floor_count = 0;
755 uint64_t nb_nans = 0, nb_infs = 0, nb_denormals = 0, abs_peak_count = 0;
756 double min_runs = 0, max_runs = 0,
757 min = DBL_MAX, max =-DBL_MAX, min_diff = DBL_MAX, max_diff = 0,
758 nmin = DBL_MAX, nmax =-DBL_MAX,
759 max_sigma_x = 0,
760 diff1_sum_x2 = 0,
761 diff1_sum = 0,
762 sigma_x2 = 0,
763 noise_floor = 0,
764 entropy = 0,
765 min_sigma_x2 = DBL_MAX,
766 max_sigma_x2 =-DBL_MAX;
767 uint8_t depth[4];
768 int c;
769
770 mask[0] = 0;
771 mask[1] = 0;
772 mask[2] =~0;
773 mask[3] = 0;
774
775 for (c = 0; c < s->nb_channels; c++) {
776 ChannelStats *p = &s->chstats[c];
777
778 if (p->nb_samples == 0 && !s->used)
779 continue;
780
781 if (p->nb_samples < s->tc_samples)
782 p->min_sigma_x2 = p->max_sigma_x2 = p->sigma_x2 / p->nb_samples;
783
784 min = FFMIN(min, p->min);
785 max = FFMAX(max, p->max);
786 nmin = FFMIN(nmin, p->nmin);
787 nmax = FFMAX(nmax, p->nmax);
788 min_diff = FFMIN(min_diff, p->min_diff);
789 max_diff = FFMAX(max_diff, p->max_diff);
790 diff1_sum_x2 += p->diff1_sum_x2;
791 diff1_sum += p->diff1_sum;
792 min_sigma_x2 = FFMIN(min_sigma_x2, p->min_sigma_x2);
793 max_sigma_x2 = FFMAX(max_sigma_x2, p->max_sigma_x2);
794 sigma_x2 += p->sigma_x2;
795 noise_floor = FFMAX(noise_floor, p->noise_floor);
796 p->entropy = calc_entropy(s, p);
797 entropy += p->entropy;
798 min_count += p->min_count;
799 max_count += p->max_count;
800 abs_peak_count += p->abs_peak_count;
801 noise_floor_count += p->noise_floor_count;
802 min_runs += p->min_runs;
803 max_runs += p->max_runs;
804 mask[0] |= p->mask[0];
805 mask[1] |= p->mask[1];
806 mask[2] &= p->mask[2];
807 mask[3] |= p->mask[3];
808 nb_samples += p->nb_samples;
809 nb_nans += p->nb_nans;
810 nb_infs += p->nb_infs;
811 nb_denormals += p->nb_denormals;
812 if (fabs(p->sigma_x) > fabs(max_sigma_x))
813 max_sigma_x = p->sigma_x;
814
815 if (s->measure_perchannel != MEASURE_NONE)
816 av_log(ctx, AV_LOG_INFO, "Channel: %d\n", c + 1);
817 if (s->measure_perchannel & MEASURE_DC_OFFSET)
818 av_log(ctx, AV_LOG_INFO, "DC offset: %f\n", p->sigma_x / p->nb_samples);
819 if (s->measure_perchannel & MEASURE_MIN_LEVEL)
820 av_log(ctx, AV_LOG_INFO, "Min level: %f\n", p->min);
821 if (s->measure_perchannel & MEASURE_MAX_LEVEL)
822 av_log(ctx, AV_LOG_INFO, "Max level: %f\n", p->max);
823 if (s->measure_perchannel & MEASURE_MIN_DIFFERENCE)
824 av_log(ctx, AV_LOG_INFO, "Min difference: %f\n", p->min_diff);
825 if (s->measure_perchannel & MEASURE_MAX_DIFFERENCE)
826 av_log(ctx, AV_LOG_INFO, "Max difference: %f\n", p->max_diff);
827 if (s->measure_perchannel & MEASURE_MEAN_DIFFERENCE)
828 av_log(ctx, AV_LOG_INFO, "Mean difference: %f\n", p->diff1_sum / (p->nb_samples - 1));
829 if (s->measure_perchannel & MEASURE_RMS_DIFFERENCE)
830 av_log(ctx, AV_LOG_INFO, "RMS difference: %f\n", sqrt(p->diff1_sum_x2 / (p->nb_samples - 1)));
831 if (s->measure_perchannel & MEASURE_PEAK_LEVEL)
832 av_log(ctx, AV_LOG_INFO, "Peak level dB: %f\n", LINEAR_TO_DB(FFMAX(-p->nmin, p->nmax)));
833 if (s->measure_perchannel & MEASURE_RMS_LEVEL)
834 av_log(ctx, AV_LOG_INFO, "RMS level dB: %f\n", LINEAR_TO_DB(sqrt(p->sigma_x2 / p->nb_samples)));
835 if (s->measure_perchannel & MEASURE_RMS_PEAK)
836 av_log(ctx, AV_LOG_INFO, "RMS peak dB: %f\n", LINEAR_TO_DB(sqrt(p->max_sigma_x2)));
837 if (s->measure_perchannel & MEASURE_RMS_TROUGH)
838 if (p->min_sigma_x2 != 1)
839 av_log(ctx, AV_LOG_INFO, "RMS trough dB: %f\n",LINEAR_TO_DB(sqrt(p->min_sigma_x2)));
840 if (s->measure_perchannel & MEASURE_CREST_FACTOR)
841 av_log(ctx, AV_LOG_INFO, "Crest factor: %f\n", p->sigma_x2 ? FFMAX(-p->nmin, p->nmax) / sqrt(p->sigma_x2 / p->nb_samples) : 1);
842 if (s->measure_perchannel & MEASURE_FLAT_FACTOR)
843 av_log(ctx, AV_LOG_INFO, "Flat factor: %f\n", LINEAR_TO_DB((p->min_runs + p->max_runs) / (p->min_count + p->max_count)));
844 if (s->measure_perchannel & MEASURE_PEAK_COUNT)
845 av_log(ctx, AV_LOG_INFO, "Peak count: %"PRId64"\n", p->min_count + p->max_count);
846 if (s->measure_perchannel & MEASURE_ABS_PEAK_COUNT)
847 av_log(ctx, AV_LOG_INFO, "Abs Peak count: %"PRId64"\n", p->abs_peak_count);
848 if (s->measure_perchannel & MEASURE_NOISE_FLOOR)
849 av_log(ctx, AV_LOG_INFO, "Noise floor dB: %f\n", LINEAR_TO_DB(p->noise_floor));
850 if (s->measure_perchannel & MEASURE_NOISE_FLOOR_COUNT)
851 av_log(ctx, AV_LOG_INFO, "Noise floor count: %"PRId64"\n", p->noise_floor_count);
852 if (s->measure_perchannel & MEASURE_ENTROPY)
853 av_log(ctx, AV_LOG_INFO, "Entropy: %f\n", p->entropy);
854 if (s->measure_perchannel & MEASURE_BIT_DEPTH) {
855 bit_depth(s, p->mask, depth);
856 av_log(ctx, AV_LOG_INFO, "Bit depth: %u/%u/%u/%u\n", depth[0], depth[1], depth[2], depth[3]);
857 }
858 if (s->measure_perchannel & MEASURE_DYNAMIC_RANGE)
859 av_log(ctx, AV_LOG_INFO, "Dynamic range: %f\n", LINEAR_TO_DB(2 * FFMAX(FFABS(p->min), FFABS(p->max))/ p->min_non_zero));
860 if (s->measure_perchannel & MEASURE_ZERO_CROSSINGS)
861 av_log(ctx, AV_LOG_INFO, "Zero crossings: %"PRId64"\n", p->zero_runs);
862 if (s->measure_perchannel & MEASURE_ZERO_CROSSINGS_RATE)
863 av_log(ctx, AV_LOG_INFO, "Zero crossings rate: %f\n", p->zero_runs/(double)p->nb_samples);
864 if ((s->is_float || s->is_double) && s->measure_perchannel & MEASURE_NUMBER_OF_NANS)
865 av_log(ctx, AV_LOG_INFO, "Number of NaNs: %"PRId64"\n", p->nb_nans);
866 if ((s->is_float || s->is_double) && s->measure_perchannel & MEASURE_NUMBER_OF_INFS)
867 av_log(ctx, AV_LOG_INFO, "Number of Infs: %"PRId64"\n", p->nb_infs);
868 if ((s->is_float || s->is_double) && s->measure_perchannel & MEASURE_NUMBER_OF_DENORMALS)
869 av_log(ctx, AV_LOG_INFO, "Number of denormals: %"PRId64"\n", p->nb_denormals);
870 }
871
872 if (nb_samples == 0 && !s->used)
873 return;
874
875 if (s->measure_overall != MEASURE_NONE)
876 av_log(ctx, AV_LOG_INFO, "Overall\n");
877 if (s->measure_overall & MEASURE_DC_OFFSET)
878 av_log(ctx, AV_LOG_INFO, "DC offset: %f\n", max_sigma_x / (nb_samples / s->nb_channels));
879 if (s->measure_overall & MEASURE_MIN_LEVEL)
880 av_log(ctx, AV_LOG_INFO, "Min level: %f\n", min);
881 if (s->measure_overall & MEASURE_MAX_LEVEL)
882 av_log(ctx, AV_LOG_INFO, "Max level: %f\n", max);
883 if (s->measure_overall & MEASURE_MIN_DIFFERENCE)
884 av_log(ctx, AV_LOG_INFO, "Min difference: %f\n", min_diff);
885 if (s->measure_overall & MEASURE_MAX_DIFFERENCE)
886 av_log(ctx, AV_LOG_INFO, "Max difference: %f\n", max_diff);
887 if (s->measure_overall & MEASURE_MEAN_DIFFERENCE)
888 av_log(ctx, AV_LOG_INFO, "Mean difference: %f\n", diff1_sum / (nb_samples - s->nb_channels));
889 if (s->measure_overall & MEASURE_RMS_DIFFERENCE)
890 av_log(ctx, AV_LOG_INFO, "RMS difference: %f\n", sqrt(diff1_sum_x2 / (nb_samples - s->nb_channels)));
891 if (s->measure_overall & MEASURE_PEAK_LEVEL)
892 av_log(ctx, AV_LOG_INFO, "Peak level dB: %f\n", LINEAR_TO_DB(FFMAX(-nmin, nmax)));
893 if (s->measure_overall & MEASURE_RMS_LEVEL)
894 av_log(ctx, AV_LOG_INFO, "RMS level dB: %f\n", LINEAR_TO_DB(sqrt(sigma_x2 / nb_samples)));
895 if (s->measure_overall & MEASURE_RMS_PEAK)
896 av_log(ctx, AV_LOG_INFO, "RMS peak dB: %f\n", LINEAR_TO_DB(sqrt(max_sigma_x2)));
897 if (s->measure_overall & MEASURE_RMS_TROUGH)
898 if (min_sigma_x2 != 1)
899 av_log(ctx, AV_LOG_INFO, "RMS trough dB: %f\n", LINEAR_TO_DB(sqrt(min_sigma_x2)));
900 if (s->measure_overall & MEASURE_FLAT_FACTOR)
901 av_log(ctx, AV_LOG_INFO, "Flat factor: %f\n", LINEAR_TO_DB((min_runs + max_runs) / (min_count + max_count)));
902 if (s->measure_overall & MEASURE_PEAK_COUNT)
903 av_log(ctx, AV_LOG_INFO, "Peak count: %f\n", (min_count + max_count) / (double)s->nb_channels);
904 if (s->measure_overall & MEASURE_ABS_PEAK_COUNT)
905 av_log(ctx, AV_LOG_INFO, "Abs Peak count: %f\n", abs_peak_count / (double)s->nb_channels);
906 if (s->measure_overall & MEASURE_NOISE_FLOOR)
907 av_log(ctx, AV_LOG_INFO, "Noise floor dB: %f\n", LINEAR_TO_DB(noise_floor));
908 if (s->measure_overall & MEASURE_NOISE_FLOOR_COUNT)
909 av_log(ctx, AV_LOG_INFO, "Noise floor count: %f\n", noise_floor_count / (double)s->nb_channels);
910 if (s->measure_overall & MEASURE_ENTROPY)
911 av_log(ctx, AV_LOG_INFO, "Entropy: %f\n", entropy / (double)s->nb_channels);
912 if (s->measure_overall & MEASURE_BIT_DEPTH) {
913 bit_depth(s, mask, depth);
914 av_log(ctx, AV_LOG_INFO, "Bit depth: %u/%u/%u/%u\n", depth[0], depth[1], depth[2], depth[3]);
915 }
916 if (s->measure_overall & MEASURE_NUMBER_OF_SAMPLES)
917 av_log(ctx, AV_LOG_INFO, "Number of samples: %"PRId64"\n", nb_samples / s->nb_channels);
918 if ((s->is_float || s->is_double) && s->measure_overall & MEASURE_NUMBER_OF_NANS)
919 av_log(ctx, AV_LOG_INFO, "Number of NaNs: %f\n", nb_nans / (float)s->nb_channels);
920 if ((s->is_float || s->is_double) && s->measure_overall & MEASURE_NUMBER_OF_INFS)
921 av_log(ctx, AV_LOG_INFO, "Number of Infs: %f\n", nb_infs / (float)s->nb_channels);
922 if ((s->is_float || s->is_double) && s->measure_overall & MEASURE_NUMBER_OF_DENORMALS)
923 av_log(ctx, AV_LOG_INFO, "Number of denormals: %f\n", nb_denormals / (float)s->nb_channels);
924 }
925
926 static av_cold void uninit(AVFilterContext *ctx)
927 {
928 AudioStatsContext *s = ctx->priv;
929
930 if (s->nb_channels)
931 print_stats(ctx);
932 if (s->chstats) {
933 for (int i = 0; i < s->nb_channels; i++) {
934 ChannelStats *p = &s->chstats[i];
935
936 av_freep(&p->win_samples);
937 av_freep(&p->sorted_samples);
938 }
939 }
940 av_freep(&s->chstats);
941 }
942
943 static const AVFilterPad astats_inputs[] = {
944 {
945 .name = "default",
946 .type = AVMEDIA_TYPE_AUDIO,
947 .filter_frame = filter_frame,
948 },
949 };
950
951 static const AVFilterPad astats_outputs[] = {
952 {
953 .name = "default",
954 .type = AVMEDIA_TYPE_AUDIO,
955 .config_props = config_output,
956 },
957 };
958
959 const AVFilter ff_af_astats = {
960 .name = "astats",
961 .description = NULL_IF_CONFIG_SMALL("Show time domain statistics about audio frames."),
962 .priv_size = sizeof(AudioStatsContext),
963 .priv_class = &astats_class,
964 .uninit = uninit,
965 FILTER_INPUTS(astats_inputs),
966 FILTER_OUTPUTS(astats_outputs),
967 FILTER_SAMPLEFMTS(AV_SAMPLE_FMT_S16, AV_SAMPLE_FMT_S16P,
968 AV_SAMPLE_FMT_S32, AV_SAMPLE_FMT_S32P,
969 AV_SAMPLE_FMT_S64, AV_SAMPLE_FMT_S64P,
970 AV_SAMPLE_FMT_FLT, AV_SAMPLE_FMT_FLTP,
971 AV_SAMPLE_FMT_DBL, AV_SAMPLE_FMT_DBLP),
972 .flags = AVFILTER_FLAG_SLICE_THREADS | AVFILTER_FLAG_METADATA_ONLY,
973 };
974