2 * Copyright (c) 2015 Paul B Mahol
4 * This file is part of FFmpeg.
6 * FFmpeg is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU Lesser General Public
8 * License as published by the Free Software Foundation; either
9 * version 2.1 of the License, or (at your option) any later version.
11 * FFmpeg is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 * Lesser General Public License for more details.
16 * You should have received a copy of the GNU Lesser General Public
17 * License along with FFmpeg; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
24 #include "libavcodec/avfft.h"
25 #include "libavutil/audio_fifo.h"
26 #include "libavutil/avassert.h"
27 #include "libavutil/avstring.h"
28 #include "libavutil/channel_layout.h"
29 #include "libavutil/intreadwrite.h"
30 #include "libavutil/opt.h"
31 #include "libavutil/parseutils.h"
37 #include "window_func.h"
39 enum DataMode { MAGNITUDE, PHASE, NB_DATA };
40 enum DisplayMode { LINE, BAR, DOT, NB_MODES };
41 enum ChannelMode { COMBINED, SEPARATE, NB_CMODES };
42 enum FrequencyScale { FS_LINEAR, FS_LOG, FS_RLOG, NB_FSCALES };
43 enum AmplitudeScale { AS_LINEAR, AS_SQRT, AS_CBRT, AS_LOG, NB_ASCALES };
45 typedef struct ShowFreqsContext {
57 FFTComplex **fft_data;
59 float *window_func_lut;
72 #define OFFSET(x) offsetof(ShowFreqsContext, x)
73 #define FLAGS AV_OPT_FLAG_FILTERING_PARAM|AV_OPT_FLAG_VIDEO_PARAM
75 static const AVOption showfreqs_options[] = {
76 { "size", "set video size", OFFSET(w), AV_OPT_TYPE_IMAGE_SIZE, {.str = "1024x512"}, 0, 0, FLAGS },
77 { "s", "set video size", OFFSET(w), AV_OPT_TYPE_IMAGE_SIZE, {.str = "1024x512"}, 0, 0, FLAGS },
78 { "mode", "set display mode", OFFSET(mode), AV_OPT_TYPE_INT, {.i64=BAR}, 0, NB_MODES-1, FLAGS, "mode" },
79 { "line", "show lines", 0, AV_OPT_TYPE_CONST, {.i64=LINE}, 0, 0, FLAGS, "mode" },
80 { "bar", "show bars", 0, AV_OPT_TYPE_CONST, {.i64=BAR}, 0, 0, FLAGS, "mode" },
81 { "dot", "show dots", 0, AV_OPT_TYPE_CONST, {.i64=DOT}, 0, 0, FLAGS, "mode" },
82 { "ascale", "set amplitude scale", OFFSET(ascale), AV_OPT_TYPE_INT, {.i64=AS_LOG}, 0, NB_ASCALES-1, FLAGS, "ascale" },
83 { "lin", "linear", 0, AV_OPT_TYPE_CONST, {.i64=AS_LINEAR}, 0, 0, FLAGS, "ascale" },
84 { "sqrt", "square root", 0, AV_OPT_TYPE_CONST, {.i64=AS_SQRT}, 0, 0, FLAGS, "ascale" },
85 { "cbrt", "cubic root", 0, AV_OPT_TYPE_CONST, {.i64=AS_CBRT}, 0, 0, FLAGS, "ascale" },
86 { "log", "logarithmic", 0, AV_OPT_TYPE_CONST, {.i64=AS_LOG}, 0, 0, FLAGS, "ascale" },
87 { "fscale", "set frequency scale", OFFSET(fscale), AV_OPT_TYPE_INT, {.i64=FS_LINEAR}, 0, NB_FSCALES-1, FLAGS, "fscale" },
88 { "lin", "linear", 0, AV_OPT_TYPE_CONST, {.i64=FS_LINEAR}, 0, 0, FLAGS, "fscale" },
89 { "log", "logarithmic", 0, AV_OPT_TYPE_CONST, {.i64=FS_LOG}, 0, 0, FLAGS, "fscale" },
90 { "rlog", "reverse logarithmic", 0, AV_OPT_TYPE_CONST, {.i64=FS_RLOG}, 0, 0, FLAGS, "fscale" },
91 { "win_size", "set window size", OFFSET(fft_size), AV_OPT_TYPE_INT, {.i64=2048}, 16, 65536, FLAGS },
92 { "win_func", "set window function", OFFSET(win_func), AV_OPT_TYPE_INT, {.i64=WFUNC_HANNING}, 0, NB_WFUNC-1, FLAGS, "win_func" },
93 { "rect", "Rectangular", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_RECT}, 0, 0, FLAGS, "win_func" },
94 { "bartlett", "Bartlett", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_BARTLETT}, 0, 0, FLAGS, "win_func" },
95 { "hanning", "Hanning", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_HANNING}, 0, 0, FLAGS, "win_func" },
96 { "hamming", "Hamming", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_HAMMING}, 0, 0, FLAGS, "win_func" },
97 { "blackman", "Blackman", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_BLACKMAN}, 0, 0, FLAGS, "win_func" },
98 { "welch", "Welch", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_WELCH}, 0, 0, FLAGS, "win_func" },
99 { "flattop", "Flat-top", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_FLATTOP}, 0, 0, FLAGS, "win_func" },
100 { "bharris", "Blackman-Harris", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_BHARRIS}, 0, 0, FLAGS, "win_func" },
101 { "bnuttall", "Blackman-Nuttall", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_BNUTTALL}, 0, 0, FLAGS, "win_func" },
102 { "bhann", "Bartlett-Hann", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_BHANN}, 0, 0, FLAGS, "win_func" },
103 { "sine", "Sine", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_SINE}, 0, 0, FLAGS, "win_func" },
104 { "nuttall", "Nuttall", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_NUTTALL}, 0, 0, FLAGS, "win_func" },
105 { "lanczos", "Lanczos", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_LANCZOS}, 0, 0, FLAGS, "win_func" },
106 { "gauss", "Gauss", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_GAUSS}, 0, 0, FLAGS, "win_func" },
107 { "tukey", "Tukey", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_TUKEY}, 0, 0, FLAGS, "win_func" },
108 { "dolph", "Dolph-Chebyshev", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_DOLPH}, 0, 0, FLAGS, "win_func" },
109 { "cauchy", "Cauchy", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_CAUCHY}, 0, 0, FLAGS, "win_func" },
110 { "parzen", "Parzen", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_PARZEN}, 0, 0, FLAGS, "win_func" },
111 { "poisson", "Poisson", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_POISSON}, 0, 0, FLAGS, "win_func" },
112 { "bohman", "Bohman", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_BOHMAN} , 0, 0, FLAGS, "win_func" },
113 { "overlap", "set window overlap", OFFSET(overlap), AV_OPT_TYPE_FLOAT, {.dbl=1.}, 0., 1., FLAGS },
114 { "averaging", "set time averaging", OFFSET(avg), AV_OPT_TYPE_INT, {.i64=1}, 0, INT32_MAX, FLAGS },
115 { "colors", "set channels colors", OFFSET(colors), AV_OPT_TYPE_STRING, {.str = "red|green|blue|yellow|orange|lime|pink|magenta|brown" }, 0, 0, FLAGS },
116 { "cmode", "set channel mode", OFFSET(cmode), AV_OPT_TYPE_INT, {.i64=COMBINED}, 0, NB_CMODES-1, FLAGS, "cmode" },
117 { "combined", "show all channels in same window", 0, AV_OPT_TYPE_CONST, {.i64=COMBINED}, 0, 0, FLAGS, "cmode" },
118 { "separate", "show each channel in own window", 0, AV_OPT_TYPE_CONST, {.i64=SEPARATE}, 0, 0, FLAGS, "cmode" },
119 { "minamp", "set minimum amplitude", OFFSET(minamp), AV_OPT_TYPE_FLOAT, {.dbl=1e-6}, FLT_MIN, 1e-6, FLAGS },
120 { "data", "set data mode", OFFSET(data_mode), AV_OPT_TYPE_INT, {.i64=MAGNITUDE}, 0, NB_DATA-1, FLAGS, "data" },
121 { "magnitude", "show magnitude", 0, AV_OPT_TYPE_CONST, {.i64=MAGNITUDE}, 0, 0, FLAGS, "data" },
122 { "phase", "show phase", 0, AV_OPT_TYPE_CONST, {.i64=PHASE}, 0, 0, FLAGS, "data" },
126 AVFILTER_DEFINE_CLASS(showfreqs);
128 static int query_formats(AVFilterContext *ctx)
130 AVFilterFormats *formats = NULL;
131 AVFilterChannelLayouts *layouts = NULL;
132 AVFilterLink *inlink = ctx->inputs[0];
133 AVFilterLink *outlink = ctx->outputs[0];
134 static const enum AVSampleFormat sample_fmts[] = { AV_SAMPLE_FMT_FLTP, AV_SAMPLE_FMT_NONE };
135 static const enum AVPixelFormat pix_fmts[] = { AV_PIX_FMT_RGBA, AV_PIX_FMT_NONE };
138 /* set input audio formats */
139 formats = ff_make_format_list(sample_fmts);
140 if ((ret = ff_formats_ref(formats, &inlink->outcfg.formats)) < 0)
143 layouts = ff_all_channel_layouts();
144 if ((ret = ff_channel_layouts_ref(layouts, &inlink->outcfg.channel_layouts)) < 0)
147 formats = ff_all_samplerates();
148 if ((ret = ff_formats_ref(formats, &inlink->outcfg.samplerates)) < 0)
151 /* set output video format */
152 formats = ff_make_format_list(pix_fmts);
153 if ((ret = ff_formats_ref(formats, &outlink->incfg.formats)) < 0)
159 static av_cold int init(AVFilterContext *ctx)
161 ShowFreqsContext *s = ctx->priv;
163 s->pts = AV_NOPTS_VALUE;
168 static int config_output(AVFilterLink *outlink)
170 AVFilterContext *ctx = outlink->src;
171 AVFilterLink *inlink = ctx->inputs[0];
172 ShowFreqsContext *s = ctx->priv;
176 s->fft_bits = av_log2(s->fft_size);
177 s->nb_freq = 1 << (s->fft_bits - 1);
178 s->win_size = s->nb_freq << 1;
179 av_audio_fifo_free(s->fifo);
181 s->fft = av_fft_init(s->fft_bits, 0);
183 av_log(ctx, AV_LOG_ERROR, "Unable to create FFT context. "
184 "The window size might be too high.\n");
185 return AVERROR(ENOMEM);
188 /* FFT buffers: x2 for each (display) channel buffer.
189 * Note: we use free and malloc instead of a realloc-like function to
190 * make sure the buffer is aligned in memory for the FFT functions. */
191 for (i = 0; i < s->nb_channels; i++) {
192 av_freep(&s->fft_data[i]);
193 av_freep(&s->avg_data[i]);
195 av_freep(&s->fft_data);
196 av_freep(&s->avg_data);
197 s->nb_channels = inlink->channels;
199 s->fft_data = av_calloc(s->nb_channels, sizeof(*s->fft_data));
201 return AVERROR(ENOMEM);
202 s->avg_data = av_calloc(s->nb_channels, sizeof(*s->avg_data));
204 return AVERROR(ENOMEM);
205 for (i = 0; i < s->nb_channels; i++) {
206 s->fft_data[i] = av_calloc(s->win_size, sizeof(**s->fft_data));
207 s->avg_data[i] = av_calloc(s->nb_freq, sizeof(**s->avg_data));
208 if (!s->fft_data[i] || !s->avg_data[i])
209 return AVERROR(ENOMEM);
212 /* pre-calc windowing function */
213 s->window_func_lut = av_realloc_f(s->window_func_lut, s->win_size,
214 sizeof(*s->window_func_lut));
215 if (!s->window_func_lut)
216 return AVERROR(ENOMEM);
217 generate_window_func(s->window_func_lut, s->win_size, s->win_func, &overlap);
218 if (s->overlap == 1.)
219 s->overlap = overlap;
220 s->hop_size = (1. - s->overlap) * s->win_size;
221 if (s->hop_size < 1) {
222 av_log(ctx, AV_LOG_ERROR, "overlap %f too big\n", s->overlap);
223 return AVERROR(EINVAL);
226 for (s->scale = 0, i = 0; i < s->win_size; i++) {
227 s->scale += s->window_func_lut[i] * s->window_func_lut[i];
230 outlink->frame_rate = av_make_q(inlink->sample_rate, s->win_size * (1.-s->overlap));
231 outlink->sample_aspect_ratio = (AVRational){1,1};
235 s->fifo = av_audio_fifo_alloc(inlink->format, inlink->channels, s->win_size);
237 return AVERROR(ENOMEM);
241 static inline void draw_dot(AVFrame *out, int x, int y, uint8_t fg[4])
244 uint32_t color = AV_RL32(out->data[0] + y * out->linesize[0] + x * 4);
246 if ((color & 0xffffff) != 0)
247 AV_WL32(out->data[0] + y * out->linesize[0] + x * 4, AV_RL32(fg) | color);
249 AV_WL32(out->data[0] + y * out->linesize[0] + x * 4, AV_RL32(fg));
252 static int get_sx(ShowFreqsContext *s, int f)
256 return (s->w/(float)s->nb_freq)*f;
258 return s->w-pow(s->w, (s->nb_freq-f-1)/(s->nb_freq-1.));
260 return pow(s->w, f/(s->nb_freq-1.));
266 static float get_bsize(ShowFreqsContext *s, int f)
270 return s->w/(float)s->nb_freq;
272 return pow(s->w, (s->nb_freq-f-1)/(s->nb_freq-1.))-
273 pow(s->w, (s->nb_freq-f-2)/(s->nb_freq-1.));
275 return pow(s->w, (f+1)/(s->nb_freq-1.))-
276 pow(s->w, f /(s->nb_freq-1.));
282 static inline void plot_freq(ShowFreqsContext *s, int ch,
283 double a, int f, uint8_t fg[4], int *prev_y,
284 AVFrame *out, AVFilterLink *outlink)
287 const float min = s->minamp;
288 const float avg = s->avg_data[ch][f];
289 const float bsize = get_bsize(s, f);
290 const int sx = get_sx(s, f);
291 int end = outlink->h;
302 a = log(av_clipd(a, min, 1)) / log(min);
311 y = a * outlink->h - 1;
314 end = (outlink->h / s->nb_channels) * (ch + 1);
315 y = (outlink->h / s->nb_channels) * ch + a * (outlink->h / s->nb_channels) - 1;
325 y = s->avg_data[ch][f] = !outlink->frame_count_in ? y : FFMIN(avg, y);
330 s->avg_data[ch][f] = avg + y * (y - avg) / (FFMIN(outlink->frame_count_in + 1, s->avg) * y);
331 y = s->avg_data[ch][f];
341 for (x = sx + 1; x < sx + bsize && x < w; x++)
342 draw_dot(out, x, y, fg);
343 for (i = y; i <= *prev_y; i++)
344 draw_dot(out, sx, i, fg);
346 for (i = *prev_y; i <= y; i++)
347 draw_dot(out, sx, i, fg);
348 for (x = sx + 1; x < sx + bsize && x < w; x++)
349 draw_dot(out, x, i - 1, fg);
354 for (x = sx; x < sx + bsize && x < w; x++)
355 for (i = y; i < end; i++)
356 draw_dot(out, x, i, fg);
359 for (x = sx; x < sx + bsize && x < w; x++)
360 draw_dot(out, x, y, fg);
365 static int plot_freqs(AVFilterLink *inlink, AVFrame *in)
367 AVFilterContext *ctx = inlink->dst;
368 AVFilterLink *outlink = ctx->outputs[0];
369 ShowFreqsContext *s = ctx->priv;
370 const int win_size = s->win_size;
371 char *colors, *color, *saveptr = NULL;
375 out = ff_get_video_buffer(outlink, outlink->w, outlink->h);
377 return AVERROR(ENOMEM);
379 for (n = 0; n < outlink->h; n++)
380 memset(out->data[0] + out->linesize[0] * n, 0, outlink->w * 4);
382 /* fill FFT input with the number of samples available */
383 for (ch = 0; ch < s->nb_channels; ch++) {
384 const float *p = (float *)in->extended_data[ch];
386 for (n = 0; n < in->nb_samples; n++) {
387 s->fft_data[ch][n].re = p[n] * s->window_func_lut[n];
388 s->fft_data[ch][n].im = 0;
390 for (; n < win_size; n++) {
391 s->fft_data[ch][n].re = 0;
392 s->fft_data[ch][n].im = 0;
396 /* run FFT on each samples set */
397 for (ch = 0; ch < s->nb_channels; ch++) {
398 av_fft_permute(s->fft, s->fft_data[ch]);
399 av_fft_calc(s->fft, s->fft_data[ch]);
402 #define RE(x, ch) s->fft_data[ch][x].re
403 #define IM(x, ch) s->fft_data[ch][x].im
404 #define M(a, b) (sqrt((a) * (a) + (b) * (b)))
405 #define P(a, b) (atan2((b), (a)))
407 colors = av_strdup(s->colors);
410 return AVERROR(ENOMEM);
413 for (ch = 0; ch < s->nb_channels; ch++) {
414 uint8_t fg[4] = { 0xff, 0xff, 0xff, 0xff };
418 color = av_strtok(ch == 0 ? colors : NULL, " |", &saveptr);
420 av_parse_color(fg, color, -1, ctx);
422 switch (s->data_mode) {
424 a = av_clipd(M(RE(0, ch), 0) / s->scale, 0, 1);
425 plot_freq(s, ch, a, 0, fg, &prev_y, out, outlink);
427 for (f = 1; f < s->nb_freq; f++) {
428 a = av_clipd(M(RE(f, ch), IM(f, ch)) / s->scale, 0, 1);
430 plot_freq(s, ch, a, f, fg, &prev_y, out, outlink);
434 a = av_clipd((M_PI + P(RE(0, ch), 0)) / (2. * M_PI), 0, 1);
435 plot_freq(s, ch, a, 0, fg, &prev_y, out, outlink);
437 for (f = 1; f < s->nb_freq; f++) {
438 a = av_clipd((M_PI + P(RE(f, ch), IM(f, ch))) / (2. * M_PI), 0, 1);
440 plot_freq(s, ch, a, f, fg, &prev_y, out, outlink);
448 out->sample_aspect_ratio = (AVRational){1,1};
449 return ff_filter_frame(outlink, out);
452 static int filter_frame(AVFilterLink *inlink)
454 AVFilterContext *ctx = inlink->dst;
455 ShowFreqsContext *s = ctx->priv;
459 fin = ff_get_audio_buffer(inlink, s->win_size);
461 ret = AVERROR(ENOMEM);
466 s->pts += s->hop_size;
467 ret = av_audio_fifo_peek(s->fifo, (void **)fin->extended_data, s->win_size);
471 ret = plot_freqs(inlink, fin);
473 av_audio_fifo_drain(s->fifo, s->hop_size);
480 static int activate(AVFilterContext *ctx)
482 AVFilterLink *inlink = ctx->inputs[0];
483 AVFilterLink *outlink = ctx->outputs[0];
484 ShowFreqsContext *s = ctx->priv;
488 FF_FILTER_FORWARD_STATUS_BACK(outlink, inlink);
490 if (av_audio_fifo_size(s->fifo) < s->win_size)
491 ret = ff_inlink_consume_samples(inlink, s->win_size, s->win_size, &in);
495 av_audio_fifo_write(s->fifo, (void **)in->extended_data, in->nb_samples);
496 if (s->pts == AV_NOPTS_VALUE)
501 if (av_audio_fifo_size(s->fifo) >= s->win_size) {
502 ret = filter_frame(inlink);
507 FF_FILTER_FORWARD_STATUS(inlink, outlink);
508 FF_FILTER_FORWARD_WANTED(outlink, inlink);
510 return FFERROR_NOT_READY;
513 static av_cold void uninit(AVFilterContext *ctx)
515 ShowFreqsContext *s = ctx->priv;
519 for (i = 0; i < s->nb_channels; i++) {
521 av_freep(&s->fft_data[i]);
523 av_freep(&s->avg_data[i]);
525 av_freep(&s->fft_data);
526 av_freep(&s->avg_data);
527 av_freep(&s->window_func_lut);
528 av_audio_fifo_free(s->fifo);
531 static const AVFilterPad showfreqs_inputs[] = {
534 .type = AVMEDIA_TYPE_AUDIO,
539 static const AVFilterPad showfreqs_outputs[] = {
542 .type = AVMEDIA_TYPE_VIDEO,
543 .config_props = config_output,
548 AVFilter ff_avf_showfreqs = {
550 .description = NULL_IF_CONFIG_SMALL("Convert input audio to a frequencies video output."),
553 .query_formats = query_formats,
554 .priv_size = sizeof(ShowFreqsContext),
555 .activate = activate,
556 .inputs = showfreqs_inputs,
557 .outputs = showfreqs_outputs,
558 .priv_class = &showfreqs_class,