[FFmpeg-devel] [PATCH] astats filter
Stefano Sabatini
stefasab at gmail.com
Wed Apr 24 19:51:05 CEST 2013
On date Tuesday 2013-04-23 12:59:08 +0000, Paul B Mahol encoded:
> Signed-off-by: Paul B Mahol <onemda at gmail.com>
> ---
> doc/filters.texi | 44 ++++++++
> libavfilter/Makefile | 1 +
> libavfilter/af_astats.c | 287 +++++++++++++++++++++++++++++++++++++++++++++++
> libavfilter/allfilters.c | 1 +
> 4 files changed, 333 insertions(+)
> create mode 100644 libavfilter/af_astats.c
>
> diff --git a/doc/filters.texi b/doc/filters.texi
> index d5fda03..1e2363d 100644
> --- a/doc/filters.texi
> +++ b/doc/filters.texi
> @@ -990,6 +990,50 @@ the data is treated as if all the planes were concatenated.
> A list of Adler-32 checksums for each data plane.
> @end table
>
> + at section astats
> +
> +Display time domain statistical information about the audio channels.
> +Statistics are calculated and displayed for each audio channel and,
> +where applicable, an overall figure is also given.
> +
> +The filter accepts the following option:
> + at table @option
> + at item length
> +Short window length. Default is 50ms.
nit: specify the unit (I think it is seconds), and range.
Also it is not clear what this "short window" refers to.
> + at end table
> +
> +A description of each shown parameter follows:
> +
> + at table @option
> + at item DC offset
> +Mean amplitude displacement from zero.
> +
> + at item Min level
> +Minimal sample level.
> +
> + at item Max level
> +Maximal sample level.
> +
> + at item Peak level dB
> + at item RMS level dB
> +Standard peak and RMS level measured in dBFS.
> +
> + at item RMS peak dB
> + at item RMS through dB
> +Peak and trough values for RMS level measured over a short window.
trough or through?
> +
> + at item Crest factor
> +Standard ratio of peak to RMS level (note: not in dB).
> +
> + at item Flat factor
> +Flatness (i.e. consecutive samples with the same value) of the signal at its peak levels
> +(i.e. either @var{Min level} or @var{Max level}).
> +
> + at item Peak count
> +Number of occasions (not the number of samples) that the signal attained either
> + at var{Min level} or @var{Max level}.
> + at end table
> +
> @section astreamsync
>
> Forward two audio streams and control the order the buffers are forwarded.
> diff --git a/libavfilter/Makefile b/libavfilter/Makefile
> index 4fce503..2b2adcb 100644
> --- a/libavfilter/Makefile
> +++ b/libavfilter/Makefile
> @@ -69,6 +69,7 @@ OBJS-$(CONFIG_ASETRATE_FILTER) += af_asetrate.o
> OBJS-$(CONFIG_ASETTB_FILTER) += f_settb.o
> OBJS-$(CONFIG_ASHOWINFO_FILTER) += af_ashowinfo.o
> OBJS-$(CONFIG_ASPLIT_FILTER) += split.o
> +OBJS-$(CONFIG_ASTATS_FILTER) += af_astats.o
> OBJS-$(CONFIG_ASTREAMSYNC_FILTER) += af_astreamsync.o
> OBJS-$(CONFIG_ASYNCTS_FILTER) += af_asyncts.o
> OBJS-$(CONFIG_ATEMPO_FILTER) += af_atempo.o
> diff --git a/libavfilter/af_astats.c b/libavfilter/af_astats.c
> new file mode 100644
> index 0000000..547cfc2
> --- /dev/null
> +++ b/libavfilter/af_astats.c
> @@ -0,0 +1,287 @@
> +/*
> + * Copyright (c) 2009 Rob Sykes <robs at users.sourceforge.net>
> + * Copyright (c) 2013 Paul B Mahol
> + *
> + * 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
> + */
> +
> +#include <float.h>
> +
> +#include "libavutil/opt.h"
> +#include "audio.h"
> +#include "avfilter.h"
> +#include "internal.h"
> +
> +typedef struct ChannelStats {
> + double last;
> + double sigma_x, sigma_x2;
> + double avg_sigma_x2, min_sigma_x2, max_sigma_x2;
> + double min, max;
> + double min_run, max_run;
> + double min_runs, max_runs;
> + uint64_t min_count, max_count;
> + uint64_t nb_samples;
> +} ChannelStats;
> +
> +typedef struct {
> + const AVClass *class;
> + ChannelStats *chstats;
> + int nb_channels;
> + uint64_t tc_samples;
> + double time_constant;
> + double mult;
better names / doxyes?
> +} AudioStatsContext;
> +
> +#define OFFSET(x) offsetof(AudioStatsContext, x)
> +#define FLAGS AV_OPT_FLAG_AUDIO_PARAM|AV_OPT_FLAG_FILTERING_PARAM
> +
> +static const AVOption astats_options[] = {
> + { "length", "set the window length", OFFSET(time_constant), AV_OPT_TYPE_DOUBLE, {.dbl=.05}, .01, 10, FLAGS },
> + {NULL},
> +};
> +
> +AVFILTER_DEFINE_CLASS(astats);
> +
> +static int query_formats(AVFilterContext *ctx)
> +{
> + AVFilterFormats *formats;
> + AVFilterChannelLayouts *layouts;
> + static const enum AVSampleFormat sample_fmts[] = {
> + AV_SAMPLE_FMT_DBL, AV_SAMPLE_FMT_DBLP,
> + AV_SAMPLE_FMT_NONE
> + };
> +
> + layouts = ff_all_channel_layouts();
> + if (!layouts)
> + return AVERROR(ENOMEM);
> + ff_set_common_channel_layouts(ctx, layouts);
> +
> + formats = ff_make_format_list(sample_fmts);
> + if (!formats)
> + return AVERROR(ENOMEM);
> + ff_set_common_formats(ctx, formats);
> +
> + formats = ff_all_samplerates();
> + if (!formats)
> + return AVERROR(ENOMEM);
> + ff_set_common_samplerates(ctx, formats);
> +
> + return 0;
> +}
> +
> +static int config_output(AVFilterLink *outlink)
> +{
> + AudioStatsContext *s = outlink->src->priv;
> + int c;
> +
> + s->chstats = av_calloc(sizeof(*s->chstats), outlink->channels);
> + if (!s->chstats)
> + return AVERROR(ENOMEM);
> + s->nb_channels = outlink->channels;
> + s->mult = exp((-1 / s->time_constant / outlink->sample_rate));
> + s->tc_samples = 5 * s->time_constant * outlink->sample_rate + .5;
> +
> + for (c = 0; c < s->nb_channels; c++) {
> + ChannelStats *p = &s->chstats[c];
> +
> + p->min = p->min_sigma_x2 = DBL_MAX;
> + p->max = p->max_sigma_x2 = DBL_MIN;
> + }
> +
> + return 0;
> +}
> +
> +static inline void stat(AudioStatsContext *s, ChannelStats *p, double d)
> +{
> + if (d < p->min) {
> + p->min = d;
> + p->min_run = 1;
> + p->min_runs = 0;
> + p->min_count = 1;
> + } else if (d == p->min) {
> + p->min_count++;
> + p->min_run = d == p->last ? p->min_run + 1 : 1;
> + } else if (p->last == p->min) {
> + p->min_runs += p->min_run * p->min_run;
> + }
> +
> + if (d > p->max) {
> + p->max = d;
> + p->max_run = 1;
> + p->max_runs = 0;
> + p->max_count = 1;
> + } else if (d == p->max) {
> + p->max_count++;
> + p->max_run = d == p->last ? p->max_run + 1 : 1;
> + } else if (p->last == p->max) {
> + p->max_runs += p->max_run * p->max_run;
> + }
> +
> + p->sigma_x += d;
> + p->sigma_x2 += d * d;
> + p->avg_sigma_x2 = p->avg_sigma_x2 * s->mult + (1.0 - s->mult) * d * d;
> + p->last = d;
> +
> + if (p->nb_samples >= s->tc_samples) {
> + p->max_sigma_x2 = FFMAX(p->max_sigma_x2, p->avg_sigma_x2);
> + p->min_sigma_x2 = FFMIN(p->min_sigma_x2, p->avg_sigma_x2);
> + }
> + p->nb_samples++;
> +}
> +
> +static int filter_frame(AVFilterLink *inlink, AVFrame *buf)
> +{
> + AudioStatsContext *s = inlink->dst->priv;
> + const int channels = s->nb_channels;
> + int i, c;
> +
> + switch (inlink->format) {
> + case AV_SAMPLE_FMT_DBLP:
> + for (c = 0; c < channels; c++) {
> + ChannelStats *p = &s->chstats[c];
> + const double *src = (const double *)buf->extended_data[c];
> +
> + for (i = 0; i < buf->nb_samples; i++, src++)
> + stat(s, p, *src);
> + }
> + break;
> + case AV_SAMPLE_FMT_DBL: {
> + const double *src = (const double *)buf->extended_data[0];
> +
> + for (i = 0; i < buf->nb_samples; i++) {
> + for (c = 0; c < channels; c++, src++) {
> + ChannelStats *p = &s->chstats[c];
> +
> + stat(s, p, *src);
you can directly use s->chstats[c]
> + }
> + }
> + break; }
> + }
> +
> + return ff_filter_frame(inlink->dst->outputs[0], buf);
> +}
> +
> +#define LINEAR_TO_DB(x) (log10(x) * 20)
> +
> +static void print_stats(AVFilterContext *ctx)
> +{
> + AudioStatsContext *s = ctx->priv;
> + uint64_t min_count = 0, max_count = 0, nb_samples;
> + double min_runs = 0, max_runs = 0,
> + min = DBL_MAX, max = DBL_MIN,
> + max_sigma_x = 0,
> + sigma_x = 0,
> + sigma_x2 = 0,
> + min_sigma_x2 = DBL_MAX,
> + max_sigma_x2 = DBL_MIN;
> + int c;
> +
> + for (c = 0; c < s->nb_channels; c++) {
> + ChannelStats *p = &s->chstats[c];
> +
> + if (p->nb_samples < s->tc_samples)
> + p->min_sigma_x2 = p->max_sigma_x2 = p->sigma_x2 / p->nb_samples;
> +
> + min = FFMIN(min, p->min);
> + max = FFMAX(max, p->max);
> + min_sigma_x2 = FFMIN(min_sigma_x2, p->min_sigma_x2);
> + max_sigma_x2 = FFMAX(max_sigma_x2, p->max_sigma_x2);
> + sigma_x += p->sigma_x;
> + sigma_x2 += p->sigma_x2;
> + min_count += p->min_count;
> + max_count += p->max_count;
> + min_runs += p->min_runs;
> + max_runs += p->max_runs;
> + nb_samples += p->nb_samples;
> + if (fabs(p->sigma_x) > fabs(max_sigma_x))
> + max_sigma_x = p->sigma_x;
> +
> + av_log(ctx, AV_LOG_INFO, "Channel: %d\n", c + 1);
> + av_log(ctx, AV_LOG_INFO, "DC offset: %f\n", p->sigma_x / p->nb_samples);
> + av_log(ctx, AV_LOG_INFO, "Min level: %f\n", p->min);
> + av_log(ctx, AV_LOG_INFO, "Max level: %f\n", p->max);
> + av_log(ctx, AV_LOG_INFO, "Peak level dB: %f\n",
> + LINEAR_TO_DB(FFMAX(-p->min, p->max)));
> + av_log(ctx, AV_LOG_INFO, "RMS level dB: %f\n",
> + LINEAR_TO_DB(sqrt(p->sigma_x2 / p->nb_samples)));
> + av_log(ctx, AV_LOG_INFO, "RMS peak dB: %f\n",
> + LINEAR_TO_DB(sqrt(p->max_sigma_x2)));
> + if (p->min_sigma_x2 != 1)
> + av_log(ctx, AV_LOG_INFO, "RMS through dB: %f\n",
> + LINEAR_TO_DB(sqrt(p->min_sigma_x2)));
> + av_log(ctx, AV_LOG_INFO, "Crest factor: %f\n",
> + p->sigma_x2 ? FFMAX(-p->min, p->max) / sqrt(p->sigma_x2 / p->nb_samples) : 1);
> + av_log(ctx, AV_LOG_INFO, "Flat factor: %f\n",
> + LINEAR_TO_DB((p->min_runs + p->max_runs) / (p->min_count + p->max_count)));
> + av_log(ctx, AV_LOG_INFO, "Peak count: %lld\n", p->min_count + p->max_count);
> + }
> +
> + av_log(ctx, AV_LOG_INFO, "Overall\n");
> + av_log(ctx, AV_LOG_INFO, "DC offset: %f\n", sigma_x / nb_samples);
> + av_log(ctx, AV_LOG_INFO, "Min level: %f\n", min);
> + av_log(ctx, AV_LOG_INFO, "Max level: %f\n", max);
> + av_log(ctx, AV_LOG_INFO, "Peak level dB: %f\n",
> + LINEAR_TO_DB(FFMAX(-min, max)));
> + av_log(ctx, AV_LOG_INFO, "RMS level dB: %f\n",
> + LINEAR_TO_DB(sqrt(sigma_x2 / nb_samples)));
> + av_log(ctx, AV_LOG_INFO, "RMS peak dB: %f\n",
> + LINEAR_TO_DB(sqrt(max_sigma_x2)));
> + if (min_sigma_x2 != 1)
> + av_log(ctx, AV_LOG_INFO, "RMS through dB: %f\n",
> + LINEAR_TO_DB(sqrt(min_sigma_x2)));
> + av_log(ctx, AV_LOG_INFO, "Flat factor: %f\n",
> + LINEAR_TO_DB((min_runs + max_runs) / (min_count + max_count)));
> + av_log(ctx, AV_LOG_INFO, "Peak count: %f\n", (min_count + max_count) / (double)s->nb_channels);
I wonder what would be a sane way to propagate this information to the
outside world (through a log, or maybe using metadata?).
[...]
--
FFmpeg = Fostering Fast Mystic Proud Efficient Gadget
More information about the ffmpeg-devel
mailing list