1. ffmpeg
  2. Commit dc72d1dde914c16d85673e80bbe3d21967e47deb
Browse code

avfilter: add audio surround upmixer

Signed-off-by: Paul B Mahol <onemda@gmail.com>

Paul B Mahol authored on 2017/05/20 03:12:04
Showing 6 changed files
Changelog
History View file @ dc72d1d
... ... 
@@ -16,6 +16,7 @@ version <next>:
16 16 
 - spec compliant VP9 muxing support in MP4
17 17 
 - remove the libnut muxer/demuxer wrappers
18 18 
 - remove the libschroedinger encoder/decoder wrappers
19 
+- surround audio filter
19 20
20 21 
 version 3.3:
21 22 
 - CrystalHD decoder moved to new decode API
doc/filters.texi
History View file @ dc72d1d
... ... 
@@ -3792,6 +3792,36 @@ channels. Default is 0.3.
3792 3792 
 Set level of input signal of original channel. Default is 0.8.
3793 3793 
 @end table
3794 3794
3795 
+@section surround
3796 
+Apply audio surround upmix filter.
3797 
+
3798 
+This filter allows to produce multichannel output from stereo audio stream.
3799 
+
3800 
+The filter accepts the following options:
3801 
+
3802 
+@table @option
3803 
+@item chl_out
3804 
+Set output channel layout. By default, this is @var{5.1}.
3805 
+
3806 
+See @ref{channel layout syntax,,the Channel Layout section in the ffmpeg-utils(1) manual,ffmpeg-utils}
3807 
+for the required syntax.
3808 
+
3809 
+@item level_in
3810 
+Set input volume level. By default, this is @var{1}.
3811 
+
3812 
+@item level_out
3813 
+Set output volume level. By default, this is @var{1}.
3814 
+
3815 
+@item lfe
3816 
+Enable LFE channel output if output channel layout has it. By default, this is enabled.
3817 
+
3818 
+@item lfe_low
3819 
+Set LFE low cut off frequency. By default, this is @var{128} Hz.
3820 
+
3821 
+@item lfe_high
3822 
+Set LFE high cut off frequency. By default, this is @var{256} Hz.
3823 
+@end table
3824 
+
3795 3825 
 @section treble
3796 3826
3797 3827 
 Boost or cut treble (upper) frequencies of the audio using a two-pole
libavfilter/Makefile
History View file @ dc72d1d
... ... 
@@ -108,6 +108,7 @@ OBJS-$(CONFIG_SILENCEREMOVE_FILTER)          += af_silenceremove.o
108 108 
 OBJS-$(CONFIG_SOFALIZER_FILTER)              += af_sofalizer.o
109 109 
 OBJS-$(CONFIG_STEREOTOOLS_FILTER)            += af_stereotools.o
110 110 
 OBJS-$(CONFIG_STEREOWIDEN_FILTER)            += af_stereowiden.o
111 
+OBJS-$(CONFIG_SURROUND_FILTER)               += af_surround.o
111 112 
 OBJS-$(CONFIG_TREBLE_FILTER)                 += af_biquads.o
112 113 
 OBJS-$(CONFIG_TREMOLO_FILTER)                += af_tremolo.o
113 114 
 OBJS-$(CONFIG_VIBRATO_FILTER)                += af_vibrato.o generate_wave_table.o
libavfilter/af_surround.c
History View file @ dc72d1d
114 115 
new file mode 100644
... ... 
@@ -0,0 +1,835 @@
0 
+/*
1 
+ * Copyright (c) 2017 Paul B Mahol
2 
+ *
3 
+ * This file is part of FFmpeg.
4 
+ *
5 
+ * FFmpeg is free software; you can redistribute it and/or
6 
+ * modify it under the terms of the GNU Lesser General Public
7 
+ * License as published by the Free Software Foundation; either
8 
+ * version 2.1 of the License, or (at your option) any later version.
9 
+ *
10 
+ * FFmpeg is distributed in the hope that it will be useful,
11 
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
12 
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
13 
+ * Lesser General Public License for more details.
14 
+ *
15 
+ * You should have received a copy of the GNU Lesser General Public
16 
+ * License along with FFmpeg; if not, write to the Free Software
17 
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
18 
+ */
19 
+
20 
+#include "libavutil/audio_fifo.h"
21 
+#include "libavutil/channel_layout.h"
22 
+#include "libavutil/opt.h"
23 
+#include "libavcodec/avfft.h"
24 
+#include "avfilter.h"
25 
+#include "audio.h"
26 
+#include "formats.h"
27 
+
28 
+typedef struct AudioSurroundContext {
29 
+    const AVClass *class;
30 
+
31 
+    char *out_channel_layout_str;
32 
+    float level_in;
33 
+    float level_out;
34 
+    int output_lfe;
35 
+    int lowcutf;
36 
+    int highcutf;
37 
+
38 
+    float lowcut;
39 
+    float highcut;
40 
+
41 
+    uint64_t out_channel_layout;
42 
+    int nb_in_channels;
43 
+    int nb_out_channels;
44 
+
45 
+    AVFrame *input;
46 
+    AVFrame *output;
47 
+    AVFrame *overlap_buffer;
48 
+
49 
+    int buf_size;
50 
+    int hop_size;
51 
+    AVAudioFifo *fifo;
52 
+    RDFTContext **rdft, **irdft;
53 
+    float *window_func_lut;
54 
+
55 
+    int64_t pts;
56 
+
57 
+    void (*upmix)(AVFilterContext *ctx,
58 
+                  float l_phase,
59 
+                  float r_phase,
60 
+                  float c_phase,
61 
+                  float mag_total,
62 
+                  float x, float y,
63 
+                  int n);
64 
+} AudioSurroundContext;
65 
+
66 
+static int query_formats(AVFilterContext *ctx)
67 
+{
68 
+    AudioSurroundContext *s = ctx->priv;
69 
+    AVFilterFormats *formats = NULL;
70 
+    AVFilterChannelLayouts *layouts = NULL;
71 
+    int ret;
72 
+
73 
+    ret = ff_add_format(&formats, AV_SAMPLE_FMT_FLTP);
74 
+    if (ret)
75 
+        return ret;
76 
+    ret = ff_set_common_formats(ctx, formats);
77 
+    if (ret)
78 
+        return ret;
79 
+
80 
+    layouts = NULL;
81 
+    ret = ff_add_channel_layout(&layouts, s->out_channel_layout);
82 
+    if (ret)
83 
+        return ret;
84 
+
85 
+    ret = ff_channel_layouts_ref(layouts, &ctx->outputs[0]->in_channel_layouts);
86 
+    if (ret)
87 
+        return ret;
88 
+
89 
+    layouts = NULL;
90 
+    ret = ff_add_channel_layout(&layouts, AV_CH_LAYOUT_STEREO);
91 
+    if (ret)
92 
+        return ret;
93 
+
94 
+    ret = ff_channel_layouts_ref(layouts, &ctx->inputs[0]->out_channel_layouts);
95 
+    if (ret)
96 
+        return ret;
97 
+
98 
+    formats = ff_all_samplerates();
99 
+    if (!formats)
100 
+        return AVERROR(ENOMEM);
101 
+    return ff_set_common_samplerates(ctx, formats);
102 
+}
103 
+
104 
+static int config_input(AVFilterLink *inlink)
105 
+{
106 
+    AVFilterContext *ctx = inlink->dst;
107 
+    AudioSurroundContext *s = ctx->priv;
108 
+    int ch;
109 
+
110 
+    s->rdft = av_calloc(inlink->channels, sizeof(*s->rdft));
111 
+    if (!s->rdft)
112 
+        return AVERROR(ENOMEM);
113 
+
114 
+    for (ch = 0; ch < inlink->channels; ch++) {
115 
+        s->rdft[ch]  = av_rdft_init(ff_log2(s->buf_size), DFT_R2C);
116 
+        if (!s->rdft[ch])
117 
+            return AVERROR(ENOMEM);
118 
+    }
119 
+    s->nb_in_channels = inlink->channels;
120 
+
121 
+    s->input = ff_get_audio_buffer(inlink, s->buf_size * 2);
122 
+    if (!s->input)
123 
+        return AVERROR(ENOMEM);
124 
+
125 
+    s->fifo = av_audio_fifo_alloc(inlink->format, inlink->channels, s->buf_size);
126 
+    if (!s->fifo)
127 
+        return AVERROR(ENOMEM);
128 
+
129 
+    s->lowcut = 1.f * s->lowcutf / (inlink->sample_rate * 0.5) * (s->buf_size / 2);
130 
+    s->highcut = 1.f * s->highcutf / (inlink->sample_rate * 0.5) * (s->buf_size / 2);
131 
+
132 
+    return 0;
133 
+}
134 
+
135 
+static int config_output(AVFilterLink *outlink)
136 
+{
137 
+    AVFilterContext *ctx = outlink->src;
138 
+    AudioSurroundContext *s = ctx->priv;
139 
+    int ch;
140 
+
141 
+    s->irdft = av_calloc(outlink->channels, sizeof(*s->irdft));
142 
+    if (!s->irdft)
143 
+        return AVERROR(ENOMEM);
144 
+
145 
+    for (ch = 0; ch < outlink->channels; ch++) {
146 
+        s->irdft[ch] = av_rdft_init(ff_log2(s->buf_size), IDFT_C2R);
147 
+        if (!s->irdft[ch])
148 
+            return AVERROR(ENOMEM);
149 
+    }
150 
+    s->nb_out_channels = outlink->channels;
151 
+
152 
+    s->output = ff_get_audio_buffer(outlink, s->buf_size * 2);
153 
+    s->overlap_buffer = ff_get_audio_buffer(outlink, s->buf_size * 2);
154 
+    if (!s->overlap_buffer || !s->output)
155 
+        return AVERROR(ENOMEM);
156 
+
157 
+    return 0;
158 
+}
159 
+
160 
+static void stereo_position(float a, float p, float *x, float *y)
161 
+{
162 
+      *x = av_clipf(a+FFMAX(0, sinf(p-M_PI_2))*FFDIFFSIGN(a,0), -1, 1);
163 
+      *y = av_clipf(cosf(a*M_PI_2+M_PI)*cosf(M_PI_2-p/M_PI)*M_LN10+1, -1, 1);
164 
+}
165 
+
166 
+static inline void get_lfe(int output_lfe, int n, float lowcut, float highcut,
167 
+                           float *lfe_mag, float *mag_total)
168 
+{
169 
+    if (output_lfe && n < highcut) {
170 
+        *lfe_mag    = n < lowcut ? 1.f : .5f*(1.f+cosf(M_PI*(lowcut-n)/(lowcut-highcut)));
171 
+        *lfe_mag   *= *mag_total;
172 
+        *mag_total -= *lfe_mag;
173 
+    } else {
174 
+        *lfe_mag = 0.f;
175 
+    }
176 
+}
177 
+
178 
+static void upmix_1_0(AVFilterContext *ctx,
179 
+                      float l_phase,
180 
+                      float r_phase,
181 
+                      float c_phase,
182 
+                      float mag_total,
183 
+                      float x, float y,
184 
+                      int n)
185 
+{
186 
+    AudioSurroundContext *s = ctx->priv;
187 
+    float mag, *dst;
188 
+
189 
+    dst = (float *)s->output->extended_data[0];
190 
+
191 
+    mag = sqrtf(1.f - fabsf(x)) * ((y + 1.f) * .5f) * mag_total;
192 
+
193 
+    dst[2 * n    ] = mag * cosf(c_phase);
194 
+    dst[2 * n + 1] = mag * sinf(c_phase);
195 
+}
196 
+
197 
+static void upmix_stereo(AVFilterContext *ctx,
198 
+                         float l_phase,
199 
+                         float r_phase,
200 
+                         float c_phase,
201 
+                         float mag_total,
202 
+                         float x, float y,
203 
+                         int n)
204 
+{
205 
+    AudioSurroundContext *s = ctx->priv;
206 
+    float l_mag, r_mag, *dstl, *dstr;
207 
+
208 
+    dstl = (float *)s->output->extended_data[0];
209 
+    dstr = (float *)s->output->extended_data[1];
210 
+
211 
+    l_mag = sqrtf(.5f * ( x + 1.f)) * ((y + 1.f) * .5f) * mag_total;
212 
+    r_mag = sqrtf(.5f * (-x + 1.f)) * ((y + 1.f) * .5f) * mag_total;
213 
+
214 
+    dstl[2 * n    ] = l_mag * cosf(l_phase);
215 
+    dstl[2 * n + 1] = l_mag * sinf(l_phase);
216 
+
217 
+    dstr[2 * n    ] = r_mag * cosf(r_phase);
218 
+    dstr[2 * n + 1] = r_mag * sinf(r_phase);
219 
+}
220 
+
221 
+static void upmix_2_1(AVFilterContext *ctx,
222 
+                      float l_phase,
223 
+                      float r_phase,
224 
+                      float c_phase,
225 
+                      float mag_total,
226 
+                      float x, float y,
227 
+                      int n)
228 
+{
229 
+    AudioSurroundContext *s = ctx->priv;
230 
+    float lfe_mag, l_mag, r_mag, *dstl, *dstr, *dstlfe;
231 
+
232 
+    dstl = (float *)s->output->extended_data[0];
233 
+    dstr = (float *)s->output->extended_data[1];
234 
+    dstlfe = (float *)s->output->extended_data[2];
235 
+
236 
+    get_lfe(s->output_lfe, n, s->lowcut, s->highcut, &lfe_mag, &mag_total);
237 
+
238 
+    l_mag = sqrtf(.5f * ( x + 1.f)) * ((y + 1.f) * .5f) * mag_total;
239 
+    r_mag = sqrtf(.5f * (-x + 1.f)) * ((y + 1.f) * .5f) * mag_total;
240 
+
241 
+    dstl[2 * n    ] = l_mag * cosf(l_phase);
242 
+    dstl[2 * n + 1] = l_mag * sinf(l_phase);
243 
+
244 
+    dstr[2 * n    ] = r_mag * cosf(r_phase);
245 
+    dstr[2 * n + 1] = r_mag * sinf(r_phase);
246 
+
247 
+    dstlfe[2 * n    ] = lfe_mag * cosf(c_phase);
248 
+    dstlfe[2 * n + 1] = lfe_mag * sinf(c_phase);
249 
+}
250 
+
251 
+static void upmix_3_0(AVFilterContext *ctx,
252 
+                      float l_phase,
253 
+                      float r_phase,
254 
+                      float c_phase,
255 
+                      float mag_total,
256 
+                      float x, float y,
257 
+                      int n)
258 
+{
259 
+    AudioSurroundContext *s = ctx->priv;
260 
+    float l_mag, r_mag, c_mag, *dstc, *dstl, *dstr;
261 
+
262 
+    dstl = (float *)s->output->extended_data[0];
263 
+    dstr = (float *)s->output->extended_data[1];
264 
+    dstc = (float *)s->output->extended_data[2];
265 
+
266 
+    c_mag = sqrtf(1.f - fabsf(x))   * ((y + 1.f) * .5f) * mag_total;
267 
+    l_mag = sqrtf(.5f * ( x + 1.f)) * ((y + 1.f) * .5f) * mag_total;
268 
+    r_mag = sqrtf(.5f * (-x + 1.f)) * ((y + 1.f) * .5f) * mag_total;
269 
+
270 
+    dstl[2 * n    ] = l_mag * cosf(l_phase);
271 
+    dstl[2 * n + 1] = l_mag * sinf(l_phase);
272 
+
273 
+    dstr[2 * n    ] = r_mag * cosf(r_phase);
274 
+    dstr[2 * n + 1] = r_mag * sinf(r_phase);
275 
+
276 
+    dstc[2 * n    ] = c_mag * cosf(c_phase);
277 
+    dstc[2 * n + 1] = c_mag * sinf(c_phase);
278 
+}
279 
+
280 
+static void upmix_3_1(AVFilterContext *ctx,
281 
+                      float l_phase,
282 
+                      float r_phase,
283 
+                      float c_phase,
284 
+                      float mag_total,
285 
+                      float x, float y,
286 
+                      int n)
287 
+{
288 
+    AudioSurroundContext *s = ctx->priv;
289 
+    float lfe_mag, l_mag, r_mag, c_mag, *dstc, *dstl, *dstr, *dstlfe;
290 
+
291 
+    dstl = (float *)s->output->extended_data[0];
292 
+    dstr = (float *)s->output->extended_data[1];
293 
+    dstc = (float *)s->output->extended_data[2];
294 
+    dstlfe = (float *)s->output->extended_data[3];
295 
+
296 
+    get_lfe(s->output_lfe, n, s->lowcut, s->highcut, &lfe_mag, &mag_total);
297 
+
298 
+    c_mag = sqrtf(1.f - fabsf(x))   * ((y + 1.f) * .5f) * mag_total;
299 
+    l_mag = sqrtf(.5f * ( x + 1.f)) * ((y + 1.f) * .5f) * mag_total;
300 
+    r_mag = sqrtf(.5f * (-x + 1.f)) * ((y + 1.f) * .5f) * mag_total;
301 
+
302 
+    dstl[2 * n    ] = l_mag * cosf(l_phase);
303 
+    dstl[2 * n + 1] = l_mag * sinf(l_phase);
304 
+
305 
+    dstr[2 * n    ] = r_mag * cosf(r_phase);
306 
+    dstr[2 * n + 1] = r_mag * sinf(r_phase);
307 
+
308 
+    dstc[2 * n    ] = c_mag * cosf(c_phase);
309 
+    dstc[2 * n + 1] = c_mag * sinf(c_phase);
310 
+
311 
+    dstlfe[2 * n    ] = lfe_mag * cosf(c_phase);
312 
+    dstlfe[2 * n + 1] = lfe_mag * sinf(c_phase);
313 
+}
314 
+
315 
+static void upmix_4_0(AVFilterContext *ctx,
316 
+                      float l_phase,
317 
+                      float r_phase,
318 
+                      float c_phase,
319 
+                      float mag_total,
320 
+                      float x, float y,
321 
+                      int n)
322 
+{
323 
+    float b_mag, l_mag, r_mag, c_mag, *dstc, *dstl, *dstr, *dstb;
324 
+    AudioSurroundContext *s = ctx->priv;
325 
+
326 
+    dstl = (float *)s->output->extended_data[0];
327 
+    dstr = (float *)s->output->extended_data[1];
328 
+    dstc = (float *)s->output->extended_data[2];
329 
+    dstb = (float *)s->output->extended_data[3];
330 
+
331 
+    c_mag = sqrtf(1.f - fabsf(x))   * ((y + 1.f) * .5f) * mag_total;
332 
+    b_mag = sqrtf(1.f - fabsf(x))   * ((1.f - y) * .5f) * mag_total;
333 
+    l_mag = sqrtf(.5f * ( x + 1.f)) * ((y + 1.f) * .5f) * mag_total;
334 
+    r_mag = sqrtf(.5f * (-x + 1.f)) * ((y + 1.f) * .5f) * mag_total;
335 
+
336 
+    dstl[2 * n    ] = l_mag * cosf(l_phase);
337 
+    dstl[2 * n + 1] = l_mag * sinf(l_phase);
338 
+
339 
+    dstr[2 * n    ] = r_mag * cosf(r_phase);
340 
+    dstr[2 * n + 1] = r_mag * sinf(r_phase);
341 
+
342 
+    dstc[2 * n    ] = c_mag * cosf(c_phase);
343 
+    dstc[2 * n + 1] = c_mag * sinf(c_phase);
344 
+
345 
+    dstb[2 * n    ] = b_mag * cosf(c_phase);
346 
+    dstb[2 * n + 1] = b_mag * sinf(c_phase);
347 
+}
348 
+
349 
+static void upmix_4_1(AVFilterContext *ctx,
350 
+                      float l_phase,
351 
+                      float r_phase,
352 
+                      float c_phase,
353 
+                      float mag_total,
354 
+                      float x, float y,
355 
+                      int n)
356 
+{
357 
+    float lfe_mag, b_mag, l_mag, r_mag, c_mag, *dstc, *dstl, *dstr, *dstb, *dstlfe;
358 
+    AudioSurroundContext *s = ctx->priv;
359 
+
360 
+    dstl = (float *)s->output->extended_data[0];
361 
+    dstr = (float *)s->output->extended_data[1];
362 
+    dstc = (float *)s->output->extended_data[2];
363 
+    dstlfe = (float *)s->output->extended_data[3];
364 
+    dstb = (float *)s->output->extended_data[4];
365 
+
366 
+    get_lfe(s->output_lfe, n, s->lowcut, s->highcut, &lfe_mag, &mag_total);
367 
+
368 
+    dstlfe[2 * n    ] = lfe_mag * cosf(c_phase);
369 
+    dstlfe[2 * n + 1] = lfe_mag * sinf(c_phase);
370 
+
371 
+    c_mag = sqrtf(1.f - fabsf(x))   * ((y + 1.f) * .5f) * mag_total;
372 
+    b_mag = sqrtf(1.f - fabsf(x))   * ((1.f - y) * .5f) * mag_total;
373 
+    l_mag = sqrtf(.5f * ( x + 1.f)) * ((y + 1.f) * .5f) * mag_total;
374 
+    r_mag = sqrtf(.5f * (-x + 1.f)) * ((y + 1.f) * .5f) * mag_total;
375 
+
376 
+    dstl[2 * n    ] = l_mag * cosf(l_phase);
377 
+    dstl[2 * n + 1] = l_mag * sinf(l_phase);
378 
+
379 
+    dstr[2 * n    ] = r_mag * cosf(r_phase);
380 
+    dstr[2 * n + 1] = r_mag * sinf(r_phase);
381 
+
382 
+    dstc[2 * n    ] = c_mag * cosf(c_phase);
383 
+    dstc[2 * n + 1] = c_mag * sinf(c_phase);
384 
+
385 
+    dstb[2 * n    ] = b_mag * cosf(c_phase);
386 
+    dstb[2 * n + 1] = b_mag * sinf(c_phase);
387 
+}
388 
+
389 
+static void upmix_5_0_back(AVFilterContext *ctx,
390 
+                           float l_phase,
391 
+                           float r_phase,
392 
+                           float c_phase,
393 
+                           float mag_total,
394 
+                           float x, float y,
395 
+                           int n)
396 
+{
397 
+    float l_mag, r_mag, ls_mag, rs_mag, c_mag, *dstc, *dstl, *dstr, *dstls, *dstrs;
398 
+    AudioSurroundContext *s = ctx->priv;
399 
+
400 
+    dstl  = (float *)s->output->extended_data[0];
401 
+    dstr  = (float *)s->output->extended_data[1];
402 
+    dstc  = (float *)s->output->extended_data[2];
403 
+    dstls = (float *)s->output->extended_data[3];
404 
+    dstrs = (float *)s->output->extended_data[4];
405 
+
406 
+    c_mag  = sqrtf(1.f - fabsf(x))   * ((y + 1.f) * .5f) * mag_total;
407 
+    l_mag  = sqrtf(.5f * ( x + 1.f)) * ((y + 1.f) * .5f) * mag_total;
408 
+    r_mag  = sqrtf(.5f * (-x + 1.f)) * ((y + 1.f) * .5f) * mag_total;
409 
+    ls_mag = sqrtf(.5f * ( x + 1.f)) * (1.f - ((y + 1.f) * .5f)) * mag_total;
410 
+    rs_mag = sqrtf(.5f * (-x + 1.f)) * (1.f - ((y + 1.f) * .5f)) * mag_total;
411 
+
412 
+    dstl[2 * n    ] = l_mag * cosf(l_phase);
413 
+    dstl[2 * n + 1] = l_mag * sinf(l_phase);
414 
+
415 
+    dstr[2 * n    ] = r_mag * cosf(r_phase);
416 
+    dstr[2 * n + 1] = r_mag * sinf(r_phase);
417 
+
418 
+    dstc[2 * n    ] = c_mag * cosf(c_phase);
419 
+    dstc[2 * n + 1] = c_mag * sinf(c_phase);
420 
+
421 
+    dstls[2 * n    ] = ls_mag * cosf(l_phase);
422 
+    dstls[2 * n + 1] = ls_mag * sinf(l_phase);
423 
+
424 
+    dstrs[2 * n    ] = rs_mag * cosf(r_phase);
425 
+    dstrs[2 * n + 1] = rs_mag * sinf(r_phase);
426 
+}
427 
+
428 
+static void upmix_5_1_back(AVFilterContext *ctx,
429 
+                           float l_phase,
430 
+                           float r_phase,
431 
+                           float c_phase,
432 
+                           float mag_total,
433 
+                           float x, float y,
434 
+                           int n)
435 
+{
436 
+    float lfe_mag, l_mag, r_mag, ls_mag, rs_mag, c_mag, *dstc, *dstl, *dstr, *dstls, *dstrs, *dstlfe;
437 
+    AudioSurroundContext *s = ctx->priv;
438 
+
439 
+    dstl  = (float *)s->output->extended_data[0];
440 
+    dstr  = (float *)s->output->extended_data[1];
441 
+    dstc  = (float *)s->output->extended_data[2];
442 
+    dstlfe = (float *)s->output->extended_data[3];
443 
+    dstls = (float *)s->output->extended_data[4];
444 
+    dstrs = (float *)s->output->extended_data[5];
445 
+
446 
+    get_lfe(s->output_lfe, n, s->lowcut, s->highcut, &lfe_mag, &mag_total);
447 
+
448 
+    c_mag  = sqrtf(1.f - fabsf(x))   * ((y + 1.f) * .5f) * mag_total;
449 
+    l_mag  = sqrtf(.5f * ( x + 1.f)) * ((y + 1.f) * .5f) * mag_total;
450 
+    r_mag  = sqrtf(.5f * (-x + 1.f)) * ((y + 1.f) * .5f) * mag_total;
451 
+    ls_mag = sqrtf(.5f * ( x + 1.f)) * (1.f - ((y + 1.f) * .5f)) * mag_total;
452 
+    rs_mag = sqrtf(.5f * (-x + 1.f)) * (1.f - ((y + 1.f) * .5f)) * mag_total;
453 
+
454 
+    dstl[2 * n    ] = l_mag * cosf(l_phase);
455 
+    dstl[2 * n + 1] = l_mag * sinf(l_phase);
456 
+
457 
+    dstr[2 * n    ] = r_mag * cosf(r_phase);
458 
+    dstr[2 * n + 1] = r_mag * sinf(r_phase);
459 
+
460 
+    dstc[2 * n    ] = c_mag * cosf(c_phase);
461 
+    dstc[2 * n + 1] = c_mag * sinf(c_phase);
462 
+
463 
+    dstlfe[2 * n    ] = lfe_mag * cosf(c_phase);
464 
+    dstlfe[2 * n + 1] = lfe_mag * sinf(c_phase);
465 
+
466 
+    dstls[2 * n    ] = ls_mag * cosf(l_phase);
467 
+    dstls[2 * n + 1] = ls_mag * sinf(l_phase);
468 
+
469 
+    dstrs[2 * n    ] = rs_mag * cosf(r_phase);
470 
+    dstrs[2 * n + 1] = rs_mag * sinf(r_phase);
471 
+}
472 
+
473 
+static void upmix_7_0(AVFilterContext *ctx,
474 
+                      float l_phase,
475 
+                      float r_phase,
476 
+                      float c_phase,
477 
+                      float mag_total,
478 
+                      float x, float y,
479 
+                      int n)
480 
+{
481 
+    float l_mag, r_mag, ls_mag, rs_mag, c_mag, lb_mag, rb_mag;
482 
+    float *dstc, *dstl, *dstr, *dstls, *dstrs, *dstlb, *dstrb;
483 
+    AudioSurroundContext *s = ctx->priv;
484 
+
485 
+    dstl  = (float *)s->output->extended_data[0];
486 
+    dstr  = (float *)s->output->extended_data[1];
487 
+    dstc  = (float *)s->output->extended_data[2];
488 
+    dstlb = (float *)s->output->extended_data[3];
489 
+    dstrb = (float *)s->output->extended_data[4];
490 
+    dstls = (float *)s->output->extended_data[5];
491 
+    dstrs = (float *)s->output->extended_data[6];
492 
+
493 
+    c_mag  = sqrtf(1.f - fabsf(x))   * ((y + 1.f) * .5f) * mag_total;
494 
+    l_mag  = sqrtf(.5f * ( x + 1.f)) * ((y + 1.f) * .5f) * mag_total;
495 
+    r_mag  = sqrtf(.5f * (-x + 1.f)) * ((y + 1.f) * .5f) * mag_total;
496 
+    lb_mag = sqrtf(.5f * ( x + 1.f)) * (1.f - ((y + 1.f) * .5f)) * mag_total;
497 
+    rb_mag = sqrtf(.5f * (-x + 1.f)) * (1.f - ((y + 1.f) * .5f)) * mag_total;
498 
+    ls_mag = sqrtf(.5f * ( x + 1.f)) * (1.f - fabsf(y)) * mag_total;
499 
+    rs_mag = sqrtf(.5f * (-x + 1.f)) * (1.f - fabsf(y)) * mag_total;
500 
+
501 
+    dstl[2 * n    ] = l_mag * cosf(l_phase);
502 
+    dstl[2 * n + 1] = l_mag * sinf(l_phase);
503 
+
504 
+    dstr[2 * n    ] = r_mag * cosf(r_phase);
505 
+    dstr[2 * n + 1] = r_mag * sinf(r_phase);
506 
+
507 
+    dstc[2 * n    ] = c_mag * cosf(c_phase);
508 
+    dstc[2 * n + 1] = c_mag * sinf(c_phase);
509 
+
510 
+    dstlb[2 * n    ] = lb_mag * cosf(l_phase);
511 
+    dstlb[2 * n + 1] = lb_mag * sinf(l_phase);
512 
+
513 
+    dstrb[2 * n    ] = rb_mag * cosf(r_phase);
514 
+    dstrb[2 * n + 1] = rb_mag * sinf(r_phase);
515 
+
516 
+    dstls[2 * n    ] = ls_mag * cosf(l_phase);
517 
+    dstls[2 * n + 1] = ls_mag * sinf(l_phase);
518 
+
519 
+    dstrs[2 * n    ] = rs_mag * cosf(r_phase);
520 
+    dstrs[2 * n + 1] = rs_mag * sinf(r_phase);
521 
+}
522 
+
523 
+static void upmix_7_1(AVFilterContext *ctx,
524 
+                      float l_phase,
525 
+                      float r_phase,
526 
+                      float c_phase,
527 
+                      float mag_total,
528 
+                      float x, float y,
529 
+                      int n)
530 
+{
531 
+    float lfe_mag, l_mag, r_mag, ls_mag, rs_mag, c_mag, lb_mag, rb_mag;
532 
+    float *dstc, *dstl, *dstr, *dstls, *dstrs, *dstlb, *dstrb, *dstlfe;
533 
+    AudioSurroundContext *s = ctx->priv;
534 
+
535 
+    dstl  = (float *)s->output->extended_data[0];
536 
+    dstr  = (float *)s->output->extended_data[1];
537 
+    dstc  = (float *)s->output->extended_data[2];
538 
+    dstlfe = (float *)s->output->extended_data[3];
539 
+    dstlb = (float *)s->output->extended_data[4];
540 
+    dstrb = (float *)s->output->extended_data[5];
541 
+    dstls = (float *)s->output->extended_data[6];
542 
+    dstrs = (float *)s->output->extended_data[7];
543 
+
544 
+    get_lfe(s->output_lfe, n, s->lowcut, s->highcut, &lfe_mag, &mag_total);
545 
+
546 
+    c_mag  = sqrtf(1.f - fabsf(x))   * ((y + 1.f) * .5f) * mag_total;
547 
+    l_mag  = sqrtf(.5f * ( x + 1.f)) * ((y + 1.f) * .5f) * mag_total;
548 
+    r_mag  = sqrtf(.5f * (-x + 1.f)) * ((y + 1.f) * .5f) * mag_total;
549 
+    lb_mag = sqrtf(.5f * ( x + 1.f)) * (1.f - ((y + 1.f) * .5f)) * mag_total;
550 
+    rb_mag = sqrtf(.5f * (-x + 1.f)) * (1.f - ((y + 1.f) * .5f)) * mag_total;
551 
+    ls_mag = sqrtf(.5f * ( x + 1.f)) * (1.f - fabsf(y)) * mag_total;
552 
+    rs_mag = sqrtf(.5f * (-x + 1.f)) * (1.f - fabsf(y)) * mag_total;
553 
+
554 
+    dstl[2 * n    ] = l_mag * cosf(l_phase);
555 
+    dstl[2 * n + 1] = l_mag * sinf(l_phase);
556 
+
557 
+    dstr[2 * n    ] = r_mag * cosf(r_phase);
558 
+    dstr[2 * n + 1] = r_mag * sinf(r_phase);
559 
+
560 
+    dstc[2 * n    ] = c_mag * cosf(c_phase);
561 
+    dstc[2 * n + 1] = c_mag * sinf(c_phase);
562 
+
563 
+    dstlfe[2 * n    ] = lfe_mag * cosf(c_phase);
564 
+    dstlfe[2 * n + 1] = lfe_mag * sinf(c_phase);
565 
+
566 
+    dstlb[2 * n    ] = lb_mag * cosf(l_phase);
567 
+    dstlb[2 * n + 1] = lb_mag * sinf(l_phase);
568 
+
569 
+    dstrb[2 * n    ] = rb_mag * cosf(r_phase);
570 
+    dstrb[2 * n + 1] = rb_mag * sinf(r_phase);
571 
+
572 
+    dstls[2 * n    ] = ls_mag * cosf(l_phase);
573 
+    dstls[2 * n + 1] = ls_mag * sinf(l_phase);
574 
+
575 
+    dstrs[2 * n    ] = rs_mag * cosf(r_phase);
576 
+    dstrs[2 * n + 1] = rs_mag * sinf(r_phase);
577 
+}
578 
+
579 
+static int init(AVFilterContext *ctx)
580 
+{
581 
+    AudioSurroundContext *s = ctx->priv;
582 
+    float overlap;
583 
+    int i;
584 
+
585 
+    if (!(s->out_channel_layout = av_get_channel_layout(s->out_channel_layout_str))) {
586 
+        av_log(ctx, AV_LOG_ERROR, "Error parsing channel layout '%s'.\n",
587 
+               s->out_channel_layout_str);
588 
+        return AVERROR(EINVAL);
589 
+    }
590 
+
591 
+    if (s->lowcutf >= s->highcutf) {
592 
+        av_log(ctx, AV_LOG_ERROR, "Low cut-off '%d' should be less than high cut-off '%d'.\n",
593 
+               s->lowcutf, s->highcutf);
594 
+        return AVERROR(EINVAL);
595 
+    }
596 
+
597 
+    switch (s->out_channel_layout) {
598 
+    case AV_CH_LAYOUT_MONO:
599 
+        s->upmix = upmix_1_0;
600 
+        break;
601 
+    case AV_CH_LAYOUT_STEREO:
602 
+        s->upmix = upmix_stereo;
603 
+        break;
604 
+    case AV_CH_LAYOUT_2POINT1:
605 
+        s->upmix = upmix_2_1;
606 
+        break;
607 
+    case AV_CH_LAYOUT_SURROUND:
608 
+        s->upmix = upmix_3_0;
609 
+        break;
610 
+    case AV_CH_LAYOUT_3POINT1:
611 
+        s->upmix = upmix_3_1;
612 
+        break;
613 
+    case AV_CH_LAYOUT_4POINT0:
614 
+        s->upmix = upmix_4_0;
615 
+        break;
616 
+    case AV_CH_LAYOUT_4POINT1:
617 
+        s->upmix = upmix_4_1;
618 
+        break;
619 
+    case AV_CH_LAYOUT_5POINT0_BACK:
620 
+        s->upmix = upmix_5_0_back;
621 
+        break;
622 
+    case AV_CH_LAYOUT_5POINT1_BACK:
623 
+        s->upmix = upmix_5_1_back;
624 
+        break;
625 
+    case AV_CH_LAYOUT_7POINT0:
626 
+        s->upmix = upmix_7_0;
627 
+        break;
628 
+    case AV_CH_LAYOUT_7POINT1:
629 
+        s->upmix = upmix_7_1;
630 
+        break;
631 
+    default:
632 
+        av_log(ctx, AV_LOG_ERROR, "Unsupported output channel layout '%s'.\n",
633 
+               s->out_channel_layout_str);
634 
+        return AVERROR(EINVAL);
635 
+    }
636 
+
637 
+    s->buf_size = 4096;
638 
+    s->pts = AV_NOPTS_VALUE;
639 
+
640 
+    s->window_func_lut = av_calloc(s->buf_size, sizeof(*s->window_func_lut));
641 
+    if (!s->window_func_lut)
642 
+        return AVERROR(ENOMEM);
643 
+
644 
+    for (i = 0; i < s->buf_size; i++)
645 
+        s->window_func_lut[i] = sqrtf(0.5 * (1 - cosf(2 * M_PI * i / s->buf_size)) / s->buf_size);
646 
+    overlap = .5;
647 
+    s->hop_size = s->buf_size * (1. - overlap);
648 
+
649 
+    return 0;
650 
+}
651 
+
652 
+static int fft_channel(AVFilterContext *ctx, void *arg, int ch, int nb_jobs)
653 
+{
654 
+    AudioSurroundContext *s = ctx->priv;
655 
+    const float level_in = s->level_in;
656 
+    float *dst;
657 
+    int n;
658 
+
659 
+    memset(s->input->extended_data[ch] + s->buf_size * sizeof(float), 0, s->buf_size * sizeof(float));
660 
+
661 
+    dst = (float *)s->input->extended_data[ch];
662 
+    for (n = 0; n < s->buf_size; n++) {
663 
+        dst[n] *= s->window_func_lut[n] * level_in;
664 
+    }
665 
+
666 
+    av_rdft_calc(s->rdft[ch], (float *)s->input->extended_data[ch]);
667 
+
668 
+    return 0;
669 
+}
670 
+
671 
+static int ifft_channel(AVFilterContext *ctx, void *arg, int ch, int nb_jobs)
672 
+{
673 
+    AudioSurroundContext *s = ctx->priv;
674 
+    const float level_out = s->level_out;
675 
+    AVFrame *out = arg;
676 
+    float *dst, *ptr;
677 
+    int n;
678 
+
679 
+    av_rdft_calc(s->irdft[ch], (float *)s->output->extended_data[ch]);
680 
+
681 
+    dst = (float *)s->output->extended_data[ch];
682 
+    ptr = (float *)s->overlap_buffer->extended_data[ch];
683 
+
684 
+    memmove(s->overlap_buffer->extended_data[ch],
685 
+            s->overlap_buffer->extended_data[ch] + s->hop_size * sizeof(float),
686 
+            s->buf_size * sizeof(float));
687 
+    memset(s->overlap_buffer->extended_data[ch] + s->buf_size * sizeof(float),
688 
+           0, s->hop_size * sizeof(float));
689 
+
690 
+    for (n = 0; n < s->buf_size; n++) {
691 
+        ptr[n] += dst[n] * s->window_func_lut[n] * level_out;
692 
+    }
693 
+
694 
+    ptr = (float *)s->overlap_buffer->extended_data[ch];
695 
+    dst = (float *)out->extended_data[ch];
696 
+    memcpy(dst, ptr, s->hop_size * sizeof(float));
697 
+
698 
+    return 0;
699 
+}
700 
+
701 
+static int filter_frame(AVFilterLink *inlink, AVFrame *in)
702 
+{
703 
+    AVFilterContext *ctx = inlink->dst;
704 
+    AVFilterLink *outlink = ctx->outputs[0];
705 
+    AudioSurroundContext *s = ctx->priv;
706 
+
707 
+    av_audio_fifo_write(s->fifo, (void **)in->extended_data,
708 
+                        in->nb_samples);
709 
+
710 
+    if (s->pts == AV_NOPTS_VALUE)
711 
+        s->pts = in->pts;
712 
+
713 
+    av_frame_free(&in);
714 
+
715 
+    while (av_audio_fifo_size(s->fifo) >= s->buf_size) {
716 
+        float *srcl, *srcr;
717 
+        AVFrame *out;
718 
+        int n, ret;
719 
+
720 
+        ret = av_audio_fifo_peek(s->fifo, (void **)s->input->extended_data, s->buf_size);
721 
+        if (ret < 0)
722 
+            return ret;
723 
+
724 
+        ctx->internal->execute(ctx, fft_channel, NULL, NULL, inlink->channels);
725 
+
726 
+        srcl = (float *)s->input->extended_data[0];
727 
+        srcr = (float *)s->input->extended_data[1];
728 
+
729 
+        for (n = 0; n < s->buf_size; n++) {
730 
+            float l_re = srcl[2 * n], r_re = srcr[2 * n];
731 
+            float l_im = srcl[2 * n + 1], r_im = srcr[2 * n + 1];
732 
+            float c_phase = atan2f(l_im + r_im, l_re + r_re);
733 
+            float l_mag = hypotf(l_re, l_im);
734 
+            float r_mag = hypotf(r_re, r_im);
735 
+            float l_phase = atan2f(l_im, l_re);
736 
+            float r_phase = atan2f(r_im, r_re);
737 
+            float phase_dif = fabsf(l_phase - r_phase);
738 
+            float mag_dif = (l_mag - r_mag) / (l_mag + r_mag);
739 
+            float mag_total = hypotf(l_mag, r_mag);
740 
+            float x, y;
741 
+
742 
+            if (phase_dif > M_PI)
743 
+                phase_dif = 2 * M_PI - phase_dif;
744 
+
745 
+            stereo_position(mag_dif, phase_dif, &x, &y);
746 
+
747 
+            s->upmix(ctx, l_phase, r_phase, c_phase, mag_total, x, y, n);
748 
+        }
749 
+
750 
+        out = ff_get_audio_buffer(outlink, s->hop_size);
751 
+        if (!out)
752 
+            return AVERROR(ENOMEM);
753 
+
754 
+        ctx->internal->execute(ctx, ifft_channel, out, NULL, outlink->channels);
755 
+
756 
+        out->pts = s->pts;
757 
+        if (s->pts != AV_NOPTS_VALUE)
758 
+            s->pts += av_rescale_q(out->nb_samples, (AVRational){1, outlink->sample_rate}, outlink->time_base);
759 
+        av_audio_fifo_drain(s->fifo, s->hop_size);
760 
+        ret = ff_filter_frame(outlink, out);
761 
+        if (ret < 0)
762 
+            return ret;
763 
+    }
764 
+
765 
+    return 0;
766 
+}
767 
+
768 
+static av_cold void uninit(AVFilterContext *ctx)
769 
+{
770 
+    AudioSurroundContext *s = ctx->priv;
771 
+    int ch;
772 
+
773 
+    av_frame_free(&s->input);
774 
+    av_frame_free(&s->output);
775 
+    av_frame_free(&s->overlap_buffer);
776 
+
777 
+    for (ch = 0; ch < s->nb_in_channels; ch++) {
778 
+        av_rdft_end(s->rdft[ch]);
779 
+    }
780 
+    for (ch = 0; ch < s->nb_out_channels; ch++) {
781 
+        av_rdft_end(s->irdft[ch]);
782 
+    }
783 
+    av_freep(&s->rdft);
784 
+    av_freep(&s->irdft);
785 
+    av_audio_fifo_free(s->fifo);
786 
+    av_freep(&s->window_func_lut);
787 
+}
788 
+
789 
+#define OFFSET(x) offsetof(AudioSurroundContext, x)
790 
+#define FLAGS AV_OPT_FLAG_AUDIO_PARAM|AV_OPT_FLAG_FILTERING_PARAM
791 
+
792 
+static const AVOption surround_options[] = {
793 
+    { "chl_out",   "set output channel layout", OFFSET(out_channel_layout_str), AV_OPT_TYPE_STRING, {.str="5.1"}, 0,   0, FLAGS },
794 
+    { "level_in",  "set input level",           OFFSET(level_in),               AV_OPT_TYPE_FLOAT,  {.dbl=1},     0,  10, FLAGS },
795 
+    { "level_out", "set output level",          OFFSET(level_out),              AV_OPT_TYPE_FLOAT,  {.dbl=1},     0,  10, FLAGS },
796 
+    { "lfe",       "output LFE",                OFFSET(output_lfe),             AV_OPT_TYPE_BOOL,   {.i64=1},     0,   1, FLAGS },
797 
+    { "lfe_low",   "LFE low cut off",           OFFSET(lowcutf),                AV_OPT_TYPE_INT,    {.i64=128},   0, 256, FLAGS },
798 
+    { "lfe_high",  "LFE high cut off",          OFFSET(highcutf),               AV_OPT_TYPE_INT,    {.i64=256},   0, 512, FLAGS },
799 
+    { NULL }
800 
+};
801 
+
802 
+AVFILTER_DEFINE_CLASS(surround);
803 
+
804 
+static const AVFilterPad inputs[] = {
805 
+    {
806 
+        .name         = "default",
807 
+        .type         = AVMEDIA_TYPE_AUDIO,
808 
+        .filter_frame = filter_frame,
809 
+        .config_props = config_input,
810 
+    },
811 
+    { NULL }
812 
+};
813 
+
814 
+static const AVFilterPad outputs[] = {
815 
+    {
816 
+        .name         = "default",
817 
+        .type         = AVMEDIA_TYPE_AUDIO,
818 
+        .config_props = config_output,
819 
+    },
820 
+    { NULL }
821 
+};
822 
+
823 
+AVFilter ff_af_surround = {
824 
+    .name           = "surround",
825 
+    .description    = NULL_IF_CONFIG_SMALL("Apply audio surround upmix filter."),
826 
+    .query_formats  = query_formats,
827 
+    .priv_size      = sizeof(AudioSurroundContext),
828 
+    .priv_class     = &surround_class,
829 
+    .init           = init,
830 
+    .uninit         = uninit,
831 
+    .inputs         = inputs,
832 
+    .outputs        = outputs,
833 
+    .flags          = AVFILTER_FLAG_SLICE_THREADS,
834 
+};
libavfilter/allfilters.c
History View file @ dc72d1d
... ... 
@@ -121,6 +121,7 @@ static void register_all(void)
121 121 
     REGISTER_FILTER(SOFALIZER,      sofalizer,      af);
122 122 
     REGISTER_FILTER(STEREOTOOLS,    stereotools,    af);
123 123 
     REGISTER_FILTER(STEREOWIDEN,    stereowiden,    af);
124 
+    REGISTER_FILTER(SURROUND,       surround,       af);
124 125 
     REGISTER_FILTER(TREBLE,         treble,         af);
125 126 
     REGISTER_FILTER(TREMOLO,        tremolo,        af);
126 127 
     REGISTER_FILTER(VIBRATO,        vibrato,        af);
libavfilter/version.h
History View file @ dc72d1d
... ... 
@@ -30,7 +30,7 @@
30 30 
 #include "libavutil/version.h"
31 31
32 32 
 #define LIBAVFILTER_VERSION_MAJOR   6
33 
-#define LIBAVFILTER_VERSION_MINOR  90
33 
+#define LIBAVFILTER_VERSION_MINOR  91
34 34 
 #define LIBAVFILTER_VERSION_MICRO 100
35 35
36 36 
 #define LIBAVFILTER_VERSION_INT AV_VERSION_INT(LIBAVFILTER_VERSION_MAJOR, \