2 * Copyright (c) 2012-2013 Clément Bœsch
3 * Copyright (c) 2013 Rudolf Polzer <divverent@xonotic.org>
5 * This file is part of FFmpeg.
7 * FFmpeg is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU Lesser General Public
9 * License as published by the Free Software Foundation; either
10 * version 2.1 of the License, or (at your option) any later version.
12 * FFmpeg is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * Lesser General Public License for more details.
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with FFmpeg; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
24 * audio to spectrum (video) transmedia filter, based on ffplay rdft showmode
25 * (by Michael Niedermayer) and lavfi/avf_showwaves (by Stefano Sabatini).
30 #include "libavcodec/avfft.h"
31 #include "libavutil/avassert.h"
32 #include "libavutil/channel_layout.h"
33 #include "libavutil/opt.h"
37 enum DisplayMode { COMBINED, SEPARATE, NB_MODES };
38 enum DisplayScale { LINEAR, SQRT, CBRT, LOG, NB_SCALES };
39 enum ColorMode { CHANNEL, INTENSITY, NB_CLMODES };
40 enum WindowFunc { WFUNC_NONE, WFUNC_HANN, WFUNC_HAMMING, WFUNC_BLACKMAN, NB_WFUNC };
41 enum SlideMode { REPLACE, SCROLL, FULLFRAME, NB_SLIDES };
47 int nb_display_channels;
49 int sliding; ///< 1 if sliding mode, 0 otherwise
50 int mode; ///< channel display mode
51 int color_mode; ///< display color scheme
53 float saturation; ///< color saturation multiplier
54 int xpos; ///< x position (current column)
55 RDFTContext *rdft; ///< Real Discrete Fourier Transform context
56 int rdft_bits; ///< number of bits (RDFT window size = 1<<rdft_bits)
57 FFTSample **rdft_data; ///< bins holder for each (displayed) channels
58 float *window_func_lut; ///< Window function LUT
60 float *combine_buffer; ///< color combining buffer (3 * h items)
61 } ShowSpectrumContext;
63 #define OFFSET(x) offsetof(ShowSpectrumContext, x)
64 #define FLAGS AV_OPT_FLAG_FILTERING_PARAM|AV_OPT_FLAG_VIDEO_PARAM
66 static const AVOption showspectrum_options[] = {
67 { "size", "set video size", OFFSET(w), AV_OPT_TYPE_IMAGE_SIZE, {.str = "640x512"}, 0, 0, FLAGS },
68 { "s", "set video size", OFFSET(w), AV_OPT_TYPE_IMAGE_SIZE, {.str = "640x512"}, 0, 0, FLAGS },
69 { "slide", "set sliding mode", OFFSET(sliding), AV_OPT_TYPE_INT, {.i64 = 0}, 0, NB_SLIDES, FLAGS, "slide" },
70 { "replace", "replace old columns with new", 0, AV_OPT_TYPE_CONST, {.i64=REPLACE}, 0, 0, FLAGS, "slide" },
71 { "scroll", "scroll from right to left", 0, AV_OPT_TYPE_CONST, {.i64=SCROLL}, 0, 0, FLAGS, "slide" },
72 { "fullframe", "return full frames", 0, AV_OPT_TYPE_CONST, {.i64=FULLFRAME}, 0, 0, FLAGS, "slide" },
73 { "mode", "set channel display mode", OFFSET(mode), AV_OPT_TYPE_INT, {.i64=COMBINED}, COMBINED, NB_MODES-1, FLAGS, "mode" },
74 { "combined", "combined mode", 0, AV_OPT_TYPE_CONST, {.i64=COMBINED}, 0, 0, FLAGS, "mode" },
75 { "separate", "separate mode", 0, AV_OPT_TYPE_CONST, {.i64=SEPARATE}, 0, 0, FLAGS, "mode" },
76 { "color", "set channel coloring", OFFSET(color_mode), AV_OPT_TYPE_INT, {.i64=CHANNEL}, CHANNEL, NB_CLMODES-1, FLAGS, "color" },
77 { "channel", "separate color for each channel", 0, AV_OPT_TYPE_CONST, {.i64=CHANNEL}, 0, 0, FLAGS, "color" },
78 { "intensity", "intensity based coloring", 0, AV_OPT_TYPE_CONST, {.i64=INTENSITY}, 0, 0, FLAGS, "color" },
79 { "scale", "set display scale", OFFSET(scale), AV_OPT_TYPE_INT, {.i64=SQRT}, LINEAR, NB_SCALES-1, FLAGS, "scale" },
80 { "sqrt", "square root", 0, AV_OPT_TYPE_CONST, {.i64=SQRT}, 0, 0, FLAGS, "scale" },
81 { "cbrt", "cubic root", 0, AV_OPT_TYPE_CONST, {.i64=CBRT}, 0, 0, FLAGS, "scale" },
82 { "log", "logarithmic", 0, AV_OPT_TYPE_CONST, {.i64=LOG}, 0, 0, FLAGS, "scale" },
83 { "lin", "linear", 0, AV_OPT_TYPE_CONST, {.i64=LINEAR}, 0, 0, FLAGS, "scale" },
84 { "saturation", "color saturation multiplier", OFFSET(saturation), AV_OPT_TYPE_FLOAT, {.dbl = 1}, -10, 10, FLAGS },
85 { "win_func", "set window function", OFFSET(win_func), AV_OPT_TYPE_INT, {.i64 = WFUNC_HANN}, 0, NB_WFUNC-1, FLAGS, "win_func" },
86 { "hann", "Hann window", 0, AV_OPT_TYPE_CONST, {.i64 = WFUNC_HANN}, 0, 0, FLAGS, "win_func" },
87 { "hamming", "Hamming window", 0, AV_OPT_TYPE_CONST, {.i64 = WFUNC_HAMMING}, 0, 0, FLAGS, "win_func" },
88 { "blackman", "Blackman window", 0, AV_OPT_TYPE_CONST, {.i64 = WFUNC_BLACKMAN}, 0, 0, FLAGS, "win_func" },
92 AVFILTER_DEFINE_CLASS(showspectrum);
96 } intensity_color_table[] = {
98 { 0.13, .03587126228984074, .1573300977624594, -.02548747583751842 },
99 { 0.30, .18572281794568020, .1772436246393981, .17475554840414750 },
100 { 0.60, .28184980583656130, -.1593064119945782, .47132074554608920 },
101 { 0.73, .65830621175547810, -.3716070802232764, .24352759331252930 },
102 { 0.78, .76318535758242900, -.4307467689263783, .16866496622310430 },
103 { 0.91, .95336363636363640, -.2045454545454546, .03313636363636363 },
107 static av_cold void uninit(AVFilterContext *ctx)
109 ShowSpectrumContext *s = ctx->priv;
112 av_freep(&s->combine_buffer);
113 av_rdft_end(s->rdft);
114 for (i = 0; i < s->nb_display_channels; i++)
115 av_freep(&s->rdft_data[i]);
116 av_freep(&s->rdft_data);
117 av_freep(&s->window_func_lut);
118 av_frame_free(&s->outpicref);
121 static int query_formats(AVFilterContext *ctx)
123 AVFilterFormats *formats = NULL;
124 AVFilterChannelLayouts *layouts = NULL;
125 AVFilterLink *inlink = ctx->inputs[0];
126 AVFilterLink *outlink = ctx->outputs[0];
127 static const enum AVSampleFormat sample_fmts[] = { AV_SAMPLE_FMT_S16P, AV_SAMPLE_FMT_NONE };
128 static const enum AVPixelFormat pix_fmts[] = { AV_PIX_FMT_YUVJ444P, AV_PIX_FMT_NONE };
130 /* set input audio formats */
131 formats = ff_make_format_list(sample_fmts);
133 return AVERROR(ENOMEM);
134 ff_formats_ref(formats, &inlink->out_formats);
136 layouts = ff_all_channel_layouts();
138 return AVERROR(ENOMEM);
139 ff_channel_layouts_ref(layouts, &inlink->out_channel_layouts);
141 formats = ff_all_samplerates();
143 return AVERROR(ENOMEM);
144 ff_formats_ref(formats, &inlink->out_samplerates);
146 /* set output video format */
147 formats = ff_make_format_list(pix_fmts);
149 return AVERROR(ENOMEM);
150 ff_formats_ref(formats, &outlink->in_formats);
155 static int config_output(AVFilterLink *outlink)
157 AVFilterContext *ctx = outlink->src;
158 AVFilterLink *inlink = ctx->inputs[0];
159 ShowSpectrumContext *s = ctx->priv;
160 int i, rdft_bits, win_size, h;
165 h = (s->mode == COMBINED) ? outlink->h : outlink->h / inlink->channels;
166 s->channel_height = h;
168 /* RDFT window size (precision) according to the requested output frame height */
169 for (rdft_bits = 1; 1 << rdft_bits < 2 * h; rdft_bits++);
170 win_size = 1 << rdft_bits;
172 /* (re-)configuration if the video output changed (or first init) */
173 if (rdft_bits != s->rdft_bits) {
176 av_rdft_end(s->rdft);
177 s->rdft = av_rdft_init(rdft_bits, DFT_R2C);
179 av_log(ctx, AV_LOG_ERROR, "Unable to create RDFT context. "
180 "The window size might be too high.\n");
181 return AVERROR(EINVAL);
183 s->rdft_bits = rdft_bits;
185 /* RDFT buffers: x2 for each (display) channel buffer.
186 * Note: we use free and malloc instead of a realloc-like function to
187 * make sure the buffer is aligned in memory for the FFT functions. */
188 for (i = 0; i < s->nb_display_channels; i++)
189 av_freep(&s->rdft_data[i]);
190 av_freep(&s->rdft_data);
191 s->nb_display_channels = inlink->channels;
193 s->rdft_data = av_calloc(s->nb_display_channels, sizeof(*s->rdft_data));
195 return AVERROR(ENOMEM);
196 for (i = 0; i < s->nb_display_channels; i++) {
197 s->rdft_data[i] = av_calloc(win_size, sizeof(**s->rdft_data));
198 if (!s->rdft_data[i])
199 return AVERROR(ENOMEM);
202 /* pre-calc windowing function */
204 av_realloc_f(s->window_func_lut, win_size,
205 sizeof(*s->window_func_lut));
206 if (!s->window_func_lut)
207 return AVERROR(ENOMEM);
208 switch (s->win_func) {
210 for (i = 0; i < win_size; i++)
211 s->window_func_lut[i] = 1.;
214 for (i = 0; i < win_size; i++)
215 s->window_func_lut[i] = .5f * (1 - cos(2*M_PI*i / (win_size-1)));
218 for (i = 0; i < win_size; i++)
219 s->window_func_lut[i] = .54f - .46f * cos(2*M_PI*i / (win_size-1));
221 case WFUNC_BLACKMAN: {
222 for (i = 0; i < win_size; i++)
223 s->window_func_lut[i] = .42f - .5f*cos(2*M_PI*i / (win_size-1)) + .08f*cos(4*M_PI*i / (win_size-1));
230 /* prepare the initial picref buffer (black frame) */
231 av_frame_free(&s->outpicref);
232 s->outpicref = outpicref =
233 ff_get_video_buffer(outlink, outlink->w, outlink->h);
235 return AVERROR(ENOMEM);
236 outlink->sample_aspect_ratio = (AVRational){1,1};
237 for (i = 0; i < outlink->h; i++) {
238 memset(outpicref->data[0] + i * outpicref->linesize[0], 0, outlink->w);
239 memset(outpicref->data[1] + i * outpicref->linesize[1], 128, outlink->w);
240 memset(outpicref->data[2] + i * outpicref->linesize[2], 128, outlink->w);
244 if (s->xpos >= outlink->w)
247 outlink->frame_rate = av_make_q(inlink->sample_rate, win_size);
248 if (s->sliding == FULLFRAME)
249 outlink->frame_rate.den *= outlink->w;
251 inlink->min_samples = inlink->max_samples = inlink->partial_buf_size =
255 av_realloc_f(s->combine_buffer, outlink->h * 3,
256 sizeof(*s->combine_buffer));
258 av_log(ctx, AV_LOG_VERBOSE, "s:%dx%d RDFT window size:%d\n",
259 s->w, s->h, win_size);
263 static int request_frame(AVFilterLink *outlink)
265 ShowSpectrumContext *s = outlink->src->priv;
266 AVFilterLink *inlink = outlink->src->inputs[0];
270 ret = ff_request_frame(inlink);
271 if (ret == AVERROR_EOF && s->sliding == FULLFRAME && s->xpos > 0 &&
273 for (i = 0; i < outlink->h; i++) {
274 memset(s->outpicref->data[0] + i * s->outpicref->linesize[0] + s->xpos, 0, outlink->w - s->xpos);
275 memset(s->outpicref->data[1] + i * s->outpicref->linesize[1] + s->xpos, 128, outlink->w - s->xpos);
276 memset(s->outpicref->data[2] + i * s->outpicref->linesize[2] + s->xpos, 128, outlink->w - s->xpos);
278 ret = ff_filter_frame(outlink, s->outpicref);
285 static int plot_spectrum_column(AVFilterLink *inlink, AVFrame *insamples)
288 AVFilterContext *ctx = inlink->dst;
289 AVFilterLink *outlink = ctx->outputs[0];
290 ShowSpectrumContext *s = ctx->priv;
291 AVFrame *outpicref = s->outpicref;
293 /* nb_freq contains the power of two superior or equal to the output image
294 * height (or half the RDFT window size) */
295 const int nb_freq = 1 << (s->rdft_bits - 1);
296 const int win_size = nb_freq << 1;
297 const double w = 1. / (sqrt(nb_freq) * 32768.);
298 int h = s->channel_height;
302 av_assert0(insamples->nb_samples == win_size);
304 /* fill RDFT input with the number of samples available */
305 for (ch = 0; ch < s->nb_display_channels; ch++) {
306 const int16_t *p = (int16_t *)insamples->extended_data[ch];
308 for (n = 0; n < win_size; n++)
309 s->rdft_data[ch][n] = p[n] * s->window_func_lut[n];
312 /* run RDFT on each samples set */
313 for (ch = 0; ch < s->nb_display_channels; ch++)
314 av_rdft_calc(s->rdft, s->rdft_data[ch]);
316 /* fill a new spectrum column */
317 #define RE(y, ch) s->rdft_data[ch][2 * (y) + 0]
318 #define IM(y, ch) s->rdft_data[ch][2 * (y) + 1]
319 #define MAGNITUDE(y, ch) hypot(RE(y, ch), IM(y, ch))
321 /* initialize buffer for combining to black */
322 for (y = 0; y < outlink->h; y++) {
323 s->combine_buffer[3 * y ] = 0;
324 s->combine_buffer[3 * y + 1] = 127.5;
325 s->combine_buffer[3 * y + 2] = 127.5;
328 for (ch = 0; ch < s->nb_display_channels; ch++) {
331 /* decide color range */
334 // reduce range by channel count
335 yf = 256.0f / s->nb_display_channels;
336 switch (s->color_mode) {
342 /* adjust saturation for mixed UV coloring */
343 /* this factor is correct for infinite channels, an approximation otherwise */
361 if (s->color_mode == CHANNEL) {
362 if (s->nb_display_channels > 1) {
363 uf *= 0.5 * sin((2 * M_PI * ch) / s->nb_display_channels);
364 vf *= 0.5 * cos((2 * M_PI * ch) / s->nb_display_channels);
373 /* draw the channel */
374 for (y = 0; y < h; y++) {
375 int row = (s->mode == COMBINED) ? y : ch * h + y;
376 float *out = &s->combine_buffer[3 * row];
379 float a = w * MAGNITUDE(y, ch);
392 a = 1 - log(FFMAX(FFMIN(1, a), 1e-6)) / log(1e-6); // zero = -120dBFS
398 if (s->color_mode == INTENSITY) {
402 for (i = 1; i < sizeof(intensity_color_table) / sizeof(*intensity_color_table) - 1; i++)
403 if (intensity_color_table[i].a >= a)
405 // i now is the first item >= the color
406 // now we know to interpolate between item i - 1 and i
407 if (a <= intensity_color_table[i - 1].a) {
408 y = intensity_color_table[i - 1].y;
409 u = intensity_color_table[i - 1].u;
410 v = intensity_color_table[i - 1].v;
411 } else if (a >= intensity_color_table[i].a) {
412 y = intensity_color_table[i].y;
413 u = intensity_color_table[i].u;
414 v = intensity_color_table[i].v;
416 float start = intensity_color_table[i - 1].a;
417 float end = intensity_color_table[i].a;
418 float lerpfrac = (a - start) / (end - start);
419 y = intensity_color_table[i - 1].y * (1.0f - lerpfrac)
420 + intensity_color_table[i].y * lerpfrac;
421 u = intensity_color_table[i - 1].u * (1.0f - lerpfrac)
422 + intensity_color_table[i].u * lerpfrac;
423 v = intensity_color_table[i - 1].v * (1.0f - lerpfrac)
424 + intensity_color_table[i].v * lerpfrac;
439 if (s->sliding == SCROLL) {
440 for (plane = 0; plane < 3; plane++) {
441 for (y = 0; y < outlink->h; y++) {
442 uint8_t *p = outpicref->data[plane] +
443 y * outpicref->linesize[plane];
444 memmove(p, p + 1, outlink->w - 1);
447 s->xpos = outlink->w - 1;
449 for (plane = 0; plane < 3; plane++) {
450 uint8_t *p = outpicref->data[plane] +
451 (outlink->h - 1) * outpicref->linesize[plane] +
453 for (y = 0; y < outlink->h; y++) {
454 *p = rint(FFMAX(0, FFMIN(s->combine_buffer[3 * y + plane], 255)));
455 p -= outpicref->linesize[plane];
459 if (s->sliding != FULLFRAME || s->xpos == 0)
460 outpicref->pts = insamples->pts;
463 if (s->xpos >= outlink->w)
465 if (s->sliding != FULLFRAME || s->xpos == 0) {
466 ret = ff_filter_frame(outlink, av_frame_clone(s->outpicref));
474 static int filter_frame(AVFilterLink *inlink, AVFrame *insamples)
476 AVFilterContext *ctx = inlink->dst;
477 ShowSpectrumContext *s = ctx->priv;
478 unsigned win_size = 1 << s->rdft_bits;
481 av_assert0(insamples->nb_samples <= win_size);
482 if (insamples->nb_samples == win_size)
483 ret = plot_spectrum_column(inlink, insamples);
485 av_frame_free(&insamples);
489 static const AVFilterPad showspectrum_inputs[] = {
492 .type = AVMEDIA_TYPE_AUDIO,
493 .filter_frame = filter_frame,
498 static const AVFilterPad showspectrum_outputs[] = {
501 .type = AVMEDIA_TYPE_VIDEO,
502 .config_props = config_output,
503 .request_frame = request_frame,
508 AVFilter ff_avf_showspectrum = {
509 .name = "showspectrum",
510 .description = NULL_IF_CONFIG_SMALL("Convert input audio to a spectrum video output."),
512 .query_formats = query_formats,
513 .priv_size = sizeof(ShowSpectrumContext),
514 .inputs = showspectrum_inputs,
515 .outputs = showspectrum_outputs,
516 .priv_class = &showspectrum_class,