FFmpeg coverage


Directory: ../../../ffmpeg/
File: src/libavfilter/af_surround.c
Date: 2024-07-26 21:54:09
Exec Total Coverage
Lines: 0 880 0.0%
Functions: 0 35 0.0%
Branches: 0 294 0.0%

Line Branch Exec Source
1 /*
2 * Copyright (c) 2017 Paul B Mahol
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 #include "libavutil/avassert.h"
22 #include "libavutil/channel_layout.h"
23 #include "libavutil/mem.h"
24 #include "libavutil/opt.h"
25 #include "libavutil/tx.h"
26 #include "avfilter.h"
27 #include "audio.h"
28 #include "filters.h"
29 #include "internal.h"
30 #include "formats.h"
31 #include "window_func.h"
32
33 enum SurroundChannel {
34 SC_FL, SC_FR, SC_FC, SC_LF, SC_BL, SC_BR, SC_BC, SC_SL, SC_SR,
35 SC_NB,
36 };
37
38 static const int ch_map[SC_NB] = {
39 [SC_FL] = AV_CHAN_FRONT_LEFT,
40 [SC_FR] = AV_CHAN_FRONT_RIGHT,
41 [SC_FC] = AV_CHAN_FRONT_CENTER,
42 [SC_LF] = AV_CHAN_LOW_FREQUENCY,
43 [SC_BL] = AV_CHAN_BACK_LEFT,
44 [SC_BR] = AV_CHAN_BACK_RIGHT,
45 [SC_BC] = AV_CHAN_BACK_CENTER,
46 [SC_SL] = AV_CHAN_SIDE_LEFT,
47 [SC_SR] = AV_CHAN_SIDE_RIGHT,
48 };
49
50 static const int sc_map[16] = {
51 [AV_CHAN_FRONT_LEFT ] = SC_FL,
52 [AV_CHAN_FRONT_RIGHT ] = SC_FR,
53 [AV_CHAN_FRONT_CENTER ] = SC_FC,
54 [AV_CHAN_LOW_FREQUENCY] = SC_LF,
55 [AV_CHAN_BACK_LEFT ] = SC_BL,
56 [AV_CHAN_BACK_RIGHT ] = SC_BR,
57 [AV_CHAN_BACK_CENTER ] = SC_BC,
58 [AV_CHAN_SIDE_LEFT ] = SC_SL,
59 [AV_CHAN_SIDE_RIGHT ] = SC_SR,
60 };
61
62 typedef struct AudioSurroundContext {
63 const AVClass *class;
64
65 AVChannelLayout out_ch_layout;
66 AVChannelLayout in_ch_layout;
67
68 float level_in;
69 float level_out;
70 float f_i[SC_NB];
71 float f_o[SC_NB];
72 int lfe_mode;
73 float smooth;
74 float angle;
75 float focus;
76 int win_size;
77 int win_func;
78 float win_gain;
79 float overlap;
80
81 float all_x;
82 float all_y;
83
84 float f_x[SC_NB];
85 float f_y[SC_NB];
86
87 float *input_levels;
88 float *output_levels;
89 int output_lfe;
90 int create_lfe;
91 int lowcutf;
92 int highcutf;
93
94 float lowcut;
95 float highcut;
96
97 int nb_in_channels;
98 int nb_out_channels;
99
100 AVFrame *factors;
101 AVFrame *sfactors;
102 AVFrame *input_in;
103 AVFrame *input;
104 AVFrame *output;
105 AVFrame *output_mag;
106 AVFrame *output_ph;
107 AVFrame *output_out;
108 AVFrame *overlap_buffer;
109 AVFrame *window;
110
111 float *x_pos;
112 float *y_pos;
113 float *l_phase;
114 float *r_phase;
115 float *c_phase;
116 float *c_mag;
117 float *lfe_mag;
118 float *lfe_phase;
119 float *mag_total;
120
121 int rdft_size;
122 int hop_size;
123 AVTXContext **rdft, **irdft;
124 av_tx_fn tx_fn, itx_fn;
125 float *window_func_lut;
126
127 void (*filter)(AVFilterContext *ctx);
128 void (*upmix)(AVFilterContext *ctx, int ch);
129 void (*upmix_5_0)(AVFilterContext *ctx,
130 float c_re, float c_im,
131 float mag_totall, float mag_totalr,
132 float fl_phase, float fr_phase,
133 float bl_phase, float br_phase,
134 float sl_phase, float sr_phase,
135 float xl, float yl,
136 float xr, float yr,
137 int n);
138 void (*upmix_5_1)(AVFilterContext *ctx,
139 float c_re, float c_im,
140 float lfe_re, float lfe_im,
141 float mag_totall, float mag_totalr,
142 float fl_phase, float fr_phase,
143 float bl_phase, float br_phase,
144 float sl_phase, float sr_phase,
145 float xl, float yl,
146 float xr, float yr,
147 int n);
148 } AudioSurroundContext;
149
150 static int query_formats(AVFilterContext *ctx)
151 {
152 AudioSurroundContext *s = ctx->priv;
153 AVFilterFormats *formats = NULL;
154 AVFilterChannelLayouts *layouts = NULL;
155 int ret;
156
157 ret = ff_add_format(&formats, AV_SAMPLE_FMT_FLTP);
158 if (ret)
159 return ret;
160 ret = ff_set_common_formats(ctx, formats);
161 if (ret)
162 return ret;
163
164 layouts = NULL;
165 ret = ff_add_channel_layout(&layouts, &s->out_ch_layout);
166 if (ret)
167 return ret;
168
169 ret = ff_channel_layouts_ref(layouts, &ctx->outputs[0]->incfg.channel_layouts);
170 if (ret)
171 return ret;
172
173 layouts = NULL;
174 ret = ff_add_channel_layout(&layouts, &s->in_ch_layout);
175 if (ret)
176 return ret;
177
178 ret = ff_channel_layouts_ref(layouts, &ctx->inputs[0]->outcfg.channel_layouts);
179 if (ret)
180 return ret;
181
182 return ff_set_common_all_samplerates(ctx);
183 }
184
185 static void set_input_levels(AVFilterContext *ctx)
186 {
187 AudioSurroundContext *s = ctx->priv;
188
189 for (int ch = 0; ch < s->nb_in_channels && s->level_in >= 0.f; ch++)
190 s->input_levels[ch] = s->level_in;
191 s->level_in = -1.f;
192
193 for (int n = 0; n < SC_NB; n++) {
194 const int ch = av_channel_layout_index_from_channel(&s->in_ch_layout, ch_map[n]);
195 if (ch >= 0)
196 s->input_levels[ch] = s->f_i[n];
197 }
198 }
199
200 static void set_output_levels(AVFilterContext *ctx)
201 {
202 AudioSurroundContext *s = ctx->priv;
203
204 for (int ch = 0; ch < s->nb_out_channels && s->level_out >= 0.f; ch++)
205 s->output_levels[ch] = s->level_out;
206 s->level_out = -1.f;
207
208 for (int n = 0; n < SC_NB; n++) {
209 const int ch = av_channel_layout_index_from_channel(&s->out_ch_layout, ch_map[n]);
210 if (ch >= 0)
211 s->output_levels[ch] = s->f_o[n];
212 }
213 }
214
215 static int config_input(AVFilterLink *inlink)
216 {
217 AVFilterContext *ctx = inlink->dst;
218 AudioSurroundContext *s = ctx->priv;
219 int ret;
220
221 s->rdft = av_calloc(inlink->ch_layout.nb_channels, sizeof(*s->rdft));
222 if (!s->rdft)
223 return AVERROR(ENOMEM);
224 s->nb_in_channels = inlink->ch_layout.nb_channels;
225
226 for (int ch = 0; ch < inlink->ch_layout.nb_channels; ch++) {
227 float scale = 1.f;
228
229 ret = av_tx_init(&s->rdft[ch], &s->tx_fn, AV_TX_FLOAT_RDFT,
230 0, s->win_size, &scale, 0);
231 if (ret < 0)
232 return ret;
233 }
234
235 s->input_levels = av_malloc_array(s->nb_in_channels, sizeof(*s->input_levels));
236 if (!s->input_levels)
237 return AVERROR(ENOMEM);
238
239 set_input_levels(ctx);
240
241 s->window = ff_get_audio_buffer(inlink, s->win_size * 2);
242 if (!s->window)
243 return AVERROR(ENOMEM);
244
245 s->input_in = ff_get_audio_buffer(inlink, s->win_size * 2);
246 if (!s->input_in)
247 return AVERROR(ENOMEM);
248
249 s->input = ff_get_audio_buffer(inlink, s->win_size + 2);
250 if (!s->input)
251 return AVERROR(ENOMEM);
252
253 s->lowcut = 1.f * s->lowcutf / (inlink->sample_rate * 0.5) * (s->win_size / 2);
254 s->highcut = 1.f * s->highcutf / (inlink->sample_rate * 0.5) * (s->win_size / 2);
255
256 return 0;
257 }
258
259 static int config_output(AVFilterLink *outlink)
260 {
261 AVFilterContext *ctx = outlink->src;
262 AudioSurroundContext *s = ctx->priv;
263 int ret;
264
265 s->irdft = av_calloc(outlink->ch_layout.nb_channels, sizeof(*s->irdft));
266 if (!s->irdft)
267 return AVERROR(ENOMEM);
268 s->nb_out_channels = outlink->ch_layout.nb_channels;
269
270 for (int ch = 0; ch < outlink->ch_layout.nb_channels; ch++) {
271 float iscale = 1.f;
272
273 ret = av_tx_init(&s->irdft[ch], &s->itx_fn, AV_TX_FLOAT_RDFT,
274 1, s->win_size, &iscale, 0);
275 if (ret < 0)
276 return ret;
277 }
278
279 s->output_levels = av_malloc_array(s->nb_out_channels, sizeof(*s->output_levels));
280 if (!s->output_levels)
281 return AVERROR(ENOMEM);
282
283 set_output_levels(ctx);
284
285 s->factors = ff_get_audio_buffer(outlink, s->win_size + 2);
286 s->sfactors = ff_get_audio_buffer(outlink, s->win_size + 2);
287 s->output_ph = ff_get_audio_buffer(outlink, s->win_size + 2);
288 s->output_mag = ff_get_audio_buffer(outlink, s->win_size + 2);
289 s->output_out = ff_get_audio_buffer(outlink, s->win_size + 2);
290 s->output = ff_get_audio_buffer(outlink, s->win_size + 2);
291 s->overlap_buffer = ff_get_audio_buffer(outlink, s->win_size * 2);
292 if (!s->overlap_buffer || !s->output || !s->output_out || !s->output_mag ||
293 !s->output_ph || !s->factors || !s->sfactors)
294 return AVERROR(ENOMEM);
295
296 s->rdft_size = s->win_size / 2 + 1;
297
298 s->x_pos = av_calloc(s->rdft_size, sizeof(*s->x_pos));
299 s->y_pos = av_calloc(s->rdft_size, sizeof(*s->y_pos));
300 s->l_phase = av_calloc(s->rdft_size, sizeof(*s->l_phase));
301 s->r_phase = av_calloc(s->rdft_size, sizeof(*s->r_phase));
302 s->c_mag = av_calloc(s->rdft_size, sizeof(*s->c_mag));
303 s->c_phase = av_calloc(s->rdft_size, sizeof(*s->c_phase));
304 s->mag_total = av_calloc(s->rdft_size, sizeof(*s->mag_total));
305 s->lfe_mag = av_calloc(s->rdft_size, sizeof(*s->lfe_mag));
306 s->lfe_phase = av_calloc(s->rdft_size, sizeof(*s->lfe_phase));
307 if (!s->x_pos || !s->y_pos || !s->l_phase || !s->r_phase || !s->lfe_phase ||
308 !s->c_phase || !s->mag_total || !s->lfe_mag || !s->c_mag)
309 return AVERROR(ENOMEM);
310
311 return 0;
312 }
313
314 static float sqrf(float x)
315 {
316 return x * x;
317 }
318
319 static float r_distance(float a)
320 {
321 return fminf(sqrtf(1.f + sqrf(tanf(a))), sqrtf(1.f + sqrf(1.f / tanf(a))));
322 }
323
324 #define MIN_MAG_SUM 0.00000001f
325
326 static void angle_transform(float *x, float *y, float angle)
327 {
328 float reference, r, a;
329
330 if (angle == 90.f)
331 return;
332
333 reference = angle * M_PIf / 180.f;
334 r = hypotf(*x, *y);
335 a = atan2f(*x, *y);
336
337 r /= r_distance(a);
338
339 if (fabsf(a) <= M_PI_4f)
340 a *= reference / M_PI_2f;
341 else
342 a = M_PIf + (-2.f * M_PIf + reference) * (M_PIf - fabsf(a)) * FFDIFFSIGN(a, 0.f) / (3.f * M_PI_2f);
343
344 r *= r_distance(a);
345
346 *x = av_clipf(sinf(a) * r, -1.f, 1.f);
347 *y = av_clipf(cosf(a) * r, -1.f, 1.f);
348 }
349
350 static void focus_transform(float *x, float *y, float focus)
351 {
352 float a, r, ra;
353
354 if (focus == 0.f)
355 return;
356
357 a = atan2f(*x, *y);
358 ra = r_distance(a);
359 r = av_clipf(hypotf(*x, *y) / ra, 0.f, 1.f);
360 r = focus > 0.f ? 1.f - powf(1.f - r, 1.f + focus * 20.f) : powf(r, 1.f - focus * 20.f);
361 r *= ra;
362 *x = av_clipf(sinf(a) * r, -1.f, 1.f);
363 *y = av_clipf(cosf(a) * r, -1.f, 1.f);
364 }
365
366 static void stereo_position(float a, float p, float *x, float *y)
367 {
368 av_assert2(a >= -1.f && a <= 1.f);
369 av_assert2(p >= 0.f && p <= M_PIf);
370 *x = av_clipf(a+a*fmaxf(0.f, p*p-M_PI_2f), -1.f, 1.f);
371 *y = av_clipf(cosf(a*M_PI_2f+M_PIf)*cosf(M_PI_2f-p/M_PIf)*M_LN10f+1.f, -1.f, 1.f);
372 }
373
374 static inline void get_lfe(int output_lfe, int n, float lowcut, float highcut,
375 float *lfe_mag, float c_mag, float *mag_total, int lfe_mode)
376 {
377 if (output_lfe && n < highcut) {
378 *lfe_mag = n < lowcut ? 1.f : .5f*(1.f+cosf(M_PIf*(lowcut-n)/(lowcut-highcut)));
379 *lfe_mag *= c_mag;
380 if (lfe_mode)
381 *mag_total -= *lfe_mag;
382 } else {
383 *lfe_mag = 0.f;
384 }
385 }
386
387 #define TRANSFORM \
388 dst[2 * n ] = mag * cosf(ph); \
389 dst[2 * n + 1] = mag * sinf(ph);
390
391 static void calculate_factors(AVFilterContext *ctx, int ch, int chan)
392 {
393 AudioSurroundContext *s = ctx->priv;
394 float *factor = (float *)s->factors->extended_data[ch];
395 const float f_x = s->f_x[sc_map[chan >= 0 ? chan : 0]];
396 const float f_y = s->f_y[sc_map[chan >= 0 ? chan : 0]];
397 const int rdft_size = s->rdft_size;
398 const float *x = s->x_pos;
399 const float *y = s->y_pos;
400
401 switch (chan) {
402 case AV_CHAN_FRONT_CENTER:
403 for (int n = 0; n < rdft_size; n++)
404 factor[n] = powf(1.f - fabsf(x[n]), f_x) * powf((y[n] + 1.f) * .5f, f_y);
405 break;
406 case AV_CHAN_FRONT_LEFT:
407 for (int n = 0; n < rdft_size; n++)
408 factor[n] = powf(.5f * ( x[n] + 1.f), f_x) * powf((y[n] + 1.f) * .5f, f_y);
409 break;
410 case AV_CHAN_FRONT_RIGHT:
411 for (int n = 0; n < rdft_size; n++)
412 factor[n] = powf(.5f * (-x[n] + 1.f), f_x) * powf((y[n] + 1.f) * .5f, f_y);
413 break;
414 case AV_CHAN_LOW_FREQUENCY:
415 for (int n = 0; n < rdft_size; n++)
416 factor[n] = powf(1.f - fabsf(x[n]), f_x) * powf((1.f - fabsf(y[n])), f_y);
417 break;
418 case AV_CHAN_BACK_CENTER:
419 for (int n = 0; n < rdft_size; n++)
420 factor[n] = powf(1.f - fabsf(x[n]), f_x) * powf((1.f - y[n]) * .5f, f_y);
421 break;
422 case AV_CHAN_BACK_LEFT:
423 for (int n = 0; n < rdft_size; n++)
424 factor[n] = powf(.5f * ( x[n] + 1.f), f_x) * powf(1.f - ((y[n] + 1.f) * .5f), f_y);
425 break;
426 case AV_CHAN_BACK_RIGHT:
427 for (int n = 0; n < rdft_size; n++)
428 factor[n] = powf(.5f * (-x[n] + 1.f), f_x) * powf(1.f - ((y[n] + 1.f) * .5f), f_y);
429 break;
430 case AV_CHAN_SIDE_LEFT:
431 for (int n = 0; n < rdft_size; n++)
432 factor[n] = powf(.5f * ( x[n] + 1.f), f_x) * powf(1.f - fabsf(y[n]), f_y);
433 break;
434 case AV_CHAN_SIDE_RIGHT:
435 for (int n = 0; n < rdft_size; n++)
436 factor[n] = powf(.5f * (-x[n] + 1.f), f_x) * powf(1.f - fabsf(y[n]), f_y);
437 break;
438 default:
439 for (int n = 0; n < rdft_size; n++)
440 factor[n] = 1.f;
441 break;
442 }
443 }
444
445 static void do_transform(AVFilterContext *ctx, int ch)
446 {
447 AudioSurroundContext *s = ctx->priv;
448 float *sfactor = (float *)s->sfactors->extended_data[ch];
449 float *factor = (float *)s->factors->extended_data[ch];
450 float *omag = (float *)s->output_mag->extended_data[ch];
451 float *oph = (float *)s->output_ph->extended_data[ch];
452 float *dst = (float *)s->output->extended_data[ch];
453 const int rdft_size = s->rdft_size;
454 const float smooth = s->smooth;
455
456 if (smooth > 0.f) {
457 for (int n = 0; n < rdft_size; n++)
458 sfactor[n] = smooth * factor[n] + (1.f - smooth) * sfactor[n];
459
460 factor = sfactor;
461 }
462
463 for (int n = 0; n < rdft_size; n++)
464 omag[n] *= factor[n];
465
466 for (int n = 0; n < rdft_size; n++) {
467 const float mag = omag[n];
468 const float ph = oph[n];
469
470 TRANSFORM
471 }
472 }
473
474 static void stereo_copy(AVFilterContext *ctx, int ch, int chan)
475 {
476 AudioSurroundContext *s = ctx->priv;
477 float *omag = (float *)s->output_mag->extended_data[ch];
478 float *oph = (float *)s->output_ph->extended_data[ch];
479 const float *mag_total = s->mag_total;
480 const int rdft_size = s->rdft_size;
481 const float *c_phase = s->c_phase;
482 const float *l_phase = s->l_phase;
483 const float *r_phase = s->r_phase;
484 const float *lfe_mag = s->lfe_mag;
485 const float *c_mag = s->c_mag;
486
487 switch (chan) {
488 case AV_CHAN_FRONT_CENTER:
489 memcpy(omag, c_mag, rdft_size * sizeof(*omag));
490 break;
491 case AV_CHAN_LOW_FREQUENCY:
492 memcpy(omag, lfe_mag, rdft_size * sizeof(*omag));
493 break;
494 case AV_CHAN_FRONT_LEFT:
495 case AV_CHAN_FRONT_RIGHT:
496 case AV_CHAN_BACK_CENTER:
497 case AV_CHAN_BACK_LEFT:
498 case AV_CHAN_BACK_RIGHT:
499 case AV_CHAN_SIDE_LEFT:
500 case AV_CHAN_SIDE_RIGHT:
501 memcpy(omag, mag_total, rdft_size * sizeof(*omag));
502 break;
503 default:
504 break;
505 }
506
507 switch (chan) {
508 case AV_CHAN_FRONT_CENTER:
509 case AV_CHAN_LOW_FREQUENCY:
510 case AV_CHAN_BACK_CENTER:
511 memcpy(oph, c_phase, rdft_size * sizeof(*oph));
512 break;
513 case AV_CHAN_FRONT_LEFT:
514 case AV_CHAN_BACK_LEFT:
515 case AV_CHAN_SIDE_LEFT:
516 memcpy(oph, l_phase, rdft_size * sizeof(*oph));
517 break;
518 case AV_CHAN_FRONT_RIGHT:
519 case AV_CHAN_BACK_RIGHT:
520 case AV_CHAN_SIDE_RIGHT:
521 memcpy(oph, r_phase, rdft_size * sizeof(*oph));
522 break;
523 default:
524 break;
525 }
526 }
527
528 static void stereo_upmix(AVFilterContext *ctx, int ch)
529 {
530 AudioSurroundContext *s = ctx->priv;
531 const int chan = av_channel_layout_channel_from_index(&s->out_ch_layout, ch);
532
533 calculate_factors(ctx, ch, chan);
534
535 stereo_copy(ctx, ch, chan);
536
537 do_transform(ctx, ch);
538 }
539
540 static void l2_1_upmix(AVFilterContext *ctx, int ch)
541 {
542 AudioSurroundContext *s = ctx->priv;
543 const int chan = av_channel_layout_channel_from_index(&s->out_ch_layout, ch);
544 float *omag = (float *)s->output_mag->extended_data[ch];
545 float *oph = (float *)s->output_ph->extended_data[ch];
546 const float *mag_total = s->mag_total;
547 const float *lfe_phase = s->lfe_phase;
548 const int rdft_size = s->rdft_size;
549 const float *c_phase = s->c_phase;
550 const float *l_phase = s->l_phase;
551 const float *r_phase = s->r_phase;
552 const float *lfe_mag = s->lfe_mag;
553 const float *c_mag = s->c_mag;
554
555 switch (chan) {
556 case AV_CHAN_LOW_FREQUENCY:
557 calculate_factors(ctx, ch, -1);
558 break;
559 default:
560 calculate_factors(ctx, ch, chan);
561 break;
562 }
563
564 switch (chan) {
565 case AV_CHAN_FRONT_CENTER:
566 memcpy(omag, c_mag, rdft_size * sizeof(*omag));
567 break;
568 case AV_CHAN_LOW_FREQUENCY:
569 memcpy(omag, lfe_mag, rdft_size * sizeof(*omag));
570 break;
571 case AV_CHAN_FRONT_LEFT:
572 case AV_CHAN_FRONT_RIGHT:
573 case AV_CHAN_BACK_CENTER:
574 case AV_CHAN_BACK_LEFT:
575 case AV_CHAN_BACK_RIGHT:
576 case AV_CHAN_SIDE_LEFT:
577 case AV_CHAN_SIDE_RIGHT:
578 memcpy(omag, mag_total, rdft_size * sizeof(*omag));
579 break;
580 default:
581 break;
582 }
583
584 switch (chan) {
585 case AV_CHAN_LOW_FREQUENCY:
586 memcpy(oph, lfe_phase, rdft_size * sizeof(*oph));
587 break;
588 case AV_CHAN_FRONT_CENTER:
589 case AV_CHAN_BACK_CENTER:
590 memcpy(oph, c_phase, rdft_size * sizeof(*oph));
591 break;
592 case AV_CHAN_FRONT_LEFT:
593 case AV_CHAN_BACK_LEFT:
594 case AV_CHAN_SIDE_LEFT:
595 memcpy(oph, l_phase, rdft_size * sizeof(*oph));
596 break;
597 case AV_CHAN_FRONT_RIGHT:
598 case AV_CHAN_BACK_RIGHT:
599 case AV_CHAN_SIDE_RIGHT:
600 memcpy(oph, r_phase, rdft_size * sizeof(*oph));
601 break;
602 default:
603 break;
604 }
605
606 do_transform(ctx, ch);
607 }
608
609 static void surround_upmix(AVFilterContext *ctx, int ch)
610 {
611 AudioSurroundContext *s = ctx->priv;
612 const int chan = av_channel_layout_channel_from_index(&s->out_ch_layout, ch);
613
614 switch (chan) {
615 case AV_CHAN_FRONT_CENTER:
616 calculate_factors(ctx, ch, -1);
617 break;
618 default:
619 calculate_factors(ctx, ch, chan);
620 break;
621 }
622
623 stereo_copy(ctx, ch, chan);
624
625 do_transform(ctx, ch);
626 }
627
628 static void upmix_7_1_5_0_side(AVFilterContext *ctx,
629 float c_re, float c_im,
630 float mag_totall, float mag_totalr,
631 float fl_phase, float fr_phase,
632 float bl_phase, float br_phase,
633 float sl_phase, float sr_phase,
634 float xl, float yl,
635 float xr, float yr,
636 int n)
637 {
638 float fl_mag, fr_mag, ls_mag, rs_mag, lb_mag, rb_mag;
639 float *dstc, *dstl, *dstr, *dstls, *dstrs, *dstlb, *dstrb, *dstlfe;
640 float lfe_mag, c_phase, mag_total = (mag_totall + mag_totalr) * 0.5f;
641 AudioSurroundContext *s = ctx->priv;
642
643 dstl = (float *)s->output->extended_data[0];
644 dstr = (float *)s->output->extended_data[1];
645 dstc = (float *)s->output->extended_data[2];
646 dstlfe = (float *)s->output->extended_data[3];
647 dstlb = (float *)s->output->extended_data[4];
648 dstrb = (float *)s->output->extended_data[5];
649 dstls = (float *)s->output->extended_data[6];
650 dstrs = (float *)s->output->extended_data[7];
651
652 c_phase = atan2f(c_im, c_re);
653
654 get_lfe(s->output_lfe, n, s->lowcut, s->highcut, &lfe_mag, hypotf(c_re, c_im), &mag_total, s->lfe_mode);
655
656 fl_mag = powf(.5f * (xl + 1.f), s->f_x[SC_FL]) * powf((yl + 1.f) * .5f, s->f_y[SC_FL]) * mag_totall;
657 fr_mag = powf(.5f * (xr + 1.f), s->f_x[SC_FR]) * powf((yr + 1.f) * .5f, s->f_y[SC_FR]) * mag_totalr;
658 lb_mag = powf(.5f * (-xl + 1.f), s->f_x[SC_BL]) * powf((yl + 1.f) * .5f, s->f_y[SC_BL]) * mag_totall;
659 rb_mag = powf(.5f * (-xr + 1.f), s->f_x[SC_BR]) * powf((yr + 1.f) * .5f, s->f_y[SC_BR]) * mag_totalr;
660 ls_mag = powf(1.f - fabsf(xl), s->f_x[SC_SL]) * powf((yl + 1.f) * .5f, s->f_y[SC_SL]) * mag_totall;
661 rs_mag = powf(1.f - fabsf(xr), s->f_x[SC_SR]) * powf((yr + 1.f) * .5f, s->f_y[SC_SR]) * mag_totalr;
662
663 dstl[2 * n ] = fl_mag * cosf(fl_phase);
664 dstl[2 * n + 1] = fl_mag * sinf(fl_phase);
665
666 dstr[2 * n ] = fr_mag * cosf(fr_phase);
667 dstr[2 * n + 1] = fr_mag * sinf(fr_phase);
668
669 dstc[2 * n ] = c_re;
670 dstc[2 * n + 1] = c_im;
671
672 dstlfe[2 * n ] = lfe_mag * cosf(c_phase);
673 dstlfe[2 * n + 1] = lfe_mag * sinf(c_phase);
674
675 dstlb[2 * n ] = lb_mag * cosf(bl_phase);
676 dstlb[2 * n + 1] = lb_mag * sinf(bl_phase);
677
678 dstrb[2 * n ] = rb_mag * cosf(br_phase);
679 dstrb[2 * n + 1] = rb_mag * sinf(br_phase);
680
681 dstls[2 * n ] = ls_mag * cosf(sl_phase);
682 dstls[2 * n + 1] = ls_mag * sinf(sl_phase);
683
684 dstrs[2 * n ] = rs_mag * cosf(sr_phase);
685 dstrs[2 * n + 1] = rs_mag * sinf(sr_phase);
686 }
687
688 static void upmix_7_1_5_1(AVFilterContext *ctx,
689 float c_re, float c_im,
690 float lfe_re, float lfe_im,
691 float mag_totall, float mag_totalr,
692 float fl_phase, float fr_phase,
693 float bl_phase, float br_phase,
694 float sl_phase, float sr_phase,
695 float xl, float yl,
696 float xr, float yr,
697 int n)
698 {
699 float fl_mag, fr_mag, ls_mag, rs_mag, lb_mag, rb_mag;
700 float *dstc, *dstl, *dstr, *dstls, *dstrs, *dstlb, *dstrb, *dstlfe;
701 AudioSurroundContext *s = ctx->priv;
702
703 dstl = (float *)s->output->extended_data[0];
704 dstr = (float *)s->output->extended_data[1];
705 dstc = (float *)s->output->extended_data[2];
706 dstlfe = (float *)s->output->extended_data[3];
707 dstlb = (float *)s->output->extended_data[4];
708 dstrb = (float *)s->output->extended_data[5];
709 dstls = (float *)s->output->extended_data[6];
710 dstrs = (float *)s->output->extended_data[7];
711
712 fl_mag = powf(.5f * (xl + 1.f), s->f_x[SC_FL]) * powf((yl + 1.f) * .5f, s->f_y[SC_FL]) * mag_totall;
713 fr_mag = powf(.5f * (xr + 1.f), s->f_x[SC_FR]) * powf((yr + 1.f) * .5f, s->f_y[SC_FR]) * mag_totalr;
714 lb_mag = powf(.5f * (-xl + 1.f), s->f_x[SC_BL]) * powf((yl + 1.f) * .5f, s->f_y[SC_BL]) * mag_totall;
715 rb_mag = powf(.5f * (-xr + 1.f), s->f_x[SC_BR]) * powf((yr + 1.f) * .5f, s->f_y[SC_BR]) * mag_totalr;
716 ls_mag = powf(1.f - fabsf(xl), s->f_x[SC_SL]) * powf((yl + 1.f) * .5f, s->f_y[SC_SL]) * mag_totall;
717 rs_mag = powf(1.f - fabsf(xr), s->f_x[SC_SR]) * powf((yr + 1.f) * .5f, s->f_y[SC_SR]) * mag_totalr;
718
719 dstl[2 * n ] = fl_mag * cosf(fl_phase);
720 dstl[2 * n + 1] = fl_mag * sinf(fl_phase);
721
722 dstr[2 * n ] = fr_mag * cosf(fr_phase);
723 dstr[2 * n + 1] = fr_mag * sinf(fr_phase);
724
725 dstc[2 * n ] = c_re;
726 dstc[2 * n + 1] = c_im;
727
728 dstlfe[2 * n ] = lfe_re;
729 dstlfe[2 * n + 1] = lfe_im;
730
731 dstlb[2 * n ] = lb_mag * cosf(bl_phase);
732 dstlb[2 * n + 1] = lb_mag * sinf(bl_phase);
733
734 dstrb[2 * n ] = rb_mag * cosf(br_phase);
735 dstrb[2 * n + 1] = rb_mag * sinf(br_phase);
736
737 dstls[2 * n ] = ls_mag * cosf(sl_phase);
738 dstls[2 * n + 1] = ls_mag * sinf(sl_phase);
739
740 dstrs[2 * n ] = rs_mag * cosf(sr_phase);
741 dstrs[2 * n + 1] = rs_mag * sinf(sr_phase);
742 }
743
744 static void filter_stereo(AVFilterContext *ctx)
745 {
746 AudioSurroundContext *s = ctx->priv;
747 const float *srcl = (const float *)s->input->extended_data[0];
748 const float *srcr = (const float *)s->input->extended_data[1];
749 const int output_lfe = s->output_lfe && s->create_lfe;
750 const int rdft_size = s->rdft_size;
751 const int lfe_mode = s->lfe_mode;
752 const float highcut = s->highcut;
753 const float lowcut = s->lowcut;
754 const float angle = s->angle;
755 const float focus = s->focus;
756 float *magtotal = s->mag_total;
757 float *lfemag = s->lfe_mag;
758 float *lphase = s->l_phase;
759 float *rphase = s->r_phase;
760 float *cphase = s->c_phase;
761 float *cmag = s->c_mag;
762 float *xpos = s->x_pos;
763 float *ypos = s->y_pos;
764
765 for (int n = 0; n < rdft_size; n++) {
766 float l_re = srcl[2 * n], r_re = srcr[2 * n];
767 float l_im = srcl[2 * n + 1], r_im = srcr[2 * n + 1];
768 float c_phase = atan2f(l_im + r_im, l_re + r_re);
769 float l_mag = hypotf(l_re, l_im);
770 float r_mag = hypotf(r_re, r_im);
771 float mag_total = hypotf(l_mag, r_mag);
772 float l_phase = atan2f(l_im, l_re);
773 float r_phase = atan2f(r_im, r_re);
774 float phase_dif = fabsf(l_phase - r_phase);
775 float mag_sum = l_mag + r_mag;
776 float c_mag = mag_sum * 0.5f;
777 float mag_dif, x, y;
778
779 mag_sum = mag_sum < MIN_MAG_SUM ? 1.f : mag_sum;
780 mag_dif = (l_mag - r_mag) / mag_sum;
781 if (phase_dif > M_PIf)
782 phase_dif = 2.f * M_PIf - phase_dif;
783
784 stereo_position(mag_dif, phase_dif, &x, &y);
785 angle_transform(&x, &y, angle);
786 focus_transform(&x, &y, focus);
787 get_lfe(output_lfe, n, lowcut, highcut, &lfemag[n], c_mag, &mag_total, lfe_mode);
788
789 xpos[n] = x;
790 ypos[n] = y;
791 lphase[n] = l_phase;
792 rphase[n] = r_phase;
793 cmag[n] = c_mag;
794 cphase[n] = c_phase;
795 magtotal[n] = mag_total;
796 }
797 }
798
799 static void filter_2_1(AVFilterContext *ctx)
800 {
801 AudioSurroundContext *s = ctx->priv;
802 const float *srcl = (const float *)s->input->extended_data[0];
803 const float *srcr = (const float *)s->input->extended_data[1];
804 const float *srclfe = (const float *)s->input->extended_data[2];
805 const int rdft_size = s->rdft_size;
806 const float angle = s->angle;
807 const float focus = s->focus;
808 float *magtotal = s->mag_total;
809 float *lfephase = s->lfe_phase;
810 float *lfemag = s->lfe_mag;
811 float *lphase = s->l_phase;
812 float *rphase = s->r_phase;
813 float *cphase = s->c_phase;
814 float *cmag = s->c_mag;
815 float *xpos = s->x_pos;
816 float *ypos = s->y_pos;
817
818 for (int n = 0; n < rdft_size; n++) {
819 float l_re = srcl[2 * n], r_re = srcr[2 * n];
820 float l_im = srcl[2 * n + 1], r_im = srcr[2 * n + 1];
821 float lfe_re = srclfe[2 * n], lfe_im = srclfe[2 * n + 1];
822 float c_phase = atan2f(l_im + r_im, l_re + r_re);
823 float l_mag = hypotf(l_re, l_im);
824 float r_mag = hypotf(r_re, r_im);
825 float lfe_mag = hypotf(lfe_re, lfe_im);
826 float lfe_phase = atan2f(lfe_im, lfe_re);
827 float mag_total = hypotf(l_mag, r_mag);
828 float l_phase = atan2f(l_im, l_re);
829 float r_phase = atan2f(r_im, r_re);
830 float phase_dif = fabsf(l_phase - r_phase);
831 float mag_sum = l_mag + r_mag;
832 float c_mag = mag_sum * 0.5f;
833 float mag_dif, x, y;
834
835 mag_sum = mag_sum < MIN_MAG_SUM ? 1.f : mag_sum;
836 mag_dif = (l_mag - r_mag) / mag_sum;
837 if (phase_dif > M_PIf)
838 phase_dif = 2.f * M_PIf - phase_dif;
839
840 stereo_position(mag_dif, phase_dif, &x, &y);
841 angle_transform(&x, &y, angle);
842 focus_transform(&x, &y, focus);
843
844 xpos[n] = x;
845 ypos[n] = y;
846 lphase[n] = l_phase;
847 rphase[n] = r_phase;
848 cmag[n] = c_mag;
849 cphase[n] = c_phase;
850 lfemag[n] = lfe_mag;
851 lfephase[n] = lfe_phase;
852 magtotal[n] = mag_total;
853 }
854 }
855
856 static void filter_surround(AVFilterContext *ctx)
857 {
858 AudioSurroundContext *s = ctx->priv;
859 const float *srcl = (const float *)s->input->extended_data[0];
860 const float *srcr = (const float *)s->input->extended_data[1];
861 const float *srcc = (const float *)s->input->extended_data[2];
862 const int output_lfe = s->output_lfe && s->create_lfe;
863 const int rdft_size = s->rdft_size;
864 const int lfe_mode = s->lfe_mode;
865 const float highcut = s->highcut;
866 const float lowcut = s->lowcut;
867 const float angle = s->angle;
868 const float focus = s->focus;
869 float *magtotal = s->mag_total;
870 float *lfemag = s->lfe_mag;
871 float *lphase = s->l_phase;
872 float *rphase = s->r_phase;
873 float *cphase = s->c_phase;
874 float *cmag = s->c_mag;
875 float *xpos = s->x_pos;
876 float *ypos = s->y_pos;
877
878 for (int n = 0; n < rdft_size; n++) {
879 float l_re = srcl[2 * n], r_re = srcr[2 * n];
880 float l_im = srcl[2 * n + 1], r_im = srcr[2 * n + 1];
881 float c_re = srcc[2 * n], c_im = srcc[2 * n + 1];
882 float c_phase = atan2f(c_im, c_re);
883 float c_mag = hypotf(c_re, c_im);
884 float l_mag = hypotf(l_re, l_im);
885 float r_mag = hypotf(r_re, r_im);
886 float mag_total = hypotf(l_mag, r_mag);
887 float l_phase = atan2f(l_im, l_re);
888 float r_phase = atan2f(r_im, r_re);
889 float phase_dif = fabsf(l_phase - r_phase);
890 float mag_sum = l_mag + r_mag;
891 float mag_dif, x, y;
892
893 mag_sum = mag_sum < MIN_MAG_SUM ? 1.f : mag_sum;
894 mag_dif = (l_mag - r_mag) / mag_sum;
895 if (phase_dif > M_PIf)
896 phase_dif = 2.f * M_PIf - phase_dif;
897
898 stereo_position(mag_dif, phase_dif, &x, &y);
899 angle_transform(&x, &y, angle);
900 focus_transform(&x, &y, focus);
901 get_lfe(output_lfe, n, lowcut, highcut, &lfemag[n], c_mag, &mag_total, lfe_mode);
902
903 xpos[n] = x;
904 ypos[n] = y;
905 lphase[n] = l_phase;
906 rphase[n] = r_phase;
907 cmag[n] = c_mag;
908 cphase[n] = c_phase;
909 magtotal[n] = mag_total;
910 }
911 }
912
913 static void filter_5_0_side(AVFilterContext *ctx)
914 {
915 AudioSurroundContext *s = ctx->priv;
916 const int rdft_size = s->rdft_size;
917 float *srcl, *srcr, *srcc, *srcsl, *srcsr;
918 int n;
919
920 srcl = (float *)s->input->extended_data[0];
921 srcr = (float *)s->input->extended_data[1];
922 srcc = (float *)s->input->extended_data[2];
923 srcsl = (float *)s->input->extended_data[3];
924 srcsr = (float *)s->input->extended_data[4];
925
926 for (n = 0; n < rdft_size; n++) {
927 float fl_re = srcl[2 * n], fr_re = srcr[2 * n];
928 float fl_im = srcl[2 * n + 1], fr_im = srcr[2 * n + 1];
929 float c_re = srcc[2 * n], c_im = srcc[2 * n + 1];
930 float sl_re = srcsl[2 * n], sl_im = srcsl[2 * n + 1];
931 float sr_re = srcsr[2 * n], sr_im = srcsr[2 * n + 1];
932 float fl_mag = hypotf(fl_re, fl_im);
933 float fr_mag = hypotf(fr_re, fr_im);
934 float fl_phase = atan2f(fl_im, fl_re);
935 float fr_phase = atan2f(fr_im, fr_re);
936 float sl_mag = hypotf(sl_re, sl_im);
937 float sr_mag = hypotf(sr_re, sr_im);
938 float sl_phase = atan2f(sl_im, sl_re);
939 float sr_phase = atan2f(sr_im, sr_re);
940 float phase_difl = fabsf(fl_phase - sl_phase);
941 float phase_difr = fabsf(fr_phase - sr_phase);
942 float magl_sum = fl_mag + sl_mag;
943 float magr_sum = fr_mag + sr_mag;
944 float mag_difl = magl_sum < MIN_MAG_SUM ? FFDIFFSIGN(fl_mag, sl_mag) : (fl_mag - sl_mag) / magl_sum;
945 float mag_difr = magr_sum < MIN_MAG_SUM ? FFDIFFSIGN(fr_mag, sr_mag) : (fr_mag - sr_mag) / magr_sum;
946 float mag_totall = hypotf(fl_mag, sl_mag);
947 float mag_totalr = hypotf(fr_mag, sr_mag);
948 float bl_phase = atan2f(fl_im + sl_im, fl_re + sl_re);
949 float br_phase = atan2f(fr_im + sr_im, fr_re + sr_re);
950 float xl, yl;
951 float xr, yr;
952
953 if (phase_difl > M_PIf)
954 phase_difl = 2.f * M_PIf - phase_difl;
955
956 if (phase_difr > M_PIf)
957 phase_difr = 2.f * M_PIf - phase_difr;
958
959 stereo_position(mag_difl, phase_difl, &xl, &yl);
960 stereo_position(mag_difr, phase_difr, &xr, &yr);
961
962 s->upmix_5_0(ctx, c_re, c_im,
963 mag_totall, mag_totalr,
964 fl_phase, fr_phase,
965 bl_phase, br_phase,
966 sl_phase, sr_phase,
967 xl, yl, xr, yr, n);
968 }
969 }
970
971 static void filter_5_1_side(AVFilterContext *ctx)
972 {
973 AudioSurroundContext *s = ctx->priv;
974 const int rdft_size = s->rdft_size;
975 float *srcl, *srcr, *srcc, *srclfe, *srcsl, *srcsr;
976 int n;
977
978 srcl = (float *)s->input->extended_data[0];
979 srcr = (float *)s->input->extended_data[1];
980 srcc = (float *)s->input->extended_data[2];
981 srclfe = (float *)s->input->extended_data[3];
982 srcsl = (float *)s->input->extended_data[4];
983 srcsr = (float *)s->input->extended_data[5];
984
985 for (n = 0; n < rdft_size; n++) {
986 float fl_re = srcl[2 * n], fr_re = srcr[2 * n];
987 float fl_im = srcl[2 * n + 1], fr_im = srcr[2 * n + 1];
988 float c_re = srcc[2 * n], c_im = srcc[2 * n + 1];
989 float lfe_re = srclfe[2 * n], lfe_im = srclfe[2 * n + 1];
990 float sl_re = srcsl[2 * n], sl_im = srcsl[2 * n + 1];
991 float sr_re = srcsr[2 * n], sr_im = srcsr[2 * n + 1];
992 float fl_mag = hypotf(fl_re, fl_im);
993 float fr_mag = hypotf(fr_re, fr_im);
994 float fl_phase = atan2f(fl_im, fl_re);
995 float fr_phase = atan2f(fr_im, fr_re);
996 float sl_mag = hypotf(sl_re, sl_im);
997 float sr_mag = hypotf(sr_re, sr_im);
998 float sl_phase = atan2f(sl_im, sl_re);
999 float sr_phase = atan2f(sr_im, sr_re);
1000 float phase_difl = fabsf(fl_phase - sl_phase);
1001 float phase_difr = fabsf(fr_phase - sr_phase);
1002 float magl_sum = fl_mag + sl_mag;
1003 float magr_sum = fr_mag + sr_mag;
1004 float mag_difl = magl_sum < MIN_MAG_SUM ? FFDIFFSIGN(fl_mag, sl_mag) : (fl_mag - sl_mag) / magl_sum;
1005 float mag_difr = magr_sum < MIN_MAG_SUM ? FFDIFFSIGN(fr_mag, sr_mag) : (fr_mag - sr_mag) / magr_sum;
1006 float mag_totall = hypotf(fl_mag, sl_mag);
1007 float mag_totalr = hypotf(fr_mag, sr_mag);
1008 float bl_phase = atan2f(fl_im + sl_im, fl_re + sl_re);
1009 float br_phase = atan2f(fr_im + sr_im, fr_re + sr_re);
1010 float xl, yl;
1011 float xr, yr;
1012
1013 if (phase_difl > M_PIf)
1014 phase_difl = 2.f * M_PIf - phase_difl;
1015
1016 if (phase_difr > M_PIf)
1017 phase_difr = 2.f * M_PIf - phase_difr;
1018
1019 stereo_position(mag_difl, phase_difl, &xl, &yl);
1020 stereo_position(mag_difr, phase_difr, &xr, &yr);
1021
1022 s->upmix_5_1(ctx, c_re, c_im, lfe_re, lfe_im,
1023 mag_totall, mag_totalr,
1024 fl_phase, fr_phase,
1025 bl_phase, br_phase,
1026 sl_phase, sr_phase,
1027 xl, yl, xr, yr, n);
1028 }
1029 }
1030
1031 static void filter_5_1_back(AVFilterContext *ctx)
1032 {
1033 AudioSurroundContext *s = ctx->priv;
1034 const int rdft_size = s->rdft_size;
1035 float *srcl, *srcr, *srcc, *srclfe, *srcbl, *srcbr;
1036 int n;
1037
1038 srcl = (float *)s->input->extended_data[0];
1039 srcr = (float *)s->input->extended_data[1];
1040 srcc = (float *)s->input->extended_data[2];
1041 srclfe = (float *)s->input->extended_data[3];
1042 srcbl = (float *)s->input->extended_data[4];
1043 srcbr = (float *)s->input->extended_data[5];
1044
1045 for (n = 0; n < rdft_size; n++) {
1046 float fl_re = srcl[2 * n], fr_re = srcr[2 * n];
1047 float fl_im = srcl[2 * n + 1], fr_im = srcr[2 * n + 1];
1048 float c_re = srcc[2 * n], c_im = srcc[2 * n + 1];
1049 float lfe_re = srclfe[2 * n], lfe_im = srclfe[2 * n + 1];
1050 float bl_re = srcbl[2 * n], bl_im = srcbl[2 * n + 1];
1051 float br_re = srcbr[2 * n], br_im = srcbr[2 * n + 1];
1052 float fl_mag = hypotf(fl_re, fl_im);
1053 float fr_mag = hypotf(fr_re, fr_im);
1054 float fl_phase = atan2f(fl_im, fl_re);
1055 float fr_phase = atan2f(fr_im, fr_re);
1056 float bl_mag = hypotf(bl_re, bl_im);
1057 float br_mag = hypotf(br_re, br_im);
1058 float bl_phase = atan2f(bl_im, bl_re);
1059 float br_phase = atan2f(br_im, br_re);
1060 float phase_difl = fabsf(fl_phase - bl_phase);
1061 float phase_difr = fabsf(fr_phase - br_phase);
1062 float magl_sum = fl_mag + bl_mag;
1063 float magr_sum = fr_mag + br_mag;
1064 float mag_difl = magl_sum < MIN_MAG_SUM ? FFDIFFSIGN(fl_mag, bl_mag) : (fl_mag - bl_mag) / magl_sum;
1065 float mag_difr = magr_sum < MIN_MAG_SUM ? FFDIFFSIGN(fr_mag, br_mag) : (fr_mag - br_mag) / magr_sum;
1066 float mag_totall = hypotf(fl_mag, bl_mag);
1067 float mag_totalr = hypotf(fr_mag, br_mag);
1068 float sl_phase = atan2f(fl_im + bl_im, fl_re + bl_re);
1069 float sr_phase = atan2f(fr_im + br_im, fr_re + br_re);
1070 float xl, yl;
1071 float xr, yr;
1072
1073 if (phase_difl > M_PIf)
1074 phase_difl = 2.f * M_PIf - phase_difl;
1075
1076 if (phase_difr > M_PIf)
1077 phase_difr = 2.f * M_PIf - phase_difr;
1078
1079 stereo_position(mag_difl, phase_difl, &xl, &yl);
1080 stereo_position(mag_difr, phase_difr, &xr, &yr);
1081
1082 s->upmix_5_1(ctx, c_re, c_im, lfe_re, lfe_im,
1083 mag_totall, mag_totalr,
1084 fl_phase, fr_phase,
1085 bl_phase, br_phase,
1086 sl_phase, sr_phase,
1087 xl, yl, xr, yr, n);
1088 }
1089 }
1090
1091 static void allchannels_spread(AVFilterContext *ctx)
1092 {
1093 AudioSurroundContext *s = ctx->priv;
1094
1095 if (s->all_x >= 0.f)
1096 for (int n = 0; n < SC_NB; n++)
1097 s->f_x[n] = s->all_x;
1098 s->all_x = -1.f;
1099 if (s->all_y >= 0.f)
1100 for (int n = 0; n < SC_NB; n++)
1101 s->f_y[n] = s->all_y;
1102 s->all_y = -1.f;
1103 }
1104
1105 static av_cold int init(AVFilterContext *ctx)
1106 {
1107 AudioSurroundContext *s = ctx->priv;
1108 int64_t in_channel_layout, out_channel_layout;
1109 char in_name[128], out_name[128];
1110 float overlap;
1111
1112 if (s->lowcutf >= s->highcutf) {
1113 av_log(ctx, AV_LOG_ERROR, "Low cut-off '%d' should be less than high cut-off '%d'.\n",
1114 s->lowcutf, s->highcutf);
1115 return AVERROR(EINVAL);
1116 }
1117
1118 in_channel_layout = s->in_ch_layout.order == AV_CHANNEL_ORDER_NATIVE ?
1119 s->in_ch_layout.u.mask : 0;
1120 out_channel_layout = s->out_ch_layout.order == AV_CHANNEL_ORDER_NATIVE ?
1121 s->out_ch_layout.u.mask : 0;
1122
1123 s->create_lfe = av_channel_layout_index_from_channel(&s->out_ch_layout,
1124 AV_CHAN_LOW_FREQUENCY) >= 0;
1125
1126 switch (in_channel_layout) {
1127 case AV_CH_LAYOUT_STEREO:
1128 s->filter = filter_stereo;
1129 s->upmix = stereo_upmix;
1130 break;
1131 case AV_CH_LAYOUT_2POINT1:
1132 s->filter = filter_2_1;
1133 s->upmix = l2_1_upmix;
1134 break;
1135 case AV_CH_LAYOUT_SURROUND:
1136 s->filter = filter_surround;
1137 s->upmix = surround_upmix;
1138 break;
1139 case AV_CH_LAYOUT_5POINT0:
1140 s->filter = filter_5_0_side;
1141 switch (out_channel_layout) {
1142 case AV_CH_LAYOUT_7POINT1:
1143 s->upmix_5_0 = upmix_7_1_5_0_side;
1144 break;
1145 default:
1146 goto fail;
1147 }
1148 break;
1149 case AV_CH_LAYOUT_5POINT1:
1150 s->filter = filter_5_1_side;
1151 switch (out_channel_layout) {
1152 case AV_CH_LAYOUT_7POINT1:
1153 s->upmix_5_1 = upmix_7_1_5_1;
1154 break;
1155 default:
1156 goto fail;
1157 }
1158 break;
1159 case AV_CH_LAYOUT_5POINT1_BACK:
1160 s->filter = filter_5_1_back;
1161 switch (out_channel_layout) {
1162 case AV_CH_LAYOUT_7POINT1:
1163 s->upmix_5_1 = upmix_7_1_5_1;
1164 break;
1165 default:
1166 goto fail;
1167 }
1168 break;
1169 default:
1170 fail:
1171 av_channel_layout_describe(&s->out_ch_layout, out_name, sizeof(out_name));
1172 av_channel_layout_describe(&s->in_ch_layout, in_name, sizeof(in_name));
1173 av_log(ctx, AV_LOG_ERROR, "Unsupported upmix: '%s' -> '%s'.\n",
1174 in_name, out_name);
1175 return AVERROR(EINVAL);
1176 }
1177
1178 s->window_func_lut = av_calloc(s->win_size, sizeof(*s->window_func_lut));
1179 if (!s->window_func_lut)
1180 return AVERROR(ENOMEM);
1181
1182 generate_window_func(s->window_func_lut, s->win_size, s->win_func, &overlap);
1183 if (s->overlap == 1)
1184 s->overlap = overlap;
1185
1186 for (int i = 0; i < s->win_size; i++)
1187 s->window_func_lut[i] = sqrtf(s->window_func_lut[i] / s->win_size);
1188 s->hop_size = FFMAX(1, s->win_size * (1. - s->overlap));
1189
1190 {
1191 float max = 0.f, *temp_lut = av_calloc(s->win_size, sizeof(*temp_lut));
1192 if (!temp_lut)
1193 return AVERROR(ENOMEM);
1194
1195 for (int j = 0; j < s->win_size; j += s->hop_size) {
1196 for (int i = 0; i < s->win_size; i++)
1197 temp_lut[(i + j) % s->win_size] += s->window_func_lut[i];
1198 }
1199
1200 for (int i = 0; i < s->win_size; i++)
1201 max = fmaxf(temp_lut[i], max);
1202 av_freep(&temp_lut);
1203
1204 s->win_gain = 1.f / (max * sqrtf(s->win_size));
1205 }
1206
1207 allchannels_spread(ctx);
1208
1209 return 0;
1210 }
1211
1212 static int fft_channel(AVFilterContext *ctx, AVFrame *in, int ch)
1213 {
1214 AudioSurroundContext *s = ctx->priv;
1215 float *src = (float *)s->input_in->extended_data[ch];
1216 float *win = (float *)s->window->extended_data[ch];
1217 const float *window_func_lut = s->window_func_lut;
1218 const int offset = s->win_size - s->hop_size;
1219 const float level_in = s->input_levels[ch];
1220 const int win_size = s->win_size;
1221
1222 memmove(src, &src[s->hop_size], offset * sizeof(float));
1223 memcpy(&src[offset], in->extended_data[ch], in->nb_samples * sizeof(float));
1224 memset(&src[offset + in->nb_samples], 0, (s->hop_size - in->nb_samples) * sizeof(float));
1225
1226 for (int n = 0; n < win_size; n++)
1227 win[n] = src[n] * window_func_lut[n] * level_in;
1228
1229 s->tx_fn(s->rdft[ch], (float *)s->input->extended_data[ch], win, sizeof(float));
1230
1231 return 0;
1232 }
1233
1234 static int fft_channels(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)
1235 {
1236 AVFrame *in = arg;
1237 const int start = (in->ch_layout.nb_channels * jobnr) / nb_jobs;
1238 const int end = (in->ch_layout.nb_channels * (jobnr+1)) / nb_jobs;
1239
1240 for (int ch = start; ch < end; ch++)
1241 fft_channel(ctx, in, ch);
1242
1243 return 0;
1244 }
1245
1246 static int ifft_channel(AVFilterContext *ctx, AVFrame *out, int ch)
1247 {
1248 AudioSurroundContext *s = ctx->priv;
1249 const float level_out = s->output_levels[ch] * s->win_gain;
1250 const float *window_func_lut = s->window_func_lut;
1251 const int win_size = s->win_size;
1252 float *dst, *ptr;
1253
1254 dst = (float *)s->output_out->extended_data[ch];
1255 ptr = (float *)s->overlap_buffer->extended_data[ch];
1256 s->itx_fn(s->irdft[ch], dst, (float *)s->output->extended_data[ch], sizeof(AVComplexFloat));
1257
1258 memmove(s->overlap_buffer->extended_data[ch],
1259 s->overlap_buffer->extended_data[ch] + s->hop_size * sizeof(float),
1260 s->win_size * sizeof(float));
1261 memset(s->overlap_buffer->extended_data[ch] + s->win_size * sizeof(float),
1262 0, s->hop_size * sizeof(float));
1263
1264 for (int n = 0; n < win_size; n++)
1265 ptr[n] += dst[n] * window_func_lut[n] * level_out;
1266
1267 ptr = (float *)s->overlap_buffer->extended_data[ch];
1268 dst = (float *)out->extended_data[ch];
1269 memcpy(dst, ptr, s->hop_size * sizeof(float));
1270
1271 return 0;
1272 }
1273
1274 static int ifft_channels(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)
1275 {
1276 AudioSurroundContext *s = ctx->priv;
1277 AVFrame *out = arg;
1278 const int start = (out->ch_layout.nb_channels * jobnr) / nb_jobs;
1279 const int end = (out->ch_layout.nb_channels * (jobnr+1)) / nb_jobs;
1280
1281 for (int ch = start; ch < end; ch++) {
1282 if (s->upmix)
1283 s->upmix(ctx, ch);
1284 ifft_channel(ctx, out, ch);
1285 }
1286
1287 return 0;
1288 }
1289
1290 static int filter_frame(AVFilterLink *inlink, AVFrame *in)
1291 {
1292 AVFilterContext *ctx = inlink->dst;
1293 AVFilterLink *outlink = ctx->outputs[0];
1294 AudioSurroundContext *s = ctx->priv;
1295 AVFrame *out;
1296
1297 ff_filter_execute(ctx, fft_channels, in, NULL,
1298 FFMIN(inlink->ch_layout.nb_channels,
1299 ff_filter_get_nb_threads(ctx)));
1300
1301 s->filter(ctx);
1302
1303 out = ff_get_audio_buffer(outlink, s->hop_size);
1304 if (!out)
1305 return AVERROR(ENOMEM);
1306
1307 ff_filter_execute(ctx, ifft_channels, out, NULL,
1308 FFMIN(outlink->ch_layout.nb_channels,
1309 ff_filter_get_nb_threads(ctx)));
1310
1311 av_frame_copy_props(out, in);
1312 out->nb_samples = in->nb_samples;
1313
1314 av_frame_free(&in);
1315 return ff_filter_frame(outlink, out);
1316 }
1317
1318 static int activate(AVFilterContext *ctx)
1319 {
1320 AVFilterLink *inlink = ctx->inputs[0];
1321 AVFilterLink *outlink = ctx->outputs[0];
1322 AudioSurroundContext *s = ctx->priv;
1323 AVFrame *in = NULL;
1324 int ret = 0, status;
1325 int64_t pts;
1326
1327 FF_FILTER_FORWARD_STATUS_BACK(outlink, inlink);
1328
1329 ret = ff_inlink_consume_samples(inlink, s->hop_size, s->hop_size, &in);
1330 if (ret < 0)
1331 return ret;
1332
1333 if (ret > 0)
1334 ret = filter_frame(inlink, in);
1335 if (ret < 0)
1336 return ret;
1337
1338 if (ff_inlink_queued_samples(inlink) >= s->hop_size) {
1339 ff_filter_set_ready(ctx, 10);
1340 return 0;
1341 }
1342
1343 if (ff_inlink_acknowledge_status(inlink, &status, &pts)) {
1344 ff_outlink_set_status(outlink, status, pts);
1345 return 0;
1346 }
1347
1348 FF_FILTER_FORWARD_WANTED(outlink, inlink);
1349
1350 return FFERROR_NOT_READY;
1351 }
1352
1353 static av_cold void uninit(AVFilterContext *ctx)
1354 {
1355 AudioSurroundContext *s = ctx->priv;
1356
1357 av_frame_free(&s->factors);
1358 av_frame_free(&s->sfactors);
1359 av_frame_free(&s->window);
1360 av_frame_free(&s->input_in);
1361 av_frame_free(&s->input);
1362 av_frame_free(&s->output);
1363 av_frame_free(&s->output_ph);
1364 av_frame_free(&s->output_mag);
1365 av_frame_free(&s->output_out);
1366 av_frame_free(&s->overlap_buffer);
1367
1368 for (int ch = 0; ch < s->nb_in_channels; ch++)
1369 av_tx_uninit(&s->rdft[ch]);
1370 for (int ch = 0; ch < s->nb_out_channels; ch++)
1371 av_tx_uninit(&s->irdft[ch]);
1372 av_freep(&s->input_levels);
1373 av_freep(&s->output_levels);
1374 av_freep(&s->rdft);
1375 av_freep(&s->irdft);
1376 av_freep(&s->window_func_lut);
1377
1378 av_freep(&s->x_pos);
1379 av_freep(&s->y_pos);
1380 av_freep(&s->l_phase);
1381 av_freep(&s->r_phase);
1382 av_freep(&s->c_mag);
1383 av_freep(&s->c_phase);
1384 av_freep(&s->mag_total);
1385 av_freep(&s->lfe_mag);
1386 av_freep(&s->lfe_phase);
1387 }
1388
1389 static int process_command(AVFilterContext *ctx, const char *cmd, const char *args,
1390 char *res, int res_len, int flags)
1391 {
1392 AudioSurroundContext *s = ctx->priv;
1393 int ret;
1394
1395 ret = ff_filter_process_command(ctx, cmd, args, res, res_len, flags);
1396 if (ret < 0)
1397 return ret;
1398
1399 s->hop_size = FFMAX(1, s->win_size * (1. - s->overlap));
1400
1401 allchannels_spread(ctx);
1402 set_input_levels(ctx);
1403 set_output_levels(ctx);
1404
1405 return 0;
1406 }
1407
1408 #define OFFSET(x) offsetof(AudioSurroundContext, x)
1409 #define FLAGS AV_OPT_FLAG_AUDIO_PARAM|AV_OPT_FLAG_FILTERING_PARAM
1410 #define TFLAGS AV_OPT_FLAG_AUDIO_PARAM|AV_OPT_FLAG_FILTERING_PARAM|AV_OPT_FLAG_RUNTIME_PARAM
1411
1412 static const AVOption surround_options[] = {
1413 { "chl_out", "set output channel layout", OFFSET(out_ch_layout), AV_OPT_TYPE_CHLAYOUT, {.str="5.1"}, 0, 0, FLAGS },
1414 { "chl_in", "set input channel layout", OFFSET(in_ch_layout), AV_OPT_TYPE_CHLAYOUT, {.str="stereo"},0, 0, FLAGS },
1415 { "level_in", "set input level", OFFSET(level_in), AV_OPT_TYPE_FLOAT, {.dbl=1}, 0, 10, TFLAGS },
1416 { "level_out", "set output level", OFFSET(level_out), AV_OPT_TYPE_FLOAT, {.dbl=1}, 0, 10, TFLAGS },
1417 { "lfe", "output LFE", OFFSET(output_lfe), AV_OPT_TYPE_BOOL, {.i64=1}, 0, 1, TFLAGS },
1418 { "lfe_low", "LFE low cut off", OFFSET(lowcutf), AV_OPT_TYPE_INT, {.i64=128}, 0, 256, FLAGS },
1419 { "lfe_high", "LFE high cut off", OFFSET(highcutf), AV_OPT_TYPE_INT, {.i64=256}, 0, 512, FLAGS },
1420 { "lfe_mode", "set LFE channel mode", OFFSET(lfe_mode), AV_OPT_TYPE_INT, {.i64=0}, 0, 1, TFLAGS, .unit = "lfe_mode" },
1421 { "add", "just add LFE channel", 0, AV_OPT_TYPE_CONST, {.i64=0}, 0, 1, TFLAGS, .unit = "lfe_mode" },
1422 { "sub", "subtract LFE channel with others", 0, AV_OPT_TYPE_CONST, {.i64=1}, 0, 1, TFLAGS, .unit = "lfe_mode" },
1423 { "smooth", "set temporal smoothness strength", OFFSET(smooth), AV_OPT_TYPE_FLOAT, {.dbl=0}, 0, 1, TFLAGS },
1424 { "angle", "set soundfield transform angle", OFFSET(angle), AV_OPT_TYPE_FLOAT, {.dbl=90}, 0, 360, TFLAGS },
1425 { "focus", "set soundfield transform focus", OFFSET(focus), AV_OPT_TYPE_FLOAT, {.dbl=0}, -1, 1, TFLAGS },
1426 { "fc_in", "set front center channel input level", OFFSET(f_i[SC_FC]), AV_OPT_TYPE_FLOAT, {.dbl=1}, 0, 10, TFLAGS },
1427 { "fc_out", "set front center channel output level", OFFSET(f_o[SC_FC]), AV_OPT_TYPE_FLOAT, {.dbl=1}, 0, 10, TFLAGS },
1428 { "fl_in", "set front left channel input level", OFFSET(f_i[SC_FL]), AV_OPT_TYPE_FLOAT, {.dbl=1}, 0, 10, TFLAGS },
1429 { "fl_out", "set front left channel output level", OFFSET(f_o[SC_FL]), AV_OPT_TYPE_FLOAT, {.dbl=1}, 0, 10, TFLAGS },
1430 { "fr_in", "set front right channel input level", OFFSET(f_i[SC_FR]), AV_OPT_TYPE_FLOAT, {.dbl=1}, 0, 10, TFLAGS },
1431 { "fr_out", "set front right channel output level", OFFSET(f_o[SC_FR]), AV_OPT_TYPE_FLOAT, {.dbl=1}, 0, 10, TFLAGS },
1432 { "sl_in", "set side left channel input level", OFFSET(f_i[SC_SL]), AV_OPT_TYPE_FLOAT, {.dbl=1}, 0, 10, TFLAGS },
1433 { "sl_out", "set side left channel output level", OFFSET(f_o[SC_SL]), AV_OPT_TYPE_FLOAT, {.dbl=1}, 0, 10, TFLAGS },
1434 { "sr_in", "set side right channel input level", OFFSET(f_i[SC_SR]), AV_OPT_TYPE_FLOAT, {.dbl=1}, 0, 10, TFLAGS },
1435 { "sr_out", "set side right channel output level", OFFSET(f_o[SC_SR]), AV_OPT_TYPE_FLOAT, {.dbl=1}, 0, 10, TFLAGS },
1436 { "bl_in", "set back left channel input level", OFFSET(f_i[SC_BL]), AV_OPT_TYPE_FLOAT, {.dbl=1}, 0, 10, TFLAGS },
1437 { "bl_out", "set back left channel output level", OFFSET(f_o[SC_BL]), AV_OPT_TYPE_FLOAT, {.dbl=1}, 0, 10, TFLAGS },
1438 { "br_in", "set back right channel input level", OFFSET(f_i[SC_BR]), AV_OPT_TYPE_FLOAT, {.dbl=1}, 0, 10, TFLAGS },
1439 { "br_out", "set back right channel output level", OFFSET(f_o[SC_BR]), AV_OPT_TYPE_FLOAT, {.dbl=1}, 0, 10, TFLAGS },
1440 { "bc_in", "set back center channel input level", OFFSET(f_i[SC_BC]), AV_OPT_TYPE_FLOAT, {.dbl=1}, 0, 10, TFLAGS },
1441 { "bc_out", "set back center channel output level", OFFSET(f_o[SC_BC]), AV_OPT_TYPE_FLOAT, {.dbl=1}, 0, 10, TFLAGS },
1442 { "lfe_in", "set lfe channel input level", OFFSET(f_i[SC_LF]), AV_OPT_TYPE_FLOAT, {.dbl=1}, 0, 10, TFLAGS },
1443 { "lfe_out", "set lfe channel output level", OFFSET(f_o[SC_LF]), AV_OPT_TYPE_FLOAT, {.dbl=1}, 0, 10, TFLAGS },
1444 { "allx", "set all channel's x spread", OFFSET(all_x), AV_OPT_TYPE_FLOAT, {.dbl=-1}, -1, 15, TFLAGS },
1445 { "ally", "set all channel's y spread", OFFSET(all_y), AV_OPT_TYPE_FLOAT, {.dbl=-1}, -1, 15, TFLAGS },
1446 { "fcx", "set front center channel x spread", OFFSET(f_x[SC_FC]), AV_OPT_TYPE_FLOAT, {.dbl=0.5}, .06, 15, TFLAGS },
1447 { "flx", "set front left channel x spread", OFFSET(f_x[SC_FL]), AV_OPT_TYPE_FLOAT, {.dbl=0.5}, .06, 15, TFLAGS },
1448 { "frx", "set front right channel x spread", OFFSET(f_x[SC_FR]), AV_OPT_TYPE_FLOAT, {.dbl=0.5}, .06, 15, TFLAGS },
1449 { "blx", "set back left channel x spread", OFFSET(f_x[SC_BL]), AV_OPT_TYPE_FLOAT, {.dbl=0.5}, .06, 15, TFLAGS },
1450 { "brx", "set back right channel x spread", OFFSET(f_x[SC_BR]), AV_OPT_TYPE_FLOAT, {.dbl=0.5}, .06, 15, TFLAGS },
1451 { "slx", "set side left channel x spread", OFFSET(f_x[SC_SL]), AV_OPT_TYPE_FLOAT, {.dbl=0.5}, .06, 15, TFLAGS },
1452 { "srx", "set side right channel x spread", OFFSET(f_x[SC_SR]), AV_OPT_TYPE_FLOAT, {.dbl=0.5}, .06, 15, TFLAGS },
1453 { "bcx", "set back center channel x spread", OFFSET(f_x[SC_BC]), AV_OPT_TYPE_FLOAT, {.dbl=0.5}, .06, 15, TFLAGS },
1454 { "fcy", "set front center channel y spread", OFFSET(f_y[SC_FC]), AV_OPT_TYPE_FLOAT, {.dbl=0.5}, .06, 15, TFLAGS },
1455 { "fly", "set front left channel y spread", OFFSET(f_y[SC_FL]), AV_OPT_TYPE_FLOAT, {.dbl=0.5}, .06, 15, TFLAGS },
1456 { "fry", "set front right channel y spread", OFFSET(f_y[SC_FR]), AV_OPT_TYPE_FLOAT, {.dbl=0.5}, .06, 15, TFLAGS },
1457 { "bly", "set back left channel y spread", OFFSET(f_y[SC_BL]), AV_OPT_TYPE_FLOAT, {.dbl=0.5}, .06, 15, TFLAGS },
1458 { "bry", "set back right channel y spread", OFFSET(f_y[SC_BR]), AV_OPT_TYPE_FLOAT, {.dbl=0.5}, .06, 15, TFLAGS },
1459 { "sly", "set side left channel y spread", OFFSET(f_y[SC_SL]), AV_OPT_TYPE_FLOAT, {.dbl=0.5}, .06, 15, TFLAGS },
1460 { "sry", "set side right channel y spread", OFFSET(f_y[SC_SR]), AV_OPT_TYPE_FLOAT, {.dbl=0.5}, .06, 15, TFLAGS },
1461 { "bcy", "set back center channel y spread", OFFSET(f_y[SC_BC]), AV_OPT_TYPE_FLOAT, {.dbl=0.5}, .06, 15, TFLAGS },
1462 { "win_size", "set window size", OFFSET(win_size), AV_OPT_TYPE_INT, {.i64=4096},1024,65536,FLAGS },
1463 WIN_FUNC_OPTION("win_func", OFFSET(win_func), FLAGS, WFUNC_HANNING),
1464 { "overlap", "set window overlap", OFFSET(overlap), AV_OPT_TYPE_FLOAT, {.dbl=0.5}, 0, 1, TFLAGS },
1465 { NULL }
1466 };
1467
1468 AVFILTER_DEFINE_CLASS(surround);
1469
1470 static const AVFilterPad inputs[] = {
1471 {
1472 .name = "default",
1473 .type = AVMEDIA_TYPE_AUDIO,
1474 .config_props = config_input,
1475 },
1476 };
1477
1478 static const AVFilterPad outputs[] = {
1479 {
1480 .name = "default",
1481 .type = AVMEDIA_TYPE_AUDIO,
1482 .config_props = config_output,
1483 },
1484 };
1485
1486 const AVFilter ff_af_surround = {
1487 .name = "surround",
1488 .description = NULL_IF_CONFIG_SMALL("Apply audio surround upmix filter."),
1489 .priv_size = sizeof(AudioSurroundContext),
1490 .priv_class = &surround_class,
1491 .init = init,
1492 .uninit = uninit,
1493 .activate = activate,
1494 FILTER_INPUTS(inputs),
1495 FILTER_OUTPUTS(outputs),
1496 FILTER_QUERY_FUNC(query_formats),
1497 .flags = AVFILTER_FLAG_SLICE_THREADS,
1498 .process_command = process_command,
1499 };
1500