[FFmpeg-devel] [PATCH 3/3] avfilter: add spectrumsynth filter

Paul B Mahol onemda at gmail.com
Mon Jan 11 00:21:53 CET 2016


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

Only gray input supported, chroma is ignored. (float pixel format would come handy)
Only fullframe slide mode implemented.
No overlaping and only rectangular window for now.
Only power of 2 height supported and tested.

---
 libavfilter/Makefile            |   1 +
 libavfilter/allfilters.c        |   1 +
 libavfilter/vaf_spectrumsynth.c | 362 ++++++++++++++++++++++++++++++++++++++++
 3 files changed, 364 insertions(+)
 create mode 100644 libavfilter/vaf_spectrumsynth.c

diff --git a/libavfilter/Makefile b/libavfilter/Makefile
index 689da73..0704c4e 100644
--- a/libavfilter/Makefile
+++ b/libavfilter/Makefile
@@ -290,6 +290,7 @@ OBJS-$(CONFIG_SHOWSPECTRUMPIC_FILTER)        += avf_showspectrum.o window_func.o
 OBJS-$(CONFIG_SHOWVOLUME_FILTER)             += avf_showvolume.o
 OBJS-$(CONFIG_SHOWWAVES_FILTER)              += avf_showwaves.o
 OBJS-$(CONFIG_SHOWWAVESPIC_FILTER)           += avf_showwaves.o
+OBJS-$(CONFIG_SPECTRUMSYNTH_FILTER)          += vaf_spectrumsynth.o
 
 # multimedia sources
 OBJS-$(CONFIG_AMOVIE_FILTER)                 += src_movie.o
diff --git a/libavfilter/allfilters.c b/libavfilter/allfilters.c
index 2267e88..d4815d6 100644
--- a/libavfilter/allfilters.c
+++ b/libavfilter/allfilters.c
@@ -310,6 +310,7 @@ void avfilter_register_all(void)
     REGISTER_FILTER(SHOWVOLUME,     showvolume,     avf);
     REGISTER_FILTER(SHOWWAVES,      showwaves,      avf);
     REGISTER_FILTER(SHOWWAVESPIC,   showwavespic,   avf);
+    REGISTER_FILTER(SPECTRUMSYNTH,  spectrumsynth,  vaf);
 
     /* multimedia sources */
     REGISTER_FILTER(AMOVIE,         amovie,         avsrc);
diff --git a/libavfilter/vaf_spectrumsynth.c b/libavfilter/vaf_spectrumsynth.c
new file mode 100644
index 0000000..aaafce2
--- /dev/null
+++ b/libavfilter/vaf_spectrumsynth.c
@@ -0,0 +1,362 @@
+/*
+ * Copyright (c) 2016 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 "libavcodec/avfft.h"
+#include "libavutil/avassert.h"
+#include "libavutil/channel_layout.h"
+#include "libavutil/opt.h"
+#include "libavutil/parseutils.h"
+#include "avfilter.h"
+#include "formats.h"
+#include "audio.h"
+#include "video.h"
+#include "internal.h"
+#include "window_func.h"
+
+enum MagnitudeScale { LINEAR, LOG, NB_SCALES };
+
+typedef struct SpectrumSynthContext {
+    const AVClass *class;
+    int sample_rate;
+    int channels;
+    int scale;
+    int win_func;
+
+    AVFrame *magnitude, *phase;
+    FFTContext *fft;            ///< Fast Fourier Transform context
+    int fft_bits;               ///< number of bits (FFT window size = 1<<fft_bits)
+    FFTComplex **fft_data;      ///< bins holder for each (displayed) channels
+    int nb_samples;
+    int64_t pts;
+} SpectrumSynthContext;
+
+#define OFFSET(x) offsetof(SpectrumSynthContext, x)
+#define A AV_OPT_FLAG_FILTERING_PARAM|AV_OPT_FLAG_AUDIO_PARAM
+
+static const AVOption spectrumsynth_options[] = {
+    { "sample_rate", "set sample rate",  OFFSET(sample_rate), AV_OPT_TYPE_INT, {.i64 = 44100}, 15,  INT_MAX, A },
+    { "channels",    "set channels",     OFFSET(channels), AV_OPT_TYPE_INT, {.i64 = 1}, 1, 8, A },
+    { "scale",       "set amplitude scale", OFFSET(scale), AV_OPT_TYPE_INT, {.i64 = LOG}, 0, NB_SCALES-1, A, "scale" },
+        { "lin",  "linear",      0, AV_OPT_TYPE_CONST, {.i64=LINEAR}, 0, 0, A, "scale" },
+        { "log",  "logarithmic", 0, AV_OPT_TYPE_CONST, {.i64=LOG},    0, 0, A, "scale" },
+    { "win_func",    "set window function", OFFSET(win_func), AV_OPT_TYPE_INT, {.i64 = 0}, 0, 0, A, "win_func" },
+        { "rect",    "Rectangular",      0, AV_OPT_TYPE_CONST, {.i64=WFUNC_RECT},     0, 0, A, "win_func" },
+    { NULL }
+};
+
+AVFILTER_DEFINE_CLASS(spectrumsynth);
+
+static int query_formats(AVFilterContext *ctx)
+{
+    SpectrumSynthContext *s = ctx->priv;
+    AVFilterFormats *formats = NULL;
+    AVFilterChannelLayouts *layout = NULL;
+    AVFilterLink *magnitude = ctx->inputs[0];
+    AVFilterLink *phase = ctx->inputs[1];
+    AVFilterLink *outlink = ctx->outputs[0];
+    static const enum AVSampleFormat sample_fmts[] = { AV_SAMPLE_FMT_FLTP, AV_SAMPLE_FMT_NONE };
+    static const enum AVPixelFormat pix_fmts[] = { AV_PIX_FMT_YUV444P, AV_PIX_FMT_YUVJ444P, AV_PIX_FMT_YUV444P16, AV_PIX_FMT_NONE };
+    int ret, sample_rates[] = { 48000, -1 };
+
+    formats = ff_make_format_list(sample_fmts);
+    if ((ret = ff_formats_ref         (formats, &outlink->in_formats        )) < 0 ||
+        (ret = ff_add_channel_layout  (&layout, FF_COUNT2LAYOUT(s->channels))) < 0 ||
+        (ret = ff_channel_layouts_ref (layout , &outlink->in_channel_layouts)) < 0)
+        return ret;
+
+    sample_rates[0] = s->sample_rate;
+    formats = ff_make_format_list(sample_rates);
+    if (!formats)
+        return AVERROR(ENOMEM);
+    if ((ret = ff_formats_ref(formats, &outlink->in_samplerates)) < 0)
+        return ret;
+
+    formats = ff_make_format_list(pix_fmts);
+    if (!formats)
+        return AVERROR(ENOMEM);
+    if ((ret = ff_formats_ref(formats, &magnitude->out_formats)) < 0)
+        return ret;
+
+    formats = ff_make_format_list(pix_fmts);
+    if (!formats)
+        return AVERROR(ENOMEM);
+    if ((ret = ff_formats_ref(formats, &phase->out_formats)) < 0)
+        return ret;
+
+    return 0;
+}
+
+static int config_output(AVFilterLink *outlink)
+{
+    AVFilterContext *ctx = outlink->src;
+    SpectrumSynthContext *s = ctx->priv;
+    int width = ctx->inputs[0]->w;
+    int height = ctx->inputs[0]->h;
+    AVRational time_base  = ctx->inputs[0]->time_base;
+    AVRational frame_rate = ctx->inputs[0]->frame_rate;
+    int ch, fft_bits;
+
+    outlink->sample_rate = s->sample_rate;
+    outlink->time_base = (AVRational){1, s->sample_rate};
+
+    if (width  != ctx->inputs[1]->w ||
+        height != ctx->inputs[1]->h) {
+        av_log(ctx, AV_LOG_ERROR,
+               "Magnitude and Phase sizes differ (%dx%d vs %dx%d).\n",
+               width, height,
+               ctx->inputs[1]->w, ctx->inputs[1]->h);
+        return AVERROR_INVALIDDATA;
+    } else if (av_cmp_q(time_base, ctx->inputs[1]->time_base) != 0) {
+        av_log(ctx, AV_LOG_ERROR,
+               "Magnitude and Phase time bases differ (%d/%d vs %d/%d).\n",
+               time_base.num, time_base.den,
+               ctx->inputs[1]->time_base.num,
+               ctx->inputs[1]->time_base.den);
+        return AVERROR_INVALIDDATA;
+    } else if (av_cmp_q(frame_rate, ctx->inputs[1]->frame_rate) != 0) {
+        av_log(ctx, AV_LOG_ERROR,
+               "Magnitude and Phase framerates differ (%d/%d vs %d/%d).\n",
+               frame_rate.num, frame_rate.den,
+               ctx->inputs[1]->frame_rate.num,
+               ctx->inputs[1]->frame_rate.den);
+        return AVERROR_INVALIDDATA;
+    }
+
+    for (fft_bits = 1; 1 << fft_bits < 2 * height / s->channels; fft_bits++);
+    s->nb_samples = 1 << fft_bits;
+
+    s->fft = av_fft_init(fft_bits, 1);
+    if (!s->fft) {
+        av_log(ctx, AV_LOG_ERROR, "Unable to create FFT context. "
+               "The window size might be too high.\n");
+        return AVERROR(EINVAL);
+    }
+    s->fft_data = av_calloc(s->channels, sizeof(*s->fft_data));
+    if (!s->fft_data)
+        return AVERROR(ENOMEM);
+    for (ch = 0; ch < s->channels; ch++) {
+        s->fft_data[ch] = av_calloc(s->nb_samples, sizeof(**s->fft_data));
+        if (!s->fft_data[ch])
+            return AVERROR(ENOMEM);
+    }
+
+    return 0;
+}
+
+static int request_frame(AVFilterLink *outlink)
+{
+    AVFilterContext *ctx = outlink->src;
+    SpectrumSynthContext *s = ctx->priv;
+    int ret;
+
+    if (!s->magnitude) {
+        ret = ff_request_frame(ctx->inputs[0]);
+        if (ret < 0)
+            return ret;
+    }
+    if (!s->phase) {
+        ret = ff_request_frame(ctx->inputs[1]);
+        if (ret < 0)
+            return ret;
+    }
+    return 0;
+}
+
+static int try_push_frame(AVFilterContext *ctx, int x)
+{
+    SpectrumSynthContext *s = ctx->priv;
+    AVFilterLink *outlink = ctx->outputs[0];
+    AVFilterLink *inlink = ctx->inputs[0];
+    int m_linesize, p_linesize;
+    int nb = s->nb_samples;
+    int h = inlink->h / s->channels;
+    int y, f, ch = 0, n;
+    float magnitude, phase;
+    AVFrame *out;
+
+    m_linesize = s->magnitude->linesize[0];
+    p_linesize = s->phase->linesize[0];
+
+    out = ff_get_audio_buffer(outlink, s->nb_samples);
+    if (!out) {
+        av_frame_free(&s->magnitude);
+        av_frame_free(&s->phase);
+        return AVERROR(ENOMEM);
+    }
+
+    for (ch = 0; ch < s->channels; ch++) {
+        int start = h * (ch + 1) - 1;
+        int end = h * ch;
+
+        switch (inlink->format) {
+        case AV_PIX_FMT_YUV444P16:
+            for (y = start, f = 0; y >= end; y--, f++) {
+                const uint16_t *m = (uint16_t *)(s->magnitude->data[0] + y * m_linesize);
+                const uint16_t *p = (uint16_t *)(s->phase->data[0] + y * p_linesize);
+
+                switch (s->scale) {
+                case LINEAR:
+                    magnitude = m[x] / (double)UINT16_MAX;
+                    break;
+                case LOG:
+                    magnitude = ff_exp10(((m[x] / (double)UINT16_MAX) - 1.) * 6.);
+                    break;
+                }
+                phase = ((p[x] / (double)UINT16_MAX) * 2. - 1.) * M_PI;
+
+                s->fft_data[ch][f].re = magnitude * cos(phase);
+                s->fft_data[ch][f].im = magnitude * sin(phase);
+            }
+            break;
+        case AV_PIX_FMT_YUVJ444P:
+        case AV_PIX_FMT_YUV444P:
+            for (y = start, f = 0; y >= end; y--, f++) {
+                const uint8_t *m = (uint8_t *)(s->magnitude->data[0] + y * m_linesize);
+                const uint8_t *p = (uint8_t *)(s->phase->data[0] + y * p_linesize);
+
+                switch (s->scale) {
+                case LINEAR:
+                    magnitude = m[x] / (double)UINT8_MAX;
+                    break;
+                case LOG:
+                    magnitude = ff_exp10(((m[x] / (double)UINT8_MAX) - 1.) * 6.);
+                    break;
+                }
+                phase = ((p[x] / (double)UINT8_MAX) * 2. - 1.) * M_PI;
+
+                s->fft_data[ch][f].re = magnitude * cos(phase);
+                s->fft_data[ch][f].im = magnitude * sin(phase);
+            }
+            break;
+        }
+
+        s->fft_data[ch][h].re = 0;
+        s->fft_data[ch][h].im = 0;
+        for (y = h + 1, f = h - 1; y < nb; y++, f--) {
+            s->fft_data[ch][y].re =  s->fft_data[ch][f].re;
+            s->fft_data[ch][y].im = -s->fft_data[ch][f].im;
+        }
+    }
+
+    for (ch = 0; ch < s->channels; ch++) {
+        float *dst = (float *)out->extended_data[ch];
+
+        av_fft_permute(s->fft, s->fft_data[ch]);
+        av_fft_calc(s->fft, s->fft_data[ch]);
+
+        for (n = 0; n < s->nb_samples; n++) {
+            dst[n] = s->fft_data[ch][n].re;
+        }
+    }
+
+    out->pts = s->pts;
+    s->pts += s->nb_samples;
+    return ff_filter_frame(outlink, out);
+}
+
+static int try_push_frames(AVFilterContext *ctx)
+{
+    SpectrumSynthContext *s = ctx->priv;
+    AVFilterLink *inlink = ctx->inputs[0];
+    int ret, x;
+
+    if (!(s->magnitude && s->phase))
+        return 0;
+
+    for (x = 0; x < inlink->w; x++) {
+        ret = try_push_frame(ctx, x);
+        if (ret < 0)
+            break;
+    }
+
+    av_frame_free(&s->magnitude);
+    av_frame_free(&s->phase);
+    return ret;
+}
+
+static int filter_frame_magnitude(AVFilterLink *inlink, AVFrame *magnitude)
+{
+    AVFilterContext *ctx = inlink->dst;
+    SpectrumSynthContext *s = ctx->priv;
+
+    s->magnitude = magnitude;
+    return try_push_frames(ctx);
+}
+
+static int filter_frame_phase(AVFilterLink *inlink, AVFrame *phase)
+{
+    AVFilterContext *ctx = inlink->dst;
+    SpectrumSynthContext *s = ctx->priv;
+
+    s->phase = phase;
+    return try_push_frames(ctx);
+}
+
+static av_cold void uninit(AVFilterContext *ctx)
+{
+    SpectrumSynthContext *s = ctx->priv;
+    int i;
+
+    av_frame_free(&s->magnitude);
+    av_frame_free(&s->phase);
+    av_fft_end(s->fft);
+    if (s->fft_data) {
+        for (i = 0; i < s->channels; i++)
+            av_freep(&s->fft_data[i]);
+    }
+    av_freep(&s->fft_data);
+}
+
+static const AVFilterPad spectrumsynth_inputs[] = {
+    {
+        .name         = "magnitude",
+        .type         = AVMEDIA_TYPE_VIDEO,
+        .filter_frame = filter_frame_magnitude,
+        .needs_fifo   = 1,
+    },
+    {
+        .name         = "phase",
+        .type         = AVMEDIA_TYPE_VIDEO,
+        .filter_frame = filter_frame_phase,
+        .needs_fifo   = 1,
+    },
+    { NULL }
+};
+
+static const AVFilterPad spectrumsynth_outputs[] = {
+    {
+        .name          = "default",
+        .type          = AVMEDIA_TYPE_AUDIO,
+        .config_props  = config_output,
+        .request_frame = request_frame,
+    },
+    { NULL }
+};
+
+AVFilter ff_vaf_spectrumsynth = {
+    .name          = "spectrumsynth",
+    .description   = NULL_IF_CONFIG_SMALL("Convert input spectrum videos to audio output."),
+    .uninit        = uninit,
+    .query_formats = query_formats,
+    .priv_size     = sizeof(SpectrumSynthContext),
+    .inputs        = spectrumsynth_inputs,
+    .outputs       = spectrumsynth_outputs,
+    .priv_class    = &spectrumsynth_class,
+};
-- 
1.9.1



More information about the ffmpeg-devel mailing list