FFmpeg
af_adynamicequalizer.c
Go to the documentation of this file.
1 /*
2  * This file is part of FFmpeg.
3  *
4  * FFmpeg is free software; you can redistribute it and/or
5  * modify it under the terms of the GNU Lesser General Public
6  * License as published by the Free Software Foundation; either
7  * version 2.1 of the License, or (at your option) any later version.
8  *
9  * FFmpeg is distributed in the hope that it will be useful,
10  * but WITHOUT ANY WARRANTY; without even the implied warranty of
11  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
12  * Lesser General Public License for more details.
13  *
14  * You should have received a copy of the GNU Lesser General Public
15  * License along with FFmpeg; if not, write to the Free Software
16  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
17  */
18 
19 #include <float.h>
20 
21 #include "libavutil/opt.h"
22 #include "avfilter.h"
23 #include "audio.h"
24 #include "formats.h"
25 
27  const AVClass *class;
28 
29  double threshold;
30  double dfrequency;
31  double dqfactor;
32  double tfrequency;
33  double tqfactor;
34  double ratio;
35  double range;
36  double makeup;
37  double knee;
38  double slew;
39  double attack;
40  double release;
41  double attack_coef;
42  double release_coef;
43  int mode;
44  int type;
45 
48 
50 {
51  AVFilterContext *ctx = inlink->dst;
53 
54  s->state = ff_get_audio_buffer(inlink, 8);
55  if (!s->state)
56  return AVERROR(ENOMEM);
57 
58  return 0;
59 }
60 
61 static double get_svf(double in, double *m, double *a, double *b)
62 {
63  const double v0 = in;
64  const double v3 = v0 - b[1];
65  const double v1 = a[0] * b[0] + a[1] * v3;
66  const double v2 = b[1] + a[1] * b[0] + a[2] * v3;
67 
68  b[0] = 2. * v1 - b[0];
69  b[1] = 2. * v2 - b[1];
70 
71  return m[0] * v0 + m[1] * v1 + m[2] * v2;
72 }
73 
74 static inline double from_dB(double x)
75 {
76  return exp(0.05 * x * M_LN10);
77 }
78 
79 static inline double to_dB(double x)
80 {
81  return 20. * log10(x);
82 }
83 
84 static inline double sqr(double x)
85 {
86  return x * x;
87 }
88 
89 static double get_gain(double in, double srate, double makeup,
90  double aattack, double iratio, double knee, double range,
91  double thresdb, double slewfactor, double *state,
92  double attack_coeff, double release_coeff, double nc)
93 {
94  double width = (6. * knee) + 0.01;
95  double cdb = 0.;
96  double Lgain = 1.;
97  double Lxg, Lxl, Lyg, Lyl, Ly1;
98  double checkwidth = 0.;
99  double slewwidth = 1.8;
100  int attslew = 0;
101 
102  Lyg = 0.;
103  Lxg = to_dB(fabs(in) + DBL_EPSILON);
104 
105  Lyg = Lxg + (iratio - 1.) * sqr(Lxg - thresdb + width * .5) / (2. * width);
106 
107  checkwidth = 2. * fabs(Lxg - thresdb);
108  if (2. * (Lxg - thresdb) < -width) {
109  Lyg = Lxg;
110  } else if (checkwidth <= width) {
111  Lyg = thresdb + (Lxg - thresdb) * iratio;
112  if (checkwidth <= slewwidth) {
113  if (Lyg >= state[2])
114  attslew = 1;
115  }
116  } else if (2. * (Lxg - thresdb) > width) {
117  Lyg = thresdb + (Lxg - thresdb) * iratio;
118  }
119 
120  attack_coeff = attslew ? aattack : attack_coeff;
121 
122  Lxl = Lxg - Lyg;
123 
124  Ly1 = fmax(Lxl, release_coeff * state[1] +(1. - release_coeff) * Lxl);
125  Lyl = attack_coeff * state[0] + (1. - attack_coeff) * Ly1;
126 
127  cdb = -Lyl;
128  Lgain = from_dB(nc * fmin(cdb - makeup, range));
129 
130  state[0] = Lyl;
131  state[1] = Ly1;
132  state[2] = Lyg;
133 
134  return Lgain;
135 }
136 
137 typedef struct ThreadData {
138  AVFrame *in, *out;
139 } ThreadData;
140 
141 static int filter_channels(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)
142 {
144  ThreadData *td = arg;
145  AVFrame *in = td->in;
146  AVFrame *out = td->out;
147  const double sample_rate = in->sample_rate;
148  const double makeup = s->makeup;
149  const double iratio = 1. / s->ratio;
150  const double range = s->range;
151  const double dfrequency = fmin(s->dfrequency, sample_rate * 0.5);
152  const double tfrequency = fmin(s->tfrequency, sample_rate * 0.5);
153  const double threshold = to_dB(s->threshold + DBL_EPSILON);
154  const double release = s->release_coef;
155  const double attack = s->attack_coef;
156  const double dqfactor = s->dqfactor;
157  const double tqfactor = s->tqfactor;
158  const double fg = tan(M_PI * tfrequency / sample_rate);
159  const double dg = tan(M_PI * dfrequency / sample_rate);
160  const int start = (in->ch_layout.nb_channels * jobnr) / nb_jobs;
161  const int end = (in->ch_layout.nb_channels * (jobnr+1)) / nb_jobs;
162  const int mode = s->mode;
163  const int type = s->type;
164  const double knee = s->knee;
165  const double slew = s->slew;
166  const double aattack = exp(-1000. / ((s->attack + 2.0 * (slew - 1.)) * sample_rate));
167  const double nc = mode == 0 ? 1. : -1.;
168  double da[3], dm[3];
169 
170  {
171  double k = 1. / dqfactor;
172 
173  da[0] = 1. / (1. + dg * (dg + k));
174  da[1] = dg * da[0];
175  da[2] = dg * da[1];
176 
177  dm[0] = 0.;
178  dm[1] = 1.;
179  dm[2] = 0.;
180  }
181 
182  for (int ch = start; ch < end; ch++) {
183  const double *src = (const double *)in->extended_data[ch];
184  double *dst = (double *)out->extended_data[ch];
185  double *state = (double *)s->state->extended_data[ch];
186 
187  for (int n = 0; n < out->nb_samples; n++) {
188  double detect, gain, v, listen;
189  double fa[3], fm[3];
190  double k, g;
191 
192  detect = listen = get_svf(src[n], dm, da, state);
193  detect = fabs(detect);
194 
195  gain = get_gain(detect, sample_rate, makeup,
196  aattack, iratio, knee, range, threshold, slew,
197  &state[4], attack, release, nc);
198 
199  switch (type) {
200  case 0:
201  k = 1. / (tqfactor * gain);
202 
203  fa[0] = 1. / (1. + fg * (fg + k));
204  fa[1] = fg * fa[0];
205  fa[2] = fg * fa[1];
206 
207  fm[0] = 1.;
208  fm[1] = k * (gain * gain - 1.);
209  fm[2] = 0.;
210  break;
211  case 1:
212  k = 1. / tqfactor;
213  g = fg / sqrt(gain);
214 
215  fa[0] = 1. / (1. + g * (g + k));
216  fa[1] = g * fa[0];
217  fa[2] = g * fa[1];
218 
219  fm[0] = 1.;
220  fm[1] = k * (gain - 1.);
221  fm[2] = gain * gain - 1.;
222  break;
223  case 2:
224  k = 1. / tqfactor;
225  g = fg / sqrt(gain);
226 
227  fa[0] = 1. / (1. + g * (g + k));
228  fa[1] = g * fa[0];
229  fa[2] = g * fa[1];
230 
231  fm[0] = gain * gain;
232  fm[1] = k * (1. - gain) * gain;
233  fm[2] = 1. - gain * gain;
234  break;
235  }
236 
237  v = get_svf(src[n], fm, fa, &state[2]);
238  v = mode == -1 ? listen : v;
239  dst[n] = ctx->is_disabled ? src[n] : v;
240  }
241  }
242 
243  return 0;
244 }
245 
246 static double get_coef(double x, double sr)
247 {
248  return exp(-1000. / (x * sr));
249 }
250 
252 {
253  AVFilterContext *ctx = inlink->dst;
254  AVFilterLink *outlink = ctx->outputs[0];
256  ThreadData td;
257  AVFrame *out;
258 
259  if (av_frame_is_writable(in)) {
260  out = in;
261  } else {
262  out = ff_get_audio_buffer(outlink, in->nb_samples);
263  if (!out) {
264  av_frame_free(&in);
265  return AVERROR(ENOMEM);
266  }
268  }
269 
270  s->attack_coef = get_coef(s->attack, in->sample_rate);
271  s->release_coef = get_coef(s->release, in->sample_rate);
272 
273  td.in = in;
274  td.out = out;
277 
278  if (out != in)
279  av_frame_free(&in);
280  return ff_filter_frame(outlink, out);
281 }
282 
284 {
286 
287  av_frame_free(&s->state);
288 }
289 
290 #define OFFSET(x) offsetof(AudioDynamicEqualizerContext, x)
291 #define FLAGS AV_OPT_FLAG_AUDIO_PARAM|AV_OPT_FLAG_FILTERING_PARAM|AV_OPT_FLAG_RUNTIME_PARAM
292 
294  { "threshold", "set detection threshold", OFFSET(threshold), AV_OPT_TYPE_DOUBLE, {.dbl=0}, 0, 100, FLAGS },
295  { "dfrequency", "set detection frequency", OFFSET(dfrequency), AV_OPT_TYPE_DOUBLE, {.dbl=1000}, 2, 1000000, FLAGS },
296  { "dqfactor", "set detection Q factor", OFFSET(dqfactor), AV_OPT_TYPE_DOUBLE, {.dbl=1}, 0.001, 1000, FLAGS },
297  { "tfrequency", "set target frequency", OFFSET(tfrequency), AV_OPT_TYPE_DOUBLE, {.dbl=1000}, 2, 1000000, FLAGS },
298  { "tqfactor", "set target Q factor", OFFSET(tqfactor), AV_OPT_TYPE_DOUBLE, {.dbl=1}, 0.001, 1000, FLAGS },
299  { "attack", "set attack duration", OFFSET(attack), AV_OPT_TYPE_DOUBLE, {.dbl=20}, 1, 2000, FLAGS },
300  { "release", "set release duration", OFFSET(release), AV_OPT_TYPE_DOUBLE, {.dbl=200}, 1, 2000, FLAGS },
301  { "knee", "set knee factor", OFFSET(knee), AV_OPT_TYPE_DOUBLE, {.dbl=1}, 0, 8, FLAGS },
302  { "ratio", "set ratio factor", OFFSET(ratio), AV_OPT_TYPE_DOUBLE, {.dbl=1}, 1, 20, FLAGS },
303  { "makeup", "set makeup gain", OFFSET(makeup), AV_OPT_TYPE_DOUBLE, {.dbl=0}, 0, 30, FLAGS },
304  { "range", "set max gain", OFFSET(range), AV_OPT_TYPE_DOUBLE, {.dbl=0}, 0, 200, FLAGS },
305  { "slew", "set slew factor", OFFSET(slew), AV_OPT_TYPE_DOUBLE, {.dbl=1}, 1, 200, FLAGS },
306  { "mode", "set mode", OFFSET(mode), AV_OPT_TYPE_INT, {.i64=0}, -1, 1, FLAGS, "mode" },
307  { "listen", 0, 0, AV_OPT_TYPE_CONST, {.i64=-1}, 0, 0, FLAGS, "mode" },
308  { "cut", 0, 0, AV_OPT_TYPE_CONST, {.i64=0}, 0, 0, FLAGS, "mode" },
309  { "boost", 0, 0, AV_OPT_TYPE_CONST, {.i64=1}, 0, 0, FLAGS, "mode" },
310  { "tftype", "set target filter type", OFFSET(type), AV_OPT_TYPE_INT, {.i64=0}, 0, 2, FLAGS, "type" },
311  { "bell", 0, 0, AV_OPT_TYPE_CONST, {.i64=0}, 0, 0, FLAGS, "type" },
312  { "lowshelf", 0, 0, AV_OPT_TYPE_CONST, {.i64=1}, 0, 0, FLAGS, "type" },
313  { "highshelf",0, 0, AV_OPT_TYPE_CONST, {.i64=2}, 0, 0, FLAGS, "type" },
314  { NULL }
315 };
316 
317 AVFILTER_DEFINE_CLASS(adynamicequalizer);
318 
319 static const AVFilterPad inputs[] = {
320  {
321  .name = "default",
322  .type = AVMEDIA_TYPE_AUDIO,
323  .filter_frame = filter_frame,
324  .config_props = config_input,
325  },
326 };
327 
328 static const AVFilterPad outputs[] = {
329  {
330  .name = "default",
331  .type = AVMEDIA_TYPE_AUDIO,
332  },
333 };
334 
336  .name = "adynamicequalizer",
337  .description = NULL_IF_CONFIG_SMALL("Apply Dynamic Equalization of input audio."),
338  .priv_size = sizeof(AudioDynamicEqualizerContext),
339  .priv_class = &adynamicequalizer_class,
340  .uninit = uninit,
346  .process_command = ff_filter_process_command,
347 };
ff_get_audio_buffer
AVFrame * ff_get_audio_buffer(AVFilterLink *link, int nb_samples)
Request an audio samples buffer with a specific set of permissions.
Definition: audio.c:100
td
#define td
Definition: regdef.h:70
AudioDynamicEqualizerContext::dfrequency
double dfrequency
Definition: af_adynamicequalizer.c:30
AVERROR
Filter the word “frame” indicates either a video frame or a group of audio as stored in an AVFrame structure Format for each input and each output the list of supported formats For video that means pixel format For audio that means channel sample they are references to shared objects When the negotiation mechanism computes the intersection of the formats supported at each end of a all references to both lists are replaced with a reference to the intersection And when a single format is eventually chosen for a link amongst the remaining all references to the list are updated That means that if a filter requires that its input and output have the same format amongst a supported all it has to do is use a reference to the same list of formats query_formats can leave some formats unset and return AVERROR(EAGAIN) to cause the negotiation mechanism toagain later. That can be used by filters with complex requirements to use the format negotiated on one link to set the formats supported on another. Frame references ownership and permissions
opt.h
AudioDynamicEqualizerContext::range
double range
Definition: af_adynamicequalizer.c:35
out
FILE * out
Definition: movenc.c:54
ff_filter_frame
int ff_filter_frame(AVFilterLink *link, AVFrame *frame)
Send a frame of data to the next filter.
Definition: avfilter.c:999
FILTER_SINGLE_SAMPLEFMT
#define FILTER_SINGLE_SAMPLEFMT(sample_fmt_)
Definition: internal.h:183
inlink
The exact code depends on how similar the blocks are and how related they are to the and needs to apply these operations to the correct inlink or outlink if there are several Macros are available to factor that when no extra processing is inlink
Definition: filter_design.txt:212
ff_af_adynamicequalizer
const AVFilter ff_af_adynamicequalizer
Definition: af_adynamicequalizer.c:335
av_frame_free
void av_frame_free(AVFrame **frame)
Free the frame and any dynamically allocated objects in it, e.g.
Definition: frame.c:111
AudioDynamicEqualizerContext::makeup
double makeup
Definition: af_adynamicequalizer.c:36
AVFrame
This structure describes decoded (raw) audio or video data.
Definition: frame.h:325
AudioDynamicEqualizerContext::type
int type
Definition: af_adynamicequalizer.c:44
config_input
static int config_input(AVFilterLink *inlink)
Definition: af_adynamicequalizer.c:49
AVOption
AVOption.
Definition: opt.h:251
b
#define b
Definition: input.c:34
FLAGS
#define FLAGS
Definition: af_adynamicequalizer.c:291
float.h
filter_channels
static int filter_channels(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)
Definition: af_adynamicequalizer.c:141
AVFilter::name
const char * name
Filter name.
Definition: avfilter.h:175
AudioDynamicEqualizerContext::mode
int mode
Definition: af_adynamicequalizer.c:43
AudioDynamicEqualizerContext::state
AVFrame * state
Definition: af_adynamicequalizer.c:46
ThreadData::out
AVFrame * out
Definition: af_adeclick.c:473
AVChannelLayout::nb_channels
int nb_channels
Number of channels in this layout.
Definition: channel_layout.h:300
ThreadData::in
AVFrame * in
Definition: af_adecorrelate.c:154
get_coef
static double get_coef(double x, double sr)
Definition: af_adynamicequalizer.c:246
sample_rate
sample_rate
Definition: ffmpeg_filter.c:153
formats.h
v0
#define v0
Definition: regdef.h:26
filter_frame
static int filter_frame(AVFilterLink *inlink, AVFrame *in)
Definition: af_adynamicequalizer.c:251
AVFrame::ch_layout
AVChannelLayout ch_layout
Channel layout of the audio data.
Definition: frame.h:704
type
it s the only field you need to keep assuming you have a context There is some magic you don t need to care about around this just let it vf type
Definition: writing_filters.txt:86
uninit
static av_cold void uninit(AVFilterContext *ctx)
Definition: af_adynamicequalizer.c:283
AVFilterPad
A filter pad used for either input or output.
Definition: internal.h:49
get_svf
static double get_svf(double in, double *m, double *a, double *b)
Definition: af_adynamicequalizer.c:61
AudioDynamicEqualizerContext::release_coef
double release_coef
Definition: af_adynamicequalizer.c:42
av_cold
#define av_cold
Definition: attributes.h:90
adynamicequalizer_options
static const AVOption adynamicequalizer_options[]
Definition: af_adynamicequalizer.c:293
width
#define width
s
#define s(width, name)
Definition: cbs_vp9.c:256
g
const char * g
Definition: vf_curves.c:117
AV_OPT_TYPE_DOUBLE
@ AV_OPT_TYPE_DOUBLE
Definition: opt.h:227
AVMEDIA_TYPE_AUDIO
@ AVMEDIA_TYPE_AUDIO
Definition: avutil.h:202
ctx
AVFormatContext * ctx
Definition: movenc.c:48
FILTER_INPUTS
#define FILTER_INPUTS(array)
Definition: internal.h:190
arg
const char * arg
Definition: jacosubdec.c:67
to_dB
static double to_dB(double x)
Definition: af_adynamicequalizer.c:79
sqr
static double sqr(double x)
Definition: af_adynamicequalizer.c:84
AudioDynamicEqualizerContext::tqfactor
double tqfactor
Definition: af_adynamicequalizer.c:33
AVClass
Describe the class of an AVClass context structure.
Definition: log.h:66
fabs
static __device__ float fabs(float a)
Definition: cuda_runtime.h:182
NULL
#define NULL
Definition: coverity.c:32
av_frame_copy_props
int av_frame_copy_props(AVFrame *dst, const AVFrame *src)
Copy only "metadata" fields from src to dst.
Definition: frame.c:596
AVFILTER_DEFINE_CLASS
AVFILTER_DEFINE_CLASS(adynamicequalizer)
exp
int8_t exp
Definition: eval.c:72
AudioDynamicEqualizerContext::slew
double slew
Definition: af_adynamicequalizer.c:38
outputs
static const AVFilterPad outputs[]
Definition: af_adynamicequalizer.c:328
NULL_IF_CONFIG_SMALL
#define NULL_IF_CONFIG_SMALL(x)
Return NULL if CONFIG_SMALL is true, otherwise the argument without modification.
Definition: internal.h:117
OFFSET
#define OFFSET(x)
Definition: af_adynamicequalizer.c:290
AVFrame::sample_rate
int sample_rate
Sample rate of the audio data.
Definition: frame.h:502
for
for(k=2;k<=8;++k)
Definition: h264pred_template.c:425
fmin
double fmin(double, double)
av_frame_is_writable
int av_frame_is_writable(AVFrame *frame)
Check if the frame data is writable.
Definition: frame.c:523
AudioDynamicEqualizerContext
Definition: af_adynamicequalizer.c:26
ff_filter_process_command
int ff_filter_process_command(AVFilterContext *ctx, const char *cmd, const char *arg, char *res, int res_len, int flags)
Generic processing of user supplied commands that are set in the same way as the filter options.
Definition: avfilter.c:863
AudioDynamicEqualizerContext::knee
double knee
Definition: af_adynamicequalizer.c:37
a
The reader does not expect b to be semantically here and if the code is changed by maybe adding a a division or other the signedness will almost certainly be mistaken To avoid this confusion a new type was SUINT is the C unsigned type but it holds a signed int to use the same example SUINT a
Definition: undefined.txt:41
M_PI
#define M_PI
Definition: mathematics.h:52
AVFrame::nb_samples
int nb_samples
number of audio samples (per channel) described by this frame
Definition: frame.h:405
AVFrame::extended_data
uint8_t ** extended_data
pointers to the data planes/channels.
Definition: frame.h:386
ff_filter_get_nb_threads
int ff_filter_get_nb_threads(AVFilterContext *ctx)
Get number of threads for current filter instance.
Definition: avfilter.c:783
ThreadData
Used for passing data between threads.
Definition: dsddec.c:68
FFMIN
#define FFMIN(a, b)
Definition: macros.h:49
AVFilterPad::name
const char * name
Pad name.
Definition: internal.h:55
AudioDynamicEqualizerContext::ratio
double ratio
Definition: af_adynamicequalizer.c:34
from_dB
static double from_dB(double x)
Definition: af_adynamicequalizer.c:74
AVFilter
Filter definition.
Definition: avfilter.h:171
AudioDynamicEqualizerContext::attack
double attack
Definition: af_adynamicequalizer.c:39
fmax
double fmax(double, double)
AudioDynamicEqualizerContext::tfrequency
double tfrequency
Definition: af_adynamicequalizer.c:32
mode
mode
Definition: ebur128.h:83
AV_OPT_TYPE_INT
@ AV_OPT_TYPE_INT
Definition: opt.h:225
avfilter.h
AV_SAMPLE_FMT_DBLP
@ AV_SAMPLE_FMT_DBLP
double, planar
Definition: samplefmt.h:67
AudioDynamicEqualizerContext::dqfactor
double dqfactor
Definition: af_adynamicequalizer.c:31
AudioDynamicEqualizerContext::release
double release
Definition: af_adynamicequalizer.c:40
state
static struct @326 state
AVFilterContext
An instance of a filter.
Definition: avfilter.h:408
AVFILTER_FLAG_SLICE_THREADS
#define AVFILTER_FLAG_SLICE_THREADS
The filter supports multithreading by splitting frames into multiple parts and processing them concur...
Definition: avfilter.h:127
audio.h
M_LN10
#define M_LN10
Definition: mathematics.h:43
AudioDynamicEqualizerContext::threshold
double threshold
Definition: af_adynamicequalizer.c:29
get_gain
static double get_gain(double in, double srate, double makeup, double aattack, double iratio, double knee, double range, double thresdb, double slewfactor, double *state, double attack_coeff, double release_coeff, double nc)
Definition: af_adynamicequalizer.c:89
AudioDynamicEqualizerContext::attack_coef
double attack_coef
Definition: af_adynamicequalizer.c:41
FILTER_OUTPUTS
#define FILTER_OUTPUTS(array)
Definition: internal.h:191
src
INIT_CLIP pixel * src
Definition: h264pred_template.c:418
AVFILTER_FLAG_SUPPORT_TIMELINE_INTERNAL
#define AVFILTER_FLAG_SUPPORT_TIMELINE_INTERNAL
Same as AVFILTER_FLAG_SUPPORT_TIMELINE_GENERIC, except that the filter will have its filter_frame() c...
Definition: avfilter.h:160
inputs
static const AVFilterPad inputs[]
Definition: af_adynamicequalizer.c:319
ff_filter_execute
static av_always_inline int ff_filter_execute(AVFilterContext *ctx, avfilter_action_func *func, void *arg, int *ret, int nb_jobs)
Definition: internal.h:142
AV_OPT_TYPE_CONST
@ AV_OPT_TYPE_CONST
Definition: opt.h:234