FFmpeg/libavfilter/af_mcompand.c
Andreas Rheinhardt b4f5201967 avfilter: Replace query_formats callback with union of list and callback
If one looks at the many query_formats callbacks in existence,
one will immediately recognize that there is one type of default
callback for video and a slightly different default callback for
audio: It is "return ff_set_common_formats_from_list(ctx, pix_fmts);"
for video with a filter-specific pix_fmts list. For audio, it is
the same with a filter-specific sample_fmts list together with
ff_set_common_all_samplerates() and ff_set_common_all_channel_counts().

This commit allows to remove the boilerplate query_formats callbacks
by replacing said callback with a union consisting the old callback
and pointers for pixel and sample format arrays. For the not uncommon
case in which these lists only contain a single entry (besides the
sentinel) enum AVPixelFormat and enum AVSampleFormat fields are also
added to the union to store them directly in the AVFilter,
thereby avoiding a relocation.

The state of said union will be contained in a new, dedicated AVFilter
field (the nb_inputs and nb_outputs fields have been shrunk to uint8_t
in order to create a hole for this new field; this is no problem, as
the maximum of all the nb_inputs is four; for nb_outputs it is only
two).

The state's default value coincides with the earlier default of
query_formats being unset, namely that the filter accepts all formats
(and also sample rates and channel counts/layouts for audio)
provided that these properties agree coincide for all inputs and
outputs.

By using different union members for audio and video filters
the type-unsafety of using the same functions for audio and video
lists will furthermore be more confined to formats.c than before.

When the new fields are used, they will also avoid allocations:
Currently something nearly equivalent to ff_default_query_formats()
is called after every successful call to a query_formats callback;
yet in the common case that the newly allocated AVFilterFormats
are not used at all (namely if there are no free links) these newly
allocated AVFilterFormats are freed again without ever being used.
Filters no longer using the callback will not exhibit this any more.

Reviewed-by: Paul B Mahol <onemda@gmail.com>
Reviewed-by: Nicolas George <george@nsup.org>
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2021-10-05 17:48:25 +02:00

664 lines
20 KiB
C

/*
* COpyright (c) 2002 Daniel Pouzzner
* Copyright (c) 1999 Chris Bagwell
* Copyright (c) 1999 Nick Bailey
* Copyright (c) 2007 Rob Sykes <robs@users.sourceforge.net>
* Copyright (c) 2013 Paul B Mahol
* Copyright (c) 2014 Andrew Kelley
*
* This file is part of FFmpeg.
*
* FFmpeg is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
* FFmpeg is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with FFmpeg; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
/**
* @file
* audio multiband compand filter
*/
#include "libavutil/avstring.h"
#include "libavutil/ffmath.h"
#include "libavutil/opt.h"
#include "libavutil/samplefmt.h"
#include "audio.h"
#include "avfilter.h"
#include "internal.h"
typedef struct CompandSegment {
double x, y;
double a, b;
} CompandSegment;
typedef struct CompandT {
CompandSegment *segments;
int nb_segments;
double in_min_lin;
double out_min_lin;
double curve_dB;
double gain_dB;
} CompandT;
#define N 4
typedef struct PrevCrossover {
double in;
double out_low;
double out_high;
} PrevCrossover[N * 2];
typedef struct Crossover {
PrevCrossover *previous;
size_t pos;
double coefs[3 *(N+1)];
} Crossover;
typedef struct CompBand {
CompandT transfer_fn;
double *attack_rate;
double *decay_rate;
double *volume;
double delay;
double topfreq;
Crossover filter;
AVFrame *delay_buf;
size_t delay_size;
ptrdiff_t delay_buf_ptr;
size_t delay_buf_cnt;
} CompBand;
typedef struct MCompandContext {
const AVClass *class;
char *args;
int nb_bands;
CompBand *bands;
AVFrame *band_buf1, *band_buf2, *band_buf3;
int band_samples;
size_t delay_buf_size;
} MCompandContext;
#define OFFSET(x) offsetof(MCompandContext, x)
#define A AV_OPT_FLAG_AUDIO_PARAM|AV_OPT_FLAG_FILTERING_PARAM
static const AVOption mcompand_options[] = {
{ "args", "set parameters for each band", OFFSET(args), AV_OPT_TYPE_STRING, { .str = "0.005,0.1 6 -47/-40,-34/-34,-17/-33 100 | 0.003,0.05 6 -47/-40,-34/-34,-17/-33 400 | 0.000625,0.0125 6 -47/-40,-34/-34,-15/-33 1600 | 0.0001,0.025 6 -47/-40,-34/-34,-31/-31,-0/-30 6400 | 0,0.025 6 -38/-31,-28/-28,-0/-25 22000" }, 0, 0, A },
{ NULL }
};
AVFILTER_DEFINE_CLASS(mcompand);
static av_cold void uninit(AVFilterContext *ctx)
{
MCompandContext *s = ctx->priv;
int i;
av_frame_free(&s->band_buf1);
av_frame_free(&s->band_buf2);
av_frame_free(&s->band_buf3);
if (s->bands) {
for (i = 0; i < s->nb_bands; i++) {
av_freep(&s->bands[i].attack_rate);
av_freep(&s->bands[i].decay_rate);
av_freep(&s->bands[i].volume);
av_freep(&s->bands[i].transfer_fn.segments);
av_freep(&s->bands[i].filter.previous);
av_frame_free(&s->bands[i].delay_buf);
}
}
av_freep(&s->bands);
}
static int query_formats(AVFilterContext *ctx)
{
static const enum AVSampleFormat sample_fmts[] = {
AV_SAMPLE_FMT_DBLP,
AV_SAMPLE_FMT_NONE
};
int ret = ff_set_common_all_channel_counts(ctx);
if (ret < 0)
return ret;
ret = ff_set_common_formats_from_list(ctx, sample_fmts);
if (ret < 0)
return ret;
return ff_set_common_all_samplerates(ctx);
}
static void count_items(char *item_str, int *nb_items, char delimiter)
{
char *p;
*nb_items = 1;
for (p = item_str; *p; p++) {
if (*p == delimiter)
(*nb_items)++;
}
}
static void update_volume(CompBand *cb, double in, int ch)
{
double delta = in - cb->volume[ch];
if (delta > 0.0)
cb->volume[ch] += delta * cb->attack_rate[ch];
else
cb->volume[ch] += delta * cb->decay_rate[ch];
}
static double get_volume(CompandT *s, double in_lin)
{
CompandSegment *cs;
double in_log, out_log;
int i;
if (in_lin <= s->in_min_lin)
return s->out_min_lin;
in_log = log(in_lin);
for (i = 1; i < s->nb_segments; i++)
if (in_log <= s->segments[i].x)
break;
cs = &s->segments[i - 1];
in_log -= cs->x;
out_log = cs->y + in_log * (cs->a * in_log + cs->b);
return exp(out_log);
}
static int parse_points(char *points, int nb_points, double radius,
CompandT *s, AVFilterContext *ctx)
{
int new_nb_items, num;
char *saveptr = NULL;
char *p = points;
int i;
#define S(x) s->segments[2 * ((x) + 1)]
for (i = 0, new_nb_items = 0; i < nb_points; i++) {
char *tstr = av_strtok(p, ",", &saveptr);
p = NULL;
if (!tstr || sscanf(tstr, "%lf/%lf", &S(i).x, &S(i).y) != 2) {
av_log(ctx, AV_LOG_ERROR,
"Invalid and/or missing input/output value.\n");
return AVERROR(EINVAL);
}
if (i && S(i - 1).x > S(i).x) {
av_log(ctx, AV_LOG_ERROR,
"Transfer function input values must be increasing.\n");
return AVERROR(EINVAL);
}
S(i).y -= S(i).x;
av_log(ctx, AV_LOG_DEBUG, "%d: x=%f y=%f\n", i, S(i).x, S(i).y);
new_nb_items++;
}
num = new_nb_items;
/* Add 0,0 if necessary */
if (num == 0 || S(num - 1).x)
num++;
#undef S
#define S(x) s->segments[2 * (x)]
/* Add a tail off segment at the start */
S(0).x = S(1).x - 2 * s->curve_dB;
S(0).y = S(1).y;
num++;
/* Join adjacent colinear segments */
for (i = 2; i < num; i++) {
double g1 = (S(i - 1).y - S(i - 2).y) * (S(i - 0).x - S(i - 1).x);
double g2 = (S(i - 0).y - S(i - 1).y) * (S(i - 1).x - S(i - 2).x);
int j;
if (fabs(g1 - g2))
continue;
num--;
for (j = --i; j < num; j++)
S(j) = S(j + 1);
}
for (i = 0; i < s->nb_segments; i += 2) {
s->segments[i].y += s->gain_dB;
s->segments[i].x *= M_LN10 / 20;
s->segments[i].y *= M_LN10 / 20;
}
#define L(x) s->segments[i - (x)]
for (i = 4; i < s->nb_segments; i += 2) {
double x, y, cx, cy, in1, in2, out1, out2, theta, len, r;
L(4).a = 0;
L(4).b = (L(2).y - L(4).y) / (L(2).x - L(4).x);
L(2).a = 0;
L(2).b = (L(0).y - L(2).y) / (L(0).x - L(2).x);
theta = atan2(L(2).y - L(4).y, L(2).x - L(4).x);
len = hypot(L(2).x - L(4).x, L(2).y - L(4).y);
r = FFMIN(radius, len);
L(3).x = L(2).x - r * cos(theta);
L(3).y = L(2).y - r * sin(theta);
theta = atan2(L(0).y - L(2).y, L(0).x - L(2).x);
len = hypot(L(0).x - L(2).x, L(0).y - L(2).y);
r = FFMIN(radius, len / 2);
x = L(2).x + r * cos(theta);
y = L(2).y + r * sin(theta);
cx = (L(3).x + L(2).x + x) / 3;
cy = (L(3).y + L(2).y + y) / 3;
L(2).x = x;
L(2).y = y;
in1 = cx - L(3).x;
out1 = cy - L(3).y;
in2 = L(2).x - L(3).x;
out2 = L(2).y - L(3).y;
L(3).a = (out2 / in2 - out1 / in1) / (in2 - in1);
L(3).b = out1 / in1 - L(3).a * in1;
}
L(3).x = 0;
L(3).y = L(2).y;
s->in_min_lin = exp(s->segments[1].x);
s->out_min_lin = exp(s->segments[1].y);
return 0;
}
static void square_quadratic(double const *x, double *y)
{
y[0] = x[0] * x[0];
y[1] = 2 * x[0] * x[1];
y[2] = 2 * x[0] * x[2] + x[1] * x[1];
y[3] = 2 * x[1] * x[2];
y[4] = x[2] * x[2];
}
static int crossover_setup(AVFilterLink *outlink, Crossover *p, double frequency)
{
double w0 = 2 * M_PI * frequency / outlink->sample_rate;
double Q = sqrt(.5), alpha = sin(w0) / (2*Q);
double x[9], norm;
int i;
if (w0 > M_PI)
return AVERROR(EINVAL);
x[0] = (1 - cos(w0))/2; /* Cf. filter_LPF in biquads.c */
x[1] = 1 - cos(w0);
x[2] = (1 - cos(w0))/2;
x[3] = (1 + cos(w0))/2; /* Cf. filter_HPF in biquads.c */
x[4] = -(1 + cos(w0));
x[5] = (1 + cos(w0))/2;
x[6] = 1 + alpha;
x[7] = -2*cos(w0);
x[8] = 1 - alpha;
for (norm = x[6], i = 0; i < 9; ++i)
x[i] /= norm;
square_quadratic(x , p->coefs);
square_quadratic(x + 3, p->coefs + 5);
square_quadratic(x + 6, p->coefs + 10);
p->previous = av_calloc(outlink->channels, sizeof(*p->previous));
if (!p->previous)
return AVERROR(ENOMEM);
return 0;
}
static int config_output(AVFilterLink *outlink)
{
AVFilterContext *ctx = outlink->src;
MCompandContext *s = ctx->priv;
int ret, ch, i, k, new_nb_items, nb_bands;
char *p = s->args, *saveptr = NULL;
int max_delay_size = 0;
count_items(s->args, &nb_bands, '|');
s->nb_bands = FFMAX(1, nb_bands);
s->bands = av_calloc(nb_bands, sizeof(*s->bands));
if (!s->bands)
return AVERROR(ENOMEM);
for (i = 0, new_nb_items = 0; i < nb_bands; i++) {
int nb_points, nb_attacks, nb_items = 0;
char *tstr2, *tstr = av_strtok(p, "|", &saveptr);
char *p2, *p3, *saveptr2 = NULL, *saveptr3 = NULL;
double radius;
if (!tstr)
return AVERROR(EINVAL);
p = NULL;
p2 = tstr;
count_items(tstr, &nb_items, ' ');
tstr2 = av_strtok(p2, " ", &saveptr2);
if (!tstr2) {
av_log(ctx, AV_LOG_ERROR, "at least one attacks/decays rate is mandatory\n");
return AVERROR(EINVAL);
}
p2 = NULL;
p3 = tstr2;
count_items(tstr2, &nb_attacks, ',');
if (!nb_attacks || nb_attacks & 1) {
av_log(ctx, AV_LOG_ERROR, "number of attacks rate plus decays rate must be even\n");
return AVERROR(EINVAL);
}
s->bands[i].attack_rate = av_calloc(outlink->channels, sizeof(double));
s->bands[i].decay_rate = av_calloc(outlink->channels, sizeof(double));
s->bands[i].volume = av_calloc(outlink->channels, sizeof(double));
if (!s->bands[i].attack_rate || !s->bands[i].decay_rate || !s->bands[i].volume)
return AVERROR(ENOMEM);
for (k = 0; k < FFMIN(nb_attacks / 2, outlink->channels); k++) {
char *tstr3 = av_strtok(p3, ",", &saveptr3);
p3 = NULL;
sscanf(tstr3, "%lf", &s->bands[i].attack_rate[k]);
tstr3 = av_strtok(p3, ",", &saveptr3);
sscanf(tstr3, "%lf", &s->bands[i].decay_rate[k]);
if (s->bands[i].attack_rate[k] > 1.0 / outlink->sample_rate) {
s->bands[i].attack_rate[k] = 1.0 - exp(-1.0 / (outlink->sample_rate * s->bands[i].attack_rate[k]));
} else {
s->bands[i].attack_rate[k] = 1.0;
}
if (s->bands[i].decay_rate[k] > 1.0 / outlink->sample_rate) {
s->bands[i].decay_rate[k] = 1.0 - exp(-1.0 / (outlink->sample_rate * s->bands[i].decay_rate[k]));
} else {
s->bands[i].decay_rate[k] = 1.0;
}
}
for (ch = k; ch < outlink->channels; ch++) {
s->bands[i].attack_rate[ch] = s->bands[i].attack_rate[k - 1];
s->bands[i].decay_rate[ch] = s->bands[i].decay_rate[k - 1];
}
tstr2 = av_strtok(p2, " ", &saveptr2);
if (!tstr2) {
av_log(ctx, AV_LOG_ERROR, "transfer function curve in dB must be set\n");
return AVERROR(EINVAL);
}
sscanf(tstr2, "%lf", &s->bands[i].transfer_fn.curve_dB);
radius = s->bands[i].transfer_fn.curve_dB * M_LN10 / 20.0;
tstr2 = av_strtok(p2, " ", &saveptr2);
if (!tstr2) {
av_log(ctx, AV_LOG_ERROR, "transfer points missing\n");
return AVERROR(EINVAL);
}
count_items(tstr2, &nb_points, ',');
s->bands[i].transfer_fn.nb_segments = (nb_points + 4) * 2;
s->bands[i].transfer_fn.segments = av_calloc(s->bands[i].transfer_fn.nb_segments,
sizeof(CompandSegment));
if (!s->bands[i].transfer_fn.segments)
return AVERROR(ENOMEM);
ret = parse_points(tstr2, nb_points, radius, &s->bands[i].transfer_fn, ctx);
if (ret < 0) {
av_log(ctx, AV_LOG_ERROR, "transfer points parsing failed\n");
return ret;
}
tstr2 = av_strtok(p2, " ", &saveptr2);
if (!tstr2) {
av_log(ctx, AV_LOG_ERROR, "crossover_frequency is missing\n");
return AVERROR(EINVAL);
}
new_nb_items += sscanf(tstr2, "%lf", &s->bands[i].topfreq) == 1;
if (s->bands[i].topfreq < 0 || s->bands[i].topfreq >= outlink->sample_rate / 2) {
av_log(ctx, AV_LOG_ERROR, "crossover_frequency: %f, should be >=0 and lower than half of sample rate: %d.\n", s->bands[i].topfreq, outlink->sample_rate / 2);
return AVERROR(EINVAL);
}
if (s->bands[i].topfreq != 0) {
ret = crossover_setup(outlink, &s->bands[i].filter, s->bands[i].topfreq);
if (ret < 0)
return ret;
}
tstr2 = av_strtok(p2, " ", &saveptr2);
if (tstr2) {
sscanf(tstr2, "%lf", &s->bands[i].delay);
max_delay_size = FFMAX(max_delay_size, s->bands[i].delay * outlink->sample_rate);
tstr2 = av_strtok(p2, " ", &saveptr2);
if (tstr2) {
double initial_volume;
sscanf(tstr2, "%lf", &initial_volume);
initial_volume = pow(10.0, initial_volume / 20);
for (k = 0; k < outlink->channels; k++) {
s->bands[i].volume[k] = initial_volume;
}
tstr2 = av_strtok(p2, " ", &saveptr2);
if (tstr2) {
sscanf(tstr2, "%lf", &s->bands[i].transfer_fn.gain_dB);
}
}
}
}
s->nb_bands = new_nb_items;
for (i = 0; max_delay_size > 0 && i < s->nb_bands; i++) {
s->bands[i].delay_buf = ff_get_audio_buffer(outlink, max_delay_size);
if (!s->bands[i].delay_buf)
return AVERROR(ENOMEM);
}
s->delay_buf_size = max_delay_size;
return 0;
}
#define CONVOLVE _ _ _ _
static void crossover(int ch, Crossover *p,
double *ibuf, double *obuf_low,
double *obuf_high, size_t len)
{
double out_low, out_high;
while (len--) {
p->pos = p->pos ? p->pos - 1 : N - 1;
#define _ out_low += p->coefs[j] * p->previous[ch][p->pos + j].in \
- p->coefs[2*N+2 + j] * p->previous[ch][p->pos + j].out_low, j++;
{
int j = 1;
out_low = p->coefs[0] * *ibuf;
CONVOLVE
*obuf_low++ = out_low;
}
#undef _
#define _ out_high += p->coefs[j+N+1] * p->previous[ch][p->pos + j].in \
- p->coefs[2*N+2 + j] * p->previous[ch][p->pos + j].out_high, j++;
{
int j = 1;
out_high = p->coefs[N+1] * *ibuf;
CONVOLVE
*obuf_high++ = out_high;
}
p->previous[ch][p->pos + N].in = p->previous[ch][p->pos].in = *ibuf++;
p->previous[ch][p->pos + N].out_low = p->previous[ch][p->pos].out_low = out_low;
p->previous[ch][p->pos + N].out_high = p->previous[ch][p->pos].out_high = out_high;
}
}
static int mcompand_channel(MCompandContext *c, CompBand *l, double *ibuf, double *obuf, int len, int ch)
{
int i;
for (i = 0; i < len; i++) {
double level_in_lin, level_out_lin, checkbuf;
/* Maintain the volume fields by simulating a leaky pump circuit */
update_volume(l, fabs(ibuf[i]), ch);
/* Volume memory is updated: perform compand */
level_in_lin = l->volume[ch];
level_out_lin = get_volume(&l->transfer_fn, level_in_lin);
if (c->delay_buf_size <= 0) {
checkbuf = ibuf[i] * level_out_lin;
obuf[i] = checkbuf;
} else {
double *delay_buf = (double *)l->delay_buf->extended_data[ch];
/* FIXME: note that this lookahead algorithm is really lame:
the response to a peak is released before the peak
arrives. */
/* because volume application delays differ band to band, but
total delay doesn't, the volume is applied in an iteration
preceding that in which the sample goes to obuf, except in
the band(s) with the longest vol app delay.
the offset between delay_buf_ptr and the sample to apply
vol to, is a constant equal to the difference between this
band's delay and the longest delay of all the bands. */
if (l->delay_buf_cnt >= l->delay_size) {
checkbuf =
delay_buf[(l->delay_buf_ptr +
c->delay_buf_size -
l->delay_size) % c->delay_buf_size] * level_out_lin;
delay_buf[(l->delay_buf_ptr + c->delay_buf_size -
l->delay_size) % c->delay_buf_size] = checkbuf;
}
if (l->delay_buf_cnt >= c->delay_buf_size) {
obuf[i] = delay_buf[l->delay_buf_ptr];
} else {
l->delay_buf_cnt++;
}
delay_buf[l->delay_buf_ptr++] = ibuf[i];
l->delay_buf_ptr %= c->delay_buf_size;
}
}
return 0;
}
static int filter_frame(AVFilterLink *inlink, AVFrame *in)
{
AVFilterContext *ctx = inlink->dst;
AVFilterLink *outlink = ctx->outputs[0];
MCompandContext *s = ctx->priv;
AVFrame *out, *abuf, *bbuf, *cbuf;
int ch, band, i;
out = ff_get_audio_buffer(outlink, in->nb_samples);
if (!out) {
av_frame_free(&in);
return AVERROR(ENOMEM);
}
if (s->band_samples < in->nb_samples) {
av_frame_free(&s->band_buf1);
av_frame_free(&s->band_buf2);
av_frame_free(&s->band_buf3);
s->band_buf1 = ff_get_audio_buffer(outlink, in->nb_samples);
s->band_buf2 = ff_get_audio_buffer(outlink, in->nb_samples);
s->band_buf3 = ff_get_audio_buffer(outlink, in->nb_samples);
s->band_samples = in->nb_samples;
}
for (ch = 0; ch < outlink->channels; ch++) {
double *a, *dst = (double *)out->extended_data[ch];
for (band = 0, abuf = in, bbuf = s->band_buf2, cbuf = s->band_buf1; band < s->nb_bands; band++) {
CompBand *b = &s->bands[band];
if (b->topfreq) {
crossover(ch, &b->filter, (double *)abuf->extended_data[ch],
(double *)bbuf->extended_data[ch], (double *)cbuf->extended_data[ch], in->nb_samples);
} else {
bbuf = abuf;
abuf = cbuf;
}
if (abuf == in)
abuf = s->band_buf3;
mcompand_channel(s, b, (double *)bbuf->extended_data[ch], (double *)abuf->extended_data[ch], out->nb_samples, ch);
a = (double *)abuf->extended_data[ch];
for (i = 0; i < out->nb_samples; i++) {
dst[i] += a[i];
}
FFSWAP(AVFrame *, abuf, cbuf);
}
}
out->pts = in->pts;
av_frame_free(&in);
return ff_filter_frame(outlink, out);
}
static int request_frame(AVFilterLink *outlink)
{
AVFilterContext *ctx = outlink->src;
int ret;
ret = ff_request_frame(ctx->inputs[0]);
return ret;
}
static const AVFilterPad mcompand_inputs[] = {
{
.name = "default",
.type = AVMEDIA_TYPE_AUDIO,
.filter_frame = filter_frame,
},
};
static const AVFilterPad mcompand_outputs[] = {
{
.name = "default",
.type = AVMEDIA_TYPE_AUDIO,
.request_frame = request_frame,
.config_props = config_output,
},
};
const AVFilter ff_af_mcompand = {
.name = "mcompand",
.description = NULL_IF_CONFIG_SMALL(
"Multiband Compress or expand audio dynamic range."),
.priv_size = sizeof(MCompandContext),
.priv_class = &mcompand_class,
.uninit = uninit,
FILTER_INPUTS(mcompand_inputs),
FILTER_OUTPUTS(mcompand_outputs),
FILTER_QUERY_FUNC(query_formats),
};