summaryrefslogtreecommitdiff
path: root/libavfilter/avf_showspectrum.c (plain)
blob: 09b5a2a51fa462f446d361f86d2374bd1ac71c77
1/*
2 * Copyright (c) 2012-2013 Clément Bœsch
3 * Copyright (c) 2013 Rudolf Polzer <divverent@xonotic.org>
4 * Copyright (c) 2015 Paul B Mahol
5 *
6 * This file is part of FFmpeg.
7 *
8 * FFmpeg is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU Lesser General Public
10 * License as published by the Free Software Foundation; either
11 * version 2.1 of the License, or (at your option) any later version.
12 *
13 * FFmpeg is distributed in the hope that it will be useful,
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16 * Lesser General Public License for more details.
17 *
18 * You should have received a copy of the GNU Lesser General Public
19 * License along with FFmpeg; if not, write to the Free Software
20 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
21 */
22
23/**
24 * @file
25 * audio to spectrum (video) transmedia filter, based on ffplay rdft showmode
26 * (by Michael Niedermayer) and lavfi/avf_showwaves (by Stefano Sabatini).
27 */
28
29#include <math.h>
30
31#include "libavcodec/avfft.h"
32#include "libavutil/audio_fifo.h"
33#include "libavutil/avassert.h"
34#include "libavutil/avstring.h"
35#include "libavutil/channel_layout.h"
36#include "libavutil/opt.h"
37#include "libavutil/xga_font_data.h"
38#include "audio.h"
39#include "video.h"
40#include "avfilter.h"
41#include "internal.h"
42#include "window_func.h"
43
44enum DisplayMode { COMBINED, SEPARATE, NB_MODES };
45enum DataMode { D_MAGNITUDE, D_PHASE, NB_DMODES };
46enum DisplayScale { LINEAR, SQRT, CBRT, LOG, FOURTHRT, FIFTHRT, NB_SCALES };
47enum ColorMode { CHANNEL, INTENSITY, RAINBOW, MORELAND, NEBULAE, FIRE, FIERY, FRUIT, COOL, NB_CLMODES };
48enum SlideMode { REPLACE, SCROLL, FULLFRAME, RSCROLL, NB_SLIDES };
49enum Orientation { VERTICAL, HORIZONTAL, NB_ORIENTATIONS };
50
51typedef struct {
52 const AVClass *class;
53 int w, h;
54 AVFrame *outpicref;
55 int nb_display_channels;
56 int orientation;
57 int channel_width;
58 int channel_height;
59 int sliding; ///< 1 if sliding mode, 0 otherwise
60 int mode; ///< channel display mode
61 int color_mode; ///< display color scheme
62 int scale;
63 float saturation; ///< color saturation multiplier
64 float rotation; ///< color rotation
65 int data;
66 int xpos; ///< x position (current column)
67 FFTContext **fft; ///< Fast Fourier Transform context
68 int fft_bits; ///< number of bits (FFT window size = 1<<fft_bits)
69 FFTComplex **fft_data; ///< bins holder for each (displayed) channels
70 float *window_func_lut; ///< Window function LUT
71 float **magnitudes;
72 float **phases;
73 int win_func;
74 int win_size;
75 double win_scale;
76 float overlap;
77 float gain;
78 int hop_size;
79 float *combine_buffer; ///< color combining buffer (3 * h items)
80 float **color_buffer; ///< color buffer (3 * h * ch items)
81 AVAudioFifo *fifo;
82 int64_t pts;
83 int single_pic;
84 int legend;
85 int start_x, start_y;
86} ShowSpectrumContext;
87
88#define OFFSET(x) offsetof(ShowSpectrumContext, x)
89#define FLAGS AV_OPT_FLAG_FILTERING_PARAM|AV_OPT_FLAG_VIDEO_PARAM
90
91static const AVOption showspectrum_options[] = {
92 { "size", "set video size", OFFSET(w), AV_OPT_TYPE_IMAGE_SIZE, {.str = "640x512"}, 0, 0, FLAGS },
93 { "s", "set video size", OFFSET(w), AV_OPT_TYPE_IMAGE_SIZE, {.str = "640x512"}, 0, 0, FLAGS },
94 { "slide", "set sliding mode", OFFSET(sliding), AV_OPT_TYPE_INT, {.i64 = 0}, 0, NB_SLIDES-1, FLAGS, "slide" },
95 { "replace", "replace old columns with new", 0, AV_OPT_TYPE_CONST, {.i64=REPLACE}, 0, 0, FLAGS, "slide" },
96 { "scroll", "scroll from right to left", 0, AV_OPT_TYPE_CONST, {.i64=SCROLL}, 0, 0, FLAGS, "slide" },
97 { "fullframe", "return full frames", 0, AV_OPT_TYPE_CONST, {.i64=FULLFRAME}, 0, 0, FLAGS, "slide" },
98 { "rscroll", "scroll from left to right", 0, AV_OPT_TYPE_CONST, {.i64=RSCROLL}, 0, 0, FLAGS, "slide" },
99 { "mode", "set channel display mode", OFFSET(mode), AV_OPT_TYPE_INT, {.i64=COMBINED}, COMBINED, NB_MODES-1, FLAGS, "mode" },
100 { "combined", "combined mode", 0, AV_OPT_TYPE_CONST, {.i64=COMBINED}, 0, 0, FLAGS, "mode" },
101 { "separate", "separate mode", 0, AV_OPT_TYPE_CONST, {.i64=SEPARATE}, 0, 0, FLAGS, "mode" },
102 { "color", "set channel coloring", OFFSET(color_mode), AV_OPT_TYPE_INT, {.i64=CHANNEL}, CHANNEL, NB_CLMODES-1, FLAGS, "color" },
103 { "channel", "separate color for each channel", 0, AV_OPT_TYPE_CONST, {.i64=CHANNEL}, 0, 0, FLAGS, "color" },
104 { "intensity", "intensity based coloring", 0, AV_OPT_TYPE_CONST, {.i64=INTENSITY}, 0, 0, FLAGS, "color" },
105 { "rainbow", "rainbow based coloring", 0, AV_OPT_TYPE_CONST, {.i64=RAINBOW}, 0, 0, FLAGS, "color" },
106 { "moreland", "moreland based coloring", 0, AV_OPT_TYPE_CONST, {.i64=MORELAND}, 0, 0, FLAGS, "color" },
107 { "nebulae", "nebulae based coloring", 0, AV_OPT_TYPE_CONST, {.i64=NEBULAE}, 0, 0, FLAGS, "color" },
108 { "fire", "fire based coloring", 0, AV_OPT_TYPE_CONST, {.i64=FIRE}, 0, 0, FLAGS, "color" },
109 { "fiery", "fiery based coloring", 0, AV_OPT_TYPE_CONST, {.i64=FIERY}, 0, 0, FLAGS, "color" },
110 { "fruit", "fruit based coloring", 0, AV_OPT_TYPE_CONST, {.i64=FRUIT}, 0, 0, FLAGS, "color" },
111 { "cool", "cool based coloring", 0, AV_OPT_TYPE_CONST, {.i64=COOL}, 0, 0, FLAGS, "color" },
112 { "scale", "set display scale", OFFSET(scale), AV_OPT_TYPE_INT, {.i64=SQRT}, LINEAR, NB_SCALES-1, FLAGS, "scale" },
113 { "lin", "linear", 0, AV_OPT_TYPE_CONST, {.i64=LINEAR}, 0, 0, FLAGS, "scale" },
114 { "sqrt", "square root", 0, AV_OPT_TYPE_CONST, {.i64=SQRT}, 0, 0, FLAGS, "scale" },
115 { "cbrt", "cubic root", 0, AV_OPT_TYPE_CONST, {.i64=CBRT}, 0, 0, FLAGS, "scale" },
116 { "log", "logarithmic", 0, AV_OPT_TYPE_CONST, {.i64=LOG}, 0, 0, FLAGS, "scale" },
117 { "4thrt","4th root", 0, AV_OPT_TYPE_CONST, {.i64=FOURTHRT}, 0, 0, FLAGS, "scale" },
118 { "5thrt","5th root", 0, AV_OPT_TYPE_CONST, {.i64=FIFTHRT}, 0, 0, FLAGS, "scale" },
119 { "saturation", "color saturation multiplier", OFFSET(saturation), AV_OPT_TYPE_FLOAT, {.dbl = 1}, -10, 10, FLAGS },
120 { "win_func", "set window function", OFFSET(win_func), AV_OPT_TYPE_INT, {.i64 = WFUNC_HANNING}, 0, NB_WFUNC-1, FLAGS, "win_func" },
121 { "rect", "Rectangular", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_RECT}, 0, 0, FLAGS, "win_func" },
122 { "bartlett", "Bartlett", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_BARTLETT}, 0, 0, FLAGS, "win_func" },
123 { "hann", "Hann", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_HANNING}, 0, 0, FLAGS, "win_func" },
124 { "hanning", "Hanning", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_HANNING}, 0, 0, FLAGS, "win_func" },
125 { "hamming", "Hamming", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_HAMMING}, 0, 0, FLAGS, "win_func" },
126 { "blackman", "Blackman", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_BLACKMAN}, 0, 0, FLAGS, "win_func" },
127 { "welch", "Welch", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_WELCH}, 0, 0, FLAGS, "win_func" },
128 { "flattop", "Flat-top", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_FLATTOP}, 0, 0, FLAGS, "win_func" },
129 { "bharris", "Blackman-Harris", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_BHARRIS}, 0, 0, FLAGS, "win_func" },
130 { "bnuttall", "Blackman-Nuttall", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_BNUTTALL}, 0, 0, FLAGS, "win_func" },
131 { "bhann", "Bartlett-Hann", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_BHANN}, 0, 0, FLAGS, "win_func" },
132 { "sine", "Sine", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_SINE}, 0, 0, FLAGS, "win_func" },
133 { "nuttall", "Nuttall", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_NUTTALL}, 0, 0, FLAGS, "win_func" },
134 { "lanczos", "Lanczos", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_LANCZOS}, 0, 0, FLAGS, "win_func" },
135 { "gauss", "Gauss", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_GAUSS}, 0, 0, FLAGS, "win_func" },
136 { "tukey", "Tukey", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_TUKEY}, 0, 0, FLAGS, "win_func" },
137 { "dolph", "Dolph-Chebyshev", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_DOLPH}, 0, 0, FLAGS, "win_func" },
138 { "cauchy", "Cauchy", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_CAUCHY}, 0, 0, FLAGS, "win_func" },
139 { "parzen", "Parzen", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_PARZEN}, 0, 0, FLAGS, "win_func" },
140 { "poisson", "Poisson", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_POISSON}, 0, 0, FLAGS, "win_func" },
141 { "orientation", "set orientation", OFFSET(orientation), AV_OPT_TYPE_INT, {.i64=VERTICAL}, 0, NB_ORIENTATIONS-1, FLAGS, "orientation" },
142 { "vertical", NULL, 0, AV_OPT_TYPE_CONST, {.i64=VERTICAL}, 0, 0, FLAGS, "orientation" },
143 { "horizontal", NULL, 0, AV_OPT_TYPE_CONST, {.i64=HORIZONTAL}, 0, 0, FLAGS, "orientation" },
144 { "overlap", "set window overlap", OFFSET(overlap), AV_OPT_TYPE_FLOAT, {.dbl = 0}, 0, 1, FLAGS },
145 { "gain", "set scale gain", OFFSET(gain), AV_OPT_TYPE_FLOAT, {.dbl = 1}, 0, 128, FLAGS },
146 { "data", "set data mode", OFFSET(data), AV_OPT_TYPE_INT, {.i64 = 0}, 0, NB_DMODES-1, FLAGS, "data" },
147 { "magnitude", NULL, 0, AV_OPT_TYPE_CONST, {.i64=D_MAGNITUDE}, 0, 0, FLAGS, "data" },
148 { "phase", NULL, 0, AV_OPT_TYPE_CONST, {.i64=D_PHASE}, 0, 0, FLAGS, "data" },
149 { "rotation", "color rotation", OFFSET(rotation), AV_OPT_TYPE_FLOAT, {.dbl = 0}, -1, 1, FLAGS },
150 { NULL }
151};
152
153AVFILTER_DEFINE_CLASS(showspectrum);
154
155static const struct ColorTable {
156 float a, y, u, v;
157} color_table[][8] = {
158 [INTENSITY] = {
159 { 0, 0, 0, 0 },
160 { 0.13, .03587126228984074, .1573300977624594, -.02548747583751842 },
161 { 0.30, .18572281794568020, .1772436246393981, .17475554840414750 },
162 { 0.60, .28184980583656130, -.1593064119945782, .47132074554608920 },
163 { 0.73, .65830621175547810, -.3716070802232764, .24352759331252930 },
164 { 0.78, .76318535758242900, -.4307467689263783, .16866496622310430 },
165 { 0.91, .95336363636363640, -.2045454545454546, .03313636363636363 },
166 { 1, 1, 0, 0 }},
167 [RAINBOW] = {
168 { 0, 0, 0, 0 },
169 { 0.13, 44/256., (189-128)/256., (138-128)/256. },
170 { 0.25, 29/256., (186-128)/256., (119-128)/256. },
171 { 0.38, 119/256., (194-128)/256., (53-128)/256. },
172 { 0.60, 111/256., (73-128)/256., (59-128)/256. },
173 { 0.73, 205/256., (19-128)/256., (149-128)/256. },
174 { 0.86, 135/256., (83-128)/256., (200-128)/256. },
175 { 1, 73/256., (95-128)/256., (225-128)/256. }},
176 [MORELAND] = {
177 { 0, 44/256., (181-128)/256., (112-128)/256. },
178 { 0.13, 126/256., (177-128)/256., (106-128)/256. },
179 { 0.25, 164/256., (163-128)/256., (109-128)/256. },
180 { 0.38, 200/256., (140-128)/256., (120-128)/256. },
181 { 0.60, 201/256., (117-128)/256., (141-128)/256. },
182 { 0.73, 177/256., (103-128)/256., (165-128)/256. },
183 { 0.86, 136/256., (100-128)/256., (183-128)/256. },
184 { 1, 68/256., (117-128)/256., (203-128)/256. }},
185 [NEBULAE] = {
186 { 0, 10/256., (134-128)/256., (132-128)/256. },
187 { 0.23, 21/256., (137-128)/256., (130-128)/256. },
188 { 0.45, 35/256., (134-128)/256., (134-128)/256. },
189 { 0.57, 51/256., (130-128)/256., (139-128)/256. },
190 { 0.67, 104/256., (116-128)/256., (162-128)/256. },
191 { 0.77, 120/256., (105-128)/256., (188-128)/256. },
192 { 0.87, 140/256., (105-128)/256., (188-128)/256. },
193 { 1, 1, 0, 0 }},
194 [FIRE] = {
195 { 0, 0, 0, 0 },
196 { 0.23, 44/256., (132-128)/256., (127-128)/256. },
197 { 0.45, 62/256., (116-128)/256., (140-128)/256. },
198 { 0.57, 75/256., (105-128)/256., (152-128)/256. },
199 { 0.67, 95/256., (91-128)/256., (166-128)/256. },
200 { 0.77, 126/256., (74-128)/256., (172-128)/256. },
201 { 0.87, 164/256., (73-128)/256., (162-128)/256. },
202 { 1, 1, 0, 0 }},
203 [FIERY] = {
204 { 0, 0, 0, 0 },
205 { 0.23, 36/256., (116-128)/256., (163-128)/256. },
206 { 0.45, 52/256., (102-128)/256., (200-128)/256. },
207 { 0.57, 116/256., (84-128)/256., (196-128)/256. },
208 { 0.67, 157/256., (67-128)/256., (181-128)/256. },
209 { 0.77, 193/256., (40-128)/256., (155-128)/256. },
210 { 0.87, 221/256., (101-128)/256., (134-128)/256. },
211 { 1, 1, 0, 0 }},
212 [FRUIT] = {
213 { 0, 0, 0, 0 },
214 { 0.20, 29/256., (136-128)/256., (119-128)/256. },
215 { 0.30, 60/256., (119-128)/256., (90-128)/256. },
216 { 0.40, 85/256., (91-128)/256., (85-128)/256. },
217 { 0.50, 116/256., (70-128)/256., (105-128)/256. },
218 { 0.60, 151/256., (50-128)/256., (146-128)/256. },
219 { 0.70, 191/256., (63-128)/256., (178-128)/256. },
220 { 1, 98/256., (80-128)/256., (221-128)/256. }},
221 [COOL] = {
222 { 0, 0, 0, 0 },
223 { .15, 0, .5, -.5 },
224 { 1, 1, -.5, .5 }},
225};
226
227static av_cold void uninit(AVFilterContext *ctx)
228{
229 ShowSpectrumContext *s = ctx->priv;
230 int i;
231
232 av_freep(&s->combine_buffer);
233 if (s->fft) {
234 for (i = 0; i < s->nb_display_channels; i++)
235 av_fft_end(s->fft[i]);
236 }
237 av_freep(&s->fft);
238 if (s->fft_data) {
239 for (i = 0; i < s->nb_display_channels; i++)
240 av_freep(&s->fft_data[i]);
241 }
242 av_freep(&s->fft_data);
243 if (s->color_buffer) {
244 for (i = 0; i < s->nb_display_channels; i++)
245 av_freep(&s->color_buffer[i]);
246 }
247 av_freep(&s->color_buffer);
248 av_freep(&s->window_func_lut);
249 if (s->magnitudes) {
250 for (i = 0; i < s->nb_display_channels; i++)
251 av_freep(&s->magnitudes[i]);
252 }
253 av_freep(&s->magnitudes);
254 av_frame_free(&s->outpicref);
255 av_audio_fifo_free(s->fifo);
256 if (s->phases) {
257 for (i = 0; i < s->nb_display_channels; i++)
258 av_freep(&s->phases[i]);
259 }
260 av_freep(&s->phases);
261}
262
263static int query_formats(AVFilterContext *ctx)
264{
265 AVFilterFormats *formats = NULL;
266 AVFilterChannelLayouts *layouts = NULL;
267 AVFilterLink *inlink = ctx->inputs[0];
268 AVFilterLink *outlink = ctx->outputs[0];
269 static const enum AVSampleFormat sample_fmts[] = { AV_SAMPLE_FMT_FLTP, AV_SAMPLE_FMT_NONE };
270 static const enum AVPixelFormat pix_fmts[] = { AV_PIX_FMT_YUV444P, AV_PIX_FMT_YUVJ444P, AV_PIX_FMT_NONE };
271 int ret;
272
273 /* set input audio formats */
274 formats = ff_make_format_list(sample_fmts);
275 if ((ret = ff_formats_ref(formats, &inlink->out_formats)) < 0)
276 return ret;
277
278 layouts = ff_all_channel_layouts();
279 if ((ret = ff_channel_layouts_ref(layouts, &inlink->out_channel_layouts)) < 0)
280 return ret;
281
282 formats = ff_all_samplerates();
283 if ((ret = ff_formats_ref(formats, &inlink->out_samplerates)) < 0)
284 return ret;
285
286 /* set output video format */
287 formats = ff_make_format_list(pix_fmts);
288 if ((ret = ff_formats_ref(formats, &outlink->in_formats)) < 0)
289 return ret;
290
291 return 0;
292}
293
294static int config_output(AVFilterLink *outlink)
295{
296 AVFilterContext *ctx = outlink->src;
297 AVFilterLink *inlink = ctx->inputs[0];
298 ShowSpectrumContext *s = ctx->priv;
299 int i, fft_bits, h, w;
300 float overlap;
301
302 if (!strcmp(ctx->filter->name, "showspectrumpic"))
303 s->single_pic = 1;
304
305 outlink->w = s->w;
306 outlink->h = s->h;
307
308 if (s->legend) {
309 s->start_x = log10(inlink->sample_rate) * 25;
310 s->start_y = 64;
311 outlink->w += s->start_x * 2;
312 outlink->h += s->start_y * 2;
313 }
314
315 h = (s->mode == COMBINED || s->orientation == HORIZONTAL) ? s->h : s->h / inlink->channels;
316 w = (s->mode == COMBINED || s->orientation == VERTICAL) ? s->w : s->w / inlink->channels;
317 s->channel_height = h;
318 s->channel_width = w;
319
320 if (s->orientation == VERTICAL) {
321 /* FFT window size (precision) according to the requested output frame height */
322 for (fft_bits = 1; 1 << fft_bits < 2 * h; fft_bits++);
323 } else {
324 /* FFT window size (precision) according to the requested output frame width */
325 for (fft_bits = 1; 1 << fft_bits < 2 * w; fft_bits++);
326 }
327 s->win_size = 1 << fft_bits;
328
329 if (!s->fft) {
330 s->fft = av_calloc(inlink->channels, sizeof(*s->fft));
331 if (!s->fft)
332 return AVERROR(ENOMEM);
333 }
334
335 /* (re-)configuration if the video output changed (or first init) */
336 if (fft_bits != s->fft_bits) {
337 AVFrame *outpicref;
338
339 s->fft_bits = fft_bits;
340
341 /* FFT buffers: x2 for each (display) channel buffer.
342 * Note: we use free and malloc instead of a realloc-like function to
343 * make sure the buffer is aligned in memory for the FFT functions. */
344 for (i = 0; i < s->nb_display_channels; i++) {
345 av_fft_end(s->fft[i]);
346 av_freep(&s->fft_data[i]);
347 }
348 av_freep(&s->fft_data);
349
350 s->nb_display_channels = inlink->channels;
351 for (i = 0; i < s->nb_display_channels; i++) {
352 s->fft[i] = av_fft_init(fft_bits, 0);
353 if (!s->fft[i]) {
354 av_log(ctx, AV_LOG_ERROR, "Unable to create FFT context. "
355 "The window size might be too high.\n");
356 return AVERROR(EINVAL);
357 }
358 }
359
360 s->magnitudes = av_calloc(s->nb_display_channels, sizeof(*s->magnitudes));
361 if (!s->magnitudes)
362 return AVERROR(ENOMEM);
363 for (i = 0; i < s->nb_display_channels; i++) {
364 s->magnitudes[i] = av_calloc(s->orientation == VERTICAL ? s->h : s->w, sizeof(**s->magnitudes));
365 if (!s->magnitudes[i])
366 return AVERROR(ENOMEM);
367 }
368
369 s->phases = av_calloc(s->nb_display_channels, sizeof(*s->phases));
370 if (!s->phases)
371 return AVERROR(ENOMEM);
372 for (i = 0; i < s->nb_display_channels; i++) {
373 s->phases[i] = av_calloc(s->orientation == VERTICAL ? s->h : s->w, sizeof(**s->phases));
374 if (!s->phases[i])
375 return AVERROR(ENOMEM);
376 }
377
378 av_freep(&s->color_buffer);
379 s->color_buffer = av_calloc(s->nb_display_channels, sizeof(*s->color_buffer));
380 if (!s->color_buffer)
381 return AVERROR(ENOMEM);
382 for (i = 0; i < s->nb_display_channels; i++) {
383 s->color_buffer[i] = av_calloc(s->orientation == VERTICAL ? s->h * 3 : s->w * 3, sizeof(**s->color_buffer));
384 if (!s->color_buffer[i])
385 return AVERROR(ENOMEM);
386 }
387
388 s->fft_data = av_calloc(s->nb_display_channels, sizeof(*s->fft_data));
389 if (!s->fft_data)
390 return AVERROR(ENOMEM);
391 for (i = 0; i < s->nb_display_channels; i++) {
392 s->fft_data[i] = av_calloc(s->win_size, sizeof(**s->fft_data));
393 if (!s->fft_data[i])
394 return AVERROR(ENOMEM);
395 }
396
397 /* pre-calc windowing function */
398 s->window_func_lut =
399 av_realloc_f(s->window_func_lut, s->win_size,
400 sizeof(*s->window_func_lut));
401 if (!s->window_func_lut)
402 return AVERROR(ENOMEM);
403 ff_generate_window_func(s->window_func_lut, s->win_size, s->win_func, &overlap);
404 if (s->overlap == 1)
405 s->overlap = overlap;
406 s->hop_size = (1. - s->overlap) * s->win_size;
407 if (s->hop_size < 1) {
408 av_log(ctx, AV_LOG_ERROR, "overlap %f too big\n", s->overlap);
409 return AVERROR(EINVAL);
410 }
411
412 for (s->win_scale = 0, i = 0; i < s->win_size; i++) {
413 s->win_scale += s->window_func_lut[i] * s->window_func_lut[i];
414 }
415 s->win_scale = 1. / sqrt(s->win_scale);
416
417 /* prepare the initial picref buffer (black frame) */
418 av_frame_free(&s->outpicref);
419 s->outpicref = outpicref =
420 ff_get_video_buffer(outlink, outlink->w, outlink->h);
421 if (!outpicref)
422 return AVERROR(ENOMEM);
423 outlink->sample_aspect_ratio = (AVRational){1,1};
424 for (i = 0; i < outlink->h; i++) {
425 memset(outpicref->data[0] + i * outpicref->linesize[0], 0, outlink->w);
426 memset(outpicref->data[1] + i * outpicref->linesize[1], 128, outlink->w);
427 memset(outpicref->data[2] + i * outpicref->linesize[2], 128, outlink->w);
428 }
429 av_frame_set_color_range(outpicref, AVCOL_RANGE_JPEG);
430 }
431
432 if ((s->orientation == VERTICAL && s->xpos >= s->w) ||
433 (s->orientation == HORIZONTAL && s->xpos >= s->h))
434 s->xpos = 0;
435
436 outlink->frame_rate = av_make_q(inlink->sample_rate, s->win_size * (1.-s->overlap));
437 if (s->orientation == VERTICAL && s->sliding == FULLFRAME)
438 outlink->frame_rate.den *= s->w;
439 if (s->orientation == HORIZONTAL && s->sliding == FULLFRAME)
440 outlink->frame_rate.den *= s->h;
441
442 if (s->orientation == VERTICAL) {
443 s->combine_buffer =
444 av_realloc_f(s->combine_buffer, s->h * 3,
445 sizeof(*s->combine_buffer));
446 } else {
447 s->combine_buffer =
448 av_realloc_f(s->combine_buffer, s->w * 3,
449 sizeof(*s->combine_buffer));
450 }
451
452 av_log(ctx, AV_LOG_VERBOSE, "s:%dx%d FFT window size:%d\n",
453 s->w, s->h, s->win_size);
454
455 av_audio_fifo_free(s->fifo);
456 s->fifo = av_audio_fifo_alloc(inlink->format, inlink->channels, s->win_size);
457 if (!s->fifo)
458 return AVERROR(ENOMEM);
459 return 0;
460}
461
462static int run_channel_fft(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)
463{
464 ShowSpectrumContext *s = ctx->priv;
465 const float *window_func_lut = s->window_func_lut;
466 AVFrame *fin = arg;
467 const int ch = jobnr;
468 int n;
469
470 /* fill FFT input with the number of samples available */
471 const float *p = (float *)fin->extended_data[ch];
472
473 for (n = 0; n < s->win_size; n++) {
474 s->fft_data[ch][n].re = p[n] * window_func_lut[n];
475 s->fft_data[ch][n].im = 0;
476 }
477
478 /* run FFT on each samples set */
479 av_fft_permute(s->fft[ch], s->fft_data[ch]);
480 av_fft_calc(s->fft[ch], s->fft_data[ch]);
481
482 return 0;
483}
484
485#define RE(y, ch) s->fft_data[ch][y].re
486#define IM(y, ch) s->fft_data[ch][y].im
487#define MAGNITUDE(y, ch) hypot(RE(y, ch), IM(y, ch))
488#define PHASE(y, ch) atan2(IM(y, ch), RE(y, ch))
489
490static int calc_channel_magnitudes(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)
491{
492 ShowSpectrumContext *s = ctx->priv;
493 const double w = s->win_scale * (s->scale == LOG ? s->win_scale : 1);
494 int y, h = s->orientation == VERTICAL ? s->h : s->w;
495 const float f = s->gain * w;
496 const int ch = jobnr;
497 float *magnitudes = s->magnitudes[ch];
498
499 for (y = 0; y < h; y++)
500 magnitudes[y] = MAGNITUDE(y, ch) * f;
501
502 return 0;
503}
504
505static int calc_channel_phases(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)
506{
507 ShowSpectrumContext *s = ctx->priv;
508 const int h = s->orientation == VERTICAL ? s->h : s->w;
509 const int ch = jobnr;
510 float *phases = s->phases[ch];
511 int y;
512
513 for (y = 0; y < h; y++)
514 phases[y] = (PHASE(y, ch) / M_PI + 1) / 2;
515
516 return 0;
517}
518
519static void acalc_magnitudes(ShowSpectrumContext *s)
520{
521 const double w = s->win_scale * (s->scale == LOG ? s->win_scale : 1);
522 int ch, y, h = s->orientation == VERTICAL ? s->h : s->w;
523 const float f = s->gain * w;
524
525 for (ch = 0; ch < s->nb_display_channels; ch++) {
526 float *magnitudes = s->magnitudes[ch];
527
528 for (y = 0; y < h; y++)
529 magnitudes[y] += MAGNITUDE(y, ch) * f;
530 }
531}
532
533static void scale_magnitudes(ShowSpectrumContext *s, float scale)
534{
535 int ch, y, h = s->orientation == VERTICAL ? s->h : s->w;
536
537 for (ch = 0; ch < s->nb_display_channels; ch++) {
538 float *magnitudes = s->magnitudes[ch];
539
540 for (y = 0; y < h; y++)
541 magnitudes[y] *= scale;
542 }
543}
544
545static void color_range(ShowSpectrumContext *s, int ch,
546 float *yf, float *uf, float *vf)
547{
548 switch (s->mode) {
549 case COMBINED:
550 // reduce range by channel count
551 *yf = 256.0f / s->nb_display_channels;
552 switch (s->color_mode) {
553 case RAINBOW:
554 case MORELAND:
555 case NEBULAE:
556 case FIRE:
557 case FIERY:
558 case FRUIT:
559 case COOL:
560 case INTENSITY:
561 *uf = *yf;
562 *vf = *yf;
563 break;
564 case CHANNEL:
565 /* adjust saturation for mixed UV coloring */
566 /* this factor is correct for infinite channels, an approximation otherwise */
567 *uf = *yf * M_PI;
568 *vf = *yf * M_PI;
569 break;
570 default:
571 av_assert0(0);
572 }
573 break;
574 case SEPARATE:
575 // full range
576 *yf = 256.0f;
577 *uf = 256.0f;
578 *vf = 256.0f;
579 break;
580 default:
581 av_assert0(0);
582 }
583
584 if (s->color_mode == CHANNEL) {
585 if (s->nb_display_channels > 1) {
586 *uf *= 0.5 * sin((2 * M_PI * ch) / s->nb_display_channels + M_PI * s->rotation);
587 *vf *= 0.5 * cos((2 * M_PI * ch) / s->nb_display_channels + M_PI * s->rotation);
588 } else {
589 *uf *= 0.5 * sin(M_PI * s->rotation);
590 *vf *= 0.5 * cos(M_PI * s->rotation + M_PI_2);
591 }
592 } else {
593 *uf += *uf * sin(M_PI * s->rotation);
594 *vf += *vf * cos(M_PI * s->rotation + M_PI_2);
595 }
596
597 *uf *= s->saturation;
598 *vf *= s->saturation;
599}
600
601static void pick_color(ShowSpectrumContext *s,
602 float yf, float uf, float vf,
603 float a, float *out)
604{
605 if (s->color_mode > CHANNEL) {
606 const int cm = s->color_mode;
607 float y, u, v;
608 int i;
609
610 for (i = 1; i < FF_ARRAY_ELEMS(color_table[cm]) - 1; i++)
611 if (color_table[cm][i].a >= a)
612 break;
613 // i now is the first item >= the color
614 // now we know to interpolate between item i - 1 and i
615 if (a <= color_table[cm][i - 1].a) {
616 y = color_table[cm][i - 1].y;
617 u = color_table[cm][i - 1].u;
618 v = color_table[cm][i - 1].v;
619 } else if (a >= color_table[cm][i].a) {
620 y = color_table[cm][i].y;
621 u = color_table[cm][i].u;
622 v = color_table[cm][i].v;
623 } else {
624 float start = color_table[cm][i - 1].a;
625 float end = color_table[cm][i].a;
626 float lerpfrac = (a - start) / (end - start);
627 y = color_table[cm][i - 1].y * (1.0f - lerpfrac)
628 + color_table[cm][i].y * lerpfrac;
629 u = color_table[cm][i - 1].u * (1.0f - lerpfrac)
630 + color_table[cm][i].u * lerpfrac;
631 v = color_table[cm][i - 1].v * (1.0f - lerpfrac)
632 + color_table[cm][i].v * lerpfrac;
633 }
634
635 out[0] = y * yf;
636 out[1] = u * uf;
637 out[2] = v * vf;
638 } else {
639 out[0] = a * yf;
640 out[1] = a * uf;
641 out[2] = a * vf;
642 }
643}
644
645static void clear_combine_buffer(ShowSpectrumContext *s, int size)
646{
647 int y;
648
649 for (y = 0; y < size; y++) {
650 s->combine_buffer[3 * y ] = 0;
651 s->combine_buffer[3 * y + 1] = 127.5;
652 s->combine_buffer[3 * y + 2] = 127.5;
653 }
654}
655
656static int plot_channel(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)
657{
658 ShowSpectrumContext *s = ctx->priv;
659 const int h = s->orientation == VERTICAL ? s->channel_height : s->channel_width;
660 const int ch = jobnr;
661 float *magnitudes = s->magnitudes[ch];
662 float *phases = s->phases[ch];
663 float yf, uf, vf;
664 int y;
665
666 /* decide color range */
667 color_range(s, ch, &yf, &uf, &vf);
668
669 /* draw the channel */
670 for (y = 0; y < h; y++) {
671 int row = (s->mode == COMBINED) ? y : ch * h + y;
672 float *out = &s->color_buffer[ch][3 * row];
673 float a;
674
675 switch (s->data) {
676 case D_MAGNITUDE:
677 /* get magnitude */
678 a = magnitudes[y];
679 break;
680 case D_PHASE:
681 /* get phase */
682 a = phases[y];
683 break;
684 default:
685 av_assert0(0);
686 }
687
688 /* apply scale */
689 switch (s->scale) {
690 case LINEAR:
691 a = av_clipf(a, 0, 1);
692 break;
693 case SQRT:
694 a = av_clipf(sqrt(a), 0, 1);
695 break;
696 case CBRT:
697 a = av_clipf(cbrt(a), 0, 1);
698 break;
699 case FOURTHRT:
700 a = av_clipf(sqrt(sqrt(a)), 0, 1);
701 break;
702 case FIFTHRT:
703 a = av_clipf(pow(a, 0.20), 0, 1);
704 break;
705 case LOG:
706 a = 1 + log10(av_clipd(a, 1e-6, 1)) / 6; // zero = -120dBFS
707 break;
708 default:
709 av_assert0(0);
710 }
711
712 pick_color(s, yf, uf, vf, a, out);
713 }
714
715 return 0;
716}
717
718static int plot_spectrum_column(AVFilterLink *inlink, AVFrame *insamples)
719{
720 AVFilterContext *ctx = inlink->dst;
721 AVFilterLink *outlink = ctx->outputs[0];
722 ShowSpectrumContext *s = ctx->priv;
723 AVFrame *outpicref = s->outpicref;
724 int ret, plane, x, y, z = s->orientation == VERTICAL ? s->h : s->w;
725
726 /* fill a new spectrum column */
727 /* initialize buffer for combining to black */
728 clear_combine_buffer(s, z);
729
730 ctx->internal->execute(ctx, plot_channel, NULL, NULL, s->nb_display_channels);
731
732 for (y = 0; y < z * 3; y++) {
733 for (x = 0; x < s->nb_display_channels; x++) {
734 s->combine_buffer[y] += s->color_buffer[x][y];
735 }
736 }
737
738 av_frame_make_writable(s->outpicref);
739 /* copy to output */
740 if (s->orientation == VERTICAL) {
741 if (s->sliding == SCROLL) {
742 for (plane = 0; plane < 3; plane++) {
743 for (y = 0; y < s->h; y++) {
744 uint8_t *p = outpicref->data[plane] +
745 y * outpicref->linesize[plane];
746 memmove(p, p + 1, s->w - 1);
747 }
748 }
749 s->xpos = s->w - 1;
750 } else if (s->sliding == RSCROLL) {
751 for (plane = 0; plane < 3; plane++) {
752 for (y = 0; y < s->h; y++) {
753 uint8_t *p = outpicref->data[plane] +
754 y * outpicref->linesize[plane];
755 memmove(p + 1, p, s->w - 1);
756 }
757 }
758 s->xpos = 0;
759 }
760 for (plane = 0; plane < 3; plane++) {
761 uint8_t *p = outpicref->data[plane] + s->start_x +
762 (outlink->h - 1 - s->start_y) * outpicref->linesize[plane] +
763 s->xpos;
764 for (y = 0; y < s->h; y++) {
765 *p = lrintf(av_clipf(s->combine_buffer[3 * y + plane], 0, 255));
766 p -= outpicref->linesize[plane];
767 }
768 }
769 } else {
770 if (s->sliding == SCROLL) {
771 for (plane = 0; plane < 3; plane++) {
772 for (y = 1; y < s->h; y++) {
773 memmove(outpicref->data[plane] + (y-1) * outpicref->linesize[plane],
774 outpicref->data[plane] + (y ) * outpicref->linesize[plane],
775 s->w);
776 }
777 }
778 s->xpos = s->h - 1;
779 } else if (s->sliding == RSCROLL) {
780 for (plane = 0; plane < 3; plane++) {
781 for (y = s->h - 1; y >= 1; y--) {
782 memmove(outpicref->data[plane] + (y ) * outpicref->linesize[plane],
783 outpicref->data[plane] + (y-1) * outpicref->linesize[plane],
784 s->w);
785 }
786 }
787 s->xpos = 0;
788 }
789 for (plane = 0; plane < 3; plane++) {
790 uint8_t *p = outpicref->data[plane] + s->start_x +
791 (s->xpos + s->start_y) * outpicref->linesize[plane];
792 for (x = 0; x < s->w; x++) {
793 *p = lrintf(av_clipf(s->combine_buffer[3 * x + plane], 0, 255));
794 p++;
795 }
796 }
797 }
798
799 if (s->sliding != FULLFRAME || s->xpos == 0)
800 outpicref->pts = insamples->pts;
801
802 s->xpos++;
803 if (s->orientation == VERTICAL && s->xpos >= s->w)
804 s->xpos = 0;
805 if (s->orientation == HORIZONTAL && s->xpos >= s->h)
806 s->xpos = 0;
807 if (!s->single_pic && (s->sliding != FULLFRAME || s->xpos == 0)) {
808 ret = ff_filter_frame(outlink, av_frame_clone(s->outpicref));
809 if (ret < 0)
810 return ret;
811 }
812
813 return s->win_size;
814}
815
816#if CONFIG_SHOWSPECTRUM_FILTER
817
818static int request_frame(AVFilterLink *outlink)
819{
820 ShowSpectrumContext *s = outlink->src->priv;
821 AVFilterLink *inlink = outlink->src->inputs[0];
822 unsigned i;
823 int ret;
824
825 ret = ff_request_frame(inlink);
826 if (ret == AVERROR_EOF && s->sliding == FULLFRAME && s->xpos > 0 &&
827 s->outpicref) {
828 if (s->orientation == VERTICAL) {
829 for (i = 0; i < outlink->h; i++) {
830 memset(s->outpicref->data[0] + i * s->outpicref->linesize[0] + s->xpos, 0, outlink->w - s->xpos);
831 memset(s->outpicref->data[1] + i * s->outpicref->linesize[1] + s->xpos, 128, outlink->w - s->xpos);
832 memset(s->outpicref->data[2] + i * s->outpicref->linesize[2] + s->xpos, 128, outlink->w - s->xpos);
833 }
834 } else {
835 for (i = s->xpos; i < outlink->h; i++) {
836 memset(s->outpicref->data[0] + i * s->outpicref->linesize[0], 0, outlink->w);
837 memset(s->outpicref->data[1] + i * s->outpicref->linesize[1], 128, outlink->w);
838 memset(s->outpicref->data[2] + i * s->outpicref->linesize[2], 128, outlink->w);
839 }
840 }
841 ret = ff_filter_frame(outlink, s->outpicref);
842 s->outpicref = NULL;
843 }
844
845 return ret;
846}
847
848static int filter_frame(AVFilterLink *inlink, AVFrame *insamples)
849{
850 AVFilterContext *ctx = inlink->dst;
851 ShowSpectrumContext *s = ctx->priv;
852 AVFrame *fin = NULL;
853 int ret = 0, consumed = 0;
854
855 if (s->pts == AV_NOPTS_VALUE)
856 s->pts = insamples->pts - av_audio_fifo_size(s->fifo);
857
858 av_audio_fifo_write(s->fifo, (void **)insamples->extended_data, insamples->nb_samples);
859 av_frame_free(&insamples);
860 while (av_audio_fifo_size(s->fifo) >= s->win_size) {
861 fin = ff_get_audio_buffer(inlink, s->win_size);
862 if (!fin) {
863 ret = AVERROR(ENOMEM);
864 goto fail;
865 }
866
867 fin->pts = s->pts + consumed;
868 consumed += s->hop_size;
869 ret = av_audio_fifo_peek(s->fifo, (void **)fin->extended_data, s->win_size);
870 if (ret < 0)
871 goto fail;
872
873 av_assert0(fin->nb_samples == s->win_size);
874
875 ctx->internal->execute(ctx, run_channel_fft, fin, NULL, s->nb_display_channels);
876
877 if (s->data == D_MAGNITUDE)
878 ctx->internal->execute(ctx, calc_channel_magnitudes, NULL, NULL, s->nb_display_channels);
879
880 if (s->data == D_PHASE)
881 ctx->internal->execute(ctx, calc_channel_phases, NULL, NULL, s->nb_display_channels);
882
883 ret = plot_spectrum_column(inlink, fin);
884 av_frame_free(&fin);
885 av_audio_fifo_drain(s->fifo, s->hop_size);
886 if (ret < 0)
887 goto fail;
888 }
889
890fail:
891 s->pts = AV_NOPTS_VALUE;
892 av_frame_free(&fin);
893 return ret;
894}
895
896static const AVFilterPad showspectrum_inputs[] = {
897 {
898 .name = "default",
899 .type = AVMEDIA_TYPE_AUDIO,
900 .filter_frame = filter_frame,
901 },
902 { NULL }
903};
904
905static const AVFilterPad showspectrum_outputs[] = {
906 {
907 .name = "default",
908 .type = AVMEDIA_TYPE_VIDEO,
909 .config_props = config_output,
910 .request_frame = request_frame,
911 },
912 { NULL }
913};
914
915AVFilter ff_avf_showspectrum = {
916 .name = "showspectrum",
917 .description = NULL_IF_CONFIG_SMALL("Convert input audio to a spectrum video output."),
918 .uninit = uninit,
919 .query_formats = query_formats,
920 .priv_size = sizeof(ShowSpectrumContext),
921 .inputs = showspectrum_inputs,
922 .outputs = showspectrum_outputs,
923 .priv_class = &showspectrum_class,
924 .flags = AVFILTER_FLAG_SLICE_THREADS,
925};
926#endif // CONFIG_SHOWSPECTRUM_FILTER
927
928#if CONFIG_SHOWSPECTRUMPIC_FILTER
929
930static const AVOption showspectrumpic_options[] = {
931 { "size", "set video size", OFFSET(w), AV_OPT_TYPE_IMAGE_SIZE, {.str = "4096x2048"}, 0, 0, FLAGS },
932 { "s", "set video size", OFFSET(w), AV_OPT_TYPE_IMAGE_SIZE, {.str = "4096x2048"}, 0, 0, FLAGS },
933 { "mode", "set channel display mode", OFFSET(mode), AV_OPT_TYPE_INT, {.i64=COMBINED}, 0, NB_MODES-1, FLAGS, "mode" },
934 { "combined", "combined mode", 0, AV_OPT_TYPE_CONST, {.i64=COMBINED}, 0, 0, FLAGS, "mode" },
935 { "separate", "separate mode", 0, AV_OPT_TYPE_CONST, {.i64=SEPARATE}, 0, 0, FLAGS, "mode" },
936 { "color", "set channel coloring", OFFSET(color_mode), AV_OPT_TYPE_INT, {.i64=INTENSITY}, 0, NB_CLMODES-1, FLAGS, "color" },
937 { "channel", "separate color for each channel", 0, AV_OPT_TYPE_CONST, {.i64=CHANNEL}, 0, 0, FLAGS, "color" },
938 { "intensity", "intensity based coloring", 0, AV_OPT_TYPE_CONST, {.i64=INTENSITY}, 0, 0, FLAGS, "color" },
939 { "rainbow", "rainbow based coloring", 0, AV_OPT_TYPE_CONST, {.i64=RAINBOW}, 0, 0, FLAGS, "color" },
940 { "moreland", "moreland based coloring", 0, AV_OPT_TYPE_CONST, {.i64=MORELAND}, 0, 0, FLAGS, "color" },
941 { "nebulae", "nebulae based coloring", 0, AV_OPT_TYPE_CONST, {.i64=NEBULAE}, 0, 0, FLAGS, "color" },
942 { "fire", "fire based coloring", 0, AV_OPT_TYPE_CONST, {.i64=FIRE}, 0, 0, FLAGS, "color" },
943 { "fiery", "fiery based coloring", 0, AV_OPT_TYPE_CONST, {.i64=FIERY}, 0, 0, FLAGS, "color" },
944 { "fruit", "fruit based coloring", 0, AV_OPT_TYPE_CONST, {.i64=FRUIT}, 0, 0, FLAGS, "color" },
945 { "cool", "cool based coloring", 0, AV_OPT_TYPE_CONST, {.i64=COOL}, 0, 0, FLAGS, "color" },
946 { "scale", "set display scale", OFFSET(scale), AV_OPT_TYPE_INT, {.i64=LOG}, 0, NB_SCALES-1, FLAGS, "scale" },
947 { "lin", "linear", 0, AV_OPT_TYPE_CONST, {.i64=LINEAR}, 0, 0, FLAGS, "scale" },
948 { "sqrt", "square root", 0, AV_OPT_TYPE_CONST, {.i64=SQRT}, 0, 0, FLAGS, "scale" },
949 { "cbrt", "cubic root", 0, AV_OPT_TYPE_CONST, {.i64=CBRT}, 0, 0, FLAGS, "scale" },
950 { "log", "logarithmic", 0, AV_OPT_TYPE_CONST, {.i64=LOG}, 0, 0, FLAGS, "scale" },
951 { "4thrt","4th root", 0, AV_OPT_TYPE_CONST, {.i64=FOURTHRT}, 0, 0, FLAGS, "scale" },
952 { "5thrt","5th root", 0, AV_OPT_TYPE_CONST, {.i64=FIFTHRT}, 0, 0, FLAGS, "scale" },
953 { "saturation", "color saturation multiplier", OFFSET(saturation), AV_OPT_TYPE_FLOAT, {.dbl = 1}, -10, 10, FLAGS },
954 { "win_func", "set window function", OFFSET(win_func), AV_OPT_TYPE_INT, {.i64 = WFUNC_HANNING}, 0, NB_WFUNC-1, FLAGS, "win_func" },
955 { "rect", "Rectangular", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_RECT}, 0, 0, FLAGS, "win_func" },
956 { "bartlett", "Bartlett", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_BARTLETT}, 0, 0, FLAGS, "win_func" },
957 { "hann", "Hann", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_HANNING}, 0, 0, FLAGS, "win_func" },
958 { "hanning", "Hanning", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_HANNING}, 0, 0, FLAGS, "win_func" },
959 { "hamming", "Hamming", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_HAMMING}, 0, 0, FLAGS, "win_func" },
960 { "blackman", "Blackman", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_BLACKMAN}, 0, 0, FLAGS, "win_func" },
961 { "welch", "Welch", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_WELCH}, 0, 0, FLAGS, "win_func" },
962 { "flattop", "Flat-top", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_FLATTOP}, 0, 0, FLAGS, "win_func" },
963 { "bharris", "Blackman-Harris", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_BHARRIS}, 0, 0, FLAGS, "win_func" },
964 { "bnuttall", "Blackman-Nuttall", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_BNUTTALL}, 0, 0, FLAGS, "win_func" },
965 { "bhann", "Bartlett-Hann", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_BHANN}, 0, 0, FLAGS, "win_func" },
966 { "sine", "Sine", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_SINE}, 0, 0, FLAGS, "win_func" },
967 { "nuttall", "Nuttall", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_NUTTALL}, 0, 0, FLAGS, "win_func" },
968 { "lanczos", "Lanczos", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_LANCZOS}, 0, 0, FLAGS, "win_func" },
969 { "gauss", "Gauss", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_GAUSS}, 0, 0, FLAGS, "win_func" },
970 { "tukey", "Tukey", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_TUKEY}, 0, 0, FLAGS, "win_func" },
971 { "dolph", "Dolph-Chebyshev", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_DOLPH}, 0, 0, FLAGS, "win_func" },
972 { "cauchy", "Cauchy", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_CAUCHY}, 0, 0, FLAGS, "win_func" },
973 { "parzen", "Parzen", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_PARZEN}, 0, 0, FLAGS, "win_func" },
974 { "poisson", "Poisson", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_POISSON}, 0, 0, FLAGS, "win_func" },
975 { "orientation", "set orientation", OFFSET(orientation), AV_OPT_TYPE_INT, {.i64=VERTICAL}, 0, NB_ORIENTATIONS-1, FLAGS, "orientation" },
976 { "vertical", NULL, 0, AV_OPT_TYPE_CONST, {.i64=VERTICAL}, 0, 0, FLAGS, "orientation" },
977 { "horizontal", NULL, 0, AV_OPT_TYPE_CONST, {.i64=HORIZONTAL}, 0, 0, FLAGS, "orientation" },
978 { "gain", "set scale gain", OFFSET(gain), AV_OPT_TYPE_FLOAT, {.dbl = 1}, 0, 128, FLAGS },
979 { "legend", "draw legend", OFFSET(legend), AV_OPT_TYPE_BOOL, {.i64 = 1}, 0, 1, FLAGS },
980 { "rotation", "color rotation", OFFSET(rotation), AV_OPT_TYPE_FLOAT, {.dbl = 0}, -1, 1, FLAGS },
981 { NULL }
982};
983
984AVFILTER_DEFINE_CLASS(showspectrumpic);
985
986static void drawtext(AVFrame *pic, int x, int y, const char *txt, int o)
987{
988 const uint8_t *font;
989 int font_height;
990 int i;
991
992 font = avpriv_cga_font, font_height = 8;
993
994 for (i = 0; txt[i]; i++) {
995 int char_y, mask;
996
997 if (o) {
998 for (char_y = font_height - 1; char_y >= 0; char_y--) {
999 uint8_t *p = pic->data[0] + (y + i * 10) * pic->linesize[0] + x;
1000 for (mask = 0x80; mask; mask >>= 1) {
1001 if (font[txt[i] * font_height + font_height - 1 - char_y] & mask)
1002 p[char_y] = ~p[char_y];
1003 p += pic->linesize[0];
1004 }
1005 }
1006 } else {
1007 uint8_t *p = pic->data[0] + y*pic->linesize[0] + (x + i*8);
1008 for (char_y = 0; char_y < font_height; char_y++) {
1009 for (mask = 0x80; mask; mask >>= 1) {
1010 if (font[txt[i] * font_height + char_y] & mask)
1011 *p = ~(*p);
1012 p++;
1013 }
1014 p += pic->linesize[0] - 8;
1015 }
1016 }
1017 }
1018}
1019
1020static int showspectrumpic_request_frame(AVFilterLink *outlink)
1021{
1022 AVFilterContext *ctx = outlink->src;
1023 ShowSpectrumContext *s = ctx->priv;
1024 AVFilterLink *inlink = ctx->inputs[0];
1025 int ret, samples;
1026
1027 ret = ff_request_frame(inlink);
1028 samples = av_audio_fifo_size(s->fifo);
1029 if (ret == AVERROR_EOF && s->outpicref && samples > 0) {
1030 int consumed = 0;
1031 int y, x = 0, sz = s->orientation == VERTICAL ? s->w : s->h;
1032 int ch, spf, spb;
1033 AVFrame *fin;
1034
1035 spf = s->win_size * (samples / ((s->win_size * sz) * ceil(samples / (float)(s->win_size * sz))));
1036 spf = FFMAX(1, spf);
1037
1038 spb = (samples / (spf * sz)) * spf;
1039
1040 fin = ff_get_audio_buffer(inlink, s->win_size);
1041 if (!fin)
1042 return AVERROR(ENOMEM);
1043
1044 while (x < sz) {
1045 ret = av_audio_fifo_peek(s->fifo, (void **)fin->extended_data, s->win_size);
1046 if (ret < 0) {
1047 av_frame_free(&fin);
1048 return ret;
1049 }
1050
1051 av_audio_fifo_drain(s->fifo, spf);
1052
1053 if (ret < s->win_size) {
1054 for (ch = 0; ch < s->nb_display_channels; ch++) {
1055 memset(fin->extended_data[ch] + ret * sizeof(float), 0,
1056 (s->win_size - ret) * sizeof(float));
1057 }
1058 }
1059
1060 ctx->internal->execute(ctx, run_channel_fft, fin, NULL, s->nb_display_channels);
1061 acalc_magnitudes(s);
1062
1063 consumed += spf;
1064 if (consumed >= spb) {
1065 int h = s->orientation == VERTICAL ? s->h : s->w;
1066
1067 scale_magnitudes(s, 1. / (consumed / spf));
1068 plot_spectrum_column(inlink, fin);
1069 consumed = 0;
1070 x++;
1071 for (ch = 0; ch < s->nb_display_channels; ch++)
1072 memset(s->magnitudes[ch], 0, h * sizeof(float));
1073 }
1074 }
1075
1076 av_frame_free(&fin);
1077 s->outpicref->pts = 0;
1078
1079 if (s->legend) {
1080 int multi = (s->mode == SEPARATE && s->color_mode == CHANNEL);
1081 float spp = samples / (float)sz;
1082 uint8_t *dst;
1083
1084 drawtext(s->outpicref, 2, outlink->h - 10, "CREATED BY LIBAVFILTER", 0);
1085
1086 dst = s->outpicref->data[0] + (s->start_y - 1) * s->outpicref->linesize[0] + s->start_x - 1;
1087 for (x = 0; x < s->w + 1; x++)
1088 dst[x] = 200;
1089 dst = s->outpicref->data[0] + (s->start_y + s->h) * s->outpicref->linesize[0] + s->start_x - 1;
1090 for (x = 0; x < s->w + 1; x++)
1091 dst[x] = 200;
1092 for (y = 0; y < s->h + 2; y++) {
1093 dst = s->outpicref->data[0] + (y + s->start_y - 1) * s->outpicref->linesize[0];
1094 dst[s->start_x - 1] = 200;
1095 dst[s->start_x + s->w] = 200;
1096 }
1097 if (s->orientation == VERTICAL) {
1098 int h = s->mode == SEPARATE ? s->h / s->nb_display_channels : s->h;
1099 for (ch = 0; ch < (s->mode == SEPARATE ? s->nb_display_channels : 1); ch++) {
1100 for (y = 0; y < h; y += 20) {
1101 dst = s->outpicref->data[0] + (s->start_y + h * (ch + 1) - y - 1) * s->outpicref->linesize[0];
1102 dst[s->start_x - 2] = 200;
1103 dst[s->start_x + s->w + 1] = 200;
1104 }
1105 for (y = 0; y < h; y += 40) {
1106 dst = s->outpicref->data[0] + (s->start_y + h * (ch + 1) - y - 1) * s->outpicref->linesize[0];
1107 dst[s->start_x - 3] = 200;
1108 dst[s->start_x + s->w + 2] = 200;
1109 }
1110 dst = s->outpicref->data[0] + (s->start_y - 2) * s->outpicref->linesize[0] + s->start_x;
1111 for (x = 0; x < s->w; x+=40)
1112 dst[x] = 200;
1113 dst = s->outpicref->data[0] + (s->start_y - 3) * s->outpicref->linesize[0] + s->start_x;
1114 for (x = 0; x < s->w; x+=80)
1115 dst[x] = 200;
1116 dst = s->outpicref->data[0] + (s->h + s->start_y + 1) * s->outpicref->linesize[0] + s->start_x;
1117 for (x = 0; x < s->w; x+=40) {
1118 dst[x] = 200;
1119 }
1120 dst = s->outpicref->data[0] + (s->h + s->start_y + 2) * s->outpicref->linesize[0] + s->start_x;
1121 for (x = 0; x < s->w; x+=80) {
1122 dst[x] = 200;
1123 }
1124 for (y = 0; y < h; y += 40) {
1125 float hertz = y * (inlink->sample_rate / 2) / (float)(1 << (int)ceil(log2(h)));
1126 char *units;
1127
1128 if (hertz == 0)
1129 units = av_asprintf("DC");
1130 else
1131 units = av_asprintf("%.2f", hertz);
1132 if (!units)
1133 return AVERROR(ENOMEM);
1134
1135 drawtext(s->outpicref, s->start_x - 8 * strlen(units) - 4, h * (ch + 1) + s->start_y - y - 4, units, 0);
1136 av_free(units);
1137 }
1138 }
1139
1140 for (x = 0; x < s->w; x+=80) {
1141 float seconds = x * spp / inlink->sample_rate;
1142 char *units;
1143
1144 if (x == 0)
1145 units = av_asprintf("0");
1146 else if (log10(seconds) > 6)
1147 units = av_asprintf("%.2fh", seconds / (60 * 60));
1148 else if (log10(seconds) > 3)
1149 units = av_asprintf("%.2fm", seconds / 60);
1150 else
1151 units = av_asprintf("%.2fs", seconds);
1152 if (!units)
1153 return AVERROR(ENOMEM);
1154
1155 drawtext(s->outpicref, s->start_x + x - 4 * strlen(units), s->h + s->start_y + 6, units, 0);
1156 drawtext(s->outpicref, s->start_x + x - 4 * strlen(units), s->start_y - 12, units, 0);
1157 av_free(units);
1158 }
1159
1160 drawtext(s->outpicref, outlink->w / 2 - 4 * 4, outlink->h - s->start_y / 2, "TIME", 0);
1161 drawtext(s->outpicref, s->start_x / 7, outlink->h / 2 - 14 * 4, "FREQUENCY (Hz)", 1);
1162 } else {
1163 int w = s->mode == SEPARATE ? s->w / s->nb_display_channels : s->w;
1164 for (y = 0; y < s->h; y += 20) {
1165 dst = s->outpicref->data[0] + (s->start_y + y) * s->outpicref->linesize[0];
1166 dst[s->start_x - 2] = 200;
1167 dst[s->start_x + s->w + 1] = 200;
1168 }
1169 for (y = 0; y < s->h; y += 40) {
1170 dst = s->outpicref->data[0] + (s->start_y + y) * s->outpicref->linesize[0];
1171 dst[s->start_x - 3] = 200;
1172 dst[s->start_x + s->w + 2] = 200;
1173 }
1174 for (ch = 0; ch < (s->mode == SEPARATE ? s->nb_display_channels : 1); ch++) {
1175 dst = s->outpicref->data[0] + (s->start_y - 2) * s->outpicref->linesize[0] + s->start_x + w * ch;
1176 for (x = 0; x < w; x+=40)
1177 dst[x] = 200;
1178 dst = s->outpicref->data[0] + (s->start_y - 3) * s->outpicref->linesize[0] + s->start_x + w * ch;
1179 for (x = 0; x < w; x+=80)
1180 dst[x] = 200;
1181 dst = s->outpicref->data[0] + (s->h + s->start_y + 1) * s->outpicref->linesize[0] + s->start_x + w * ch;
1182 for (x = 0; x < w; x+=40) {
1183 dst[x] = 200;
1184 }
1185 dst = s->outpicref->data[0] + (s->h + s->start_y + 2) * s->outpicref->linesize[0] + s->start_x + w * ch;
1186 for (x = 0; x < w; x+=80) {
1187 dst[x] = 200;
1188 }
1189 for (x = 0; x < w; x += 80) {
1190 float hertz = x * (inlink->sample_rate / 2) / (float)(1 << (int)ceil(log2(w)));
1191 char *units;
1192
1193 if (hertz == 0)
1194 units = av_asprintf("DC");
1195 else
1196 units = av_asprintf("%.2f", hertz);
1197 if (!units)
1198 return AVERROR(ENOMEM);
1199
1200 drawtext(s->outpicref, s->start_x - 4 * strlen(units) + x + w * ch, s->start_y - 12, units, 0);
1201 drawtext(s->outpicref, s->start_x - 4 * strlen(units) + x + w * ch, s->h + s->start_y + 6, units, 0);
1202 av_free(units);
1203 }
1204 }
1205 for (y = 0; y < s->h; y+=40) {
1206 float seconds = y * spp / inlink->sample_rate;
1207 char *units;
1208
1209 if (x == 0)
1210 units = av_asprintf("0");
1211 else if (log10(seconds) > 6)
1212 units = av_asprintf("%.2fh", seconds / (60 * 60));
1213 else if (log10(seconds) > 3)
1214 units = av_asprintf("%.2fm", seconds / 60);
1215 else
1216 units = av_asprintf("%.2fs", seconds);
1217 if (!units)
1218 return AVERROR(ENOMEM);
1219
1220 drawtext(s->outpicref, s->start_x - 8 * strlen(units) - 4, s->start_y + y - 4, units, 0);
1221 av_free(units);
1222 }
1223 drawtext(s->outpicref, s->start_x / 7, outlink->h / 2 - 4 * 4, "TIME", 1);
1224 drawtext(s->outpicref, outlink->w / 2 - 14 * 4, outlink->h - s->start_y / 2, "FREQUENCY (Hz)", 0);
1225 }
1226
1227 for (ch = 0; ch < (multi ? s->nb_display_channels : 1); ch++) {
1228 int h = multi ? s->h / s->nb_display_channels : s->h;
1229
1230 for (y = 0; y < h; y++) {
1231 float out[3] = { 0., 127.5, 127.5};
1232 int chn;
1233
1234 for (chn = 0; chn < (s->mode == SEPARATE ? 1 : s->nb_display_channels); chn++) {
1235 float yf, uf, vf;
1236 int channel = (multi) ? s->nb_display_channels - ch - 1 : chn;
1237 float lout[3];
1238
1239 color_range(s, channel, &yf, &uf, &vf);
1240 pick_color(s, yf, uf, vf, y / (float)h, lout);
1241 out[0] += lout[0];
1242 out[1] += lout[1];
1243 out[2] += lout[2];
1244 }
1245 memset(s->outpicref->data[0]+(s->start_y + h * (ch + 1) - y - 1) * s->outpicref->linesize[0] + s->w + s->start_x + 20, av_clip_uint8(out[0]), 10);
1246 memset(s->outpicref->data[1]+(s->start_y + h * (ch + 1) - y - 1) * s->outpicref->linesize[1] + s->w + s->start_x + 20, av_clip_uint8(out[1]), 10);
1247 memset(s->outpicref->data[2]+(s->start_y + h * (ch + 1) - y - 1) * s->outpicref->linesize[2] + s->w + s->start_x + 20, av_clip_uint8(out[2]), 10);
1248 }
1249
1250 for (y = 0; ch == 0 && y < h; y += h / 10) {
1251 float value = 120.0 * log10(1. - y / (float)h);
1252 char *text;
1253
1254 if (value < -120)
1255 break;
1256 text = av_asprintf("%.0f dB", value);
1257 if (!text)
1258 continue;
1259 drawtext(s->outpicref, s->w + s->start_x + 35, s->start_y + y - 5, text, 0);
1260 av_free(text);
1261 }
1262 }
1263 }
1264
1265 ret = ff_filter_frame(outlink, s->outpicref);
1266 s->outpicref = NULL;
1267 }
1268
1269 return ret;
1270}
1271
1272static int showspectrumpic_filter_frame(AVFilterLink *inlink, AVFrame *insamples)
1273{
1274 AVFilterContext *ctx = inlink->dst;
1275 ShowSpectrumContext *s = ctx->priv;
1276 int ret;
1277
1278 ret = av_audio_fifo_write(s->fifo, (void **)insamples->extended_data, insamples->nb_samples);
1279 av_frame_free(&insamples);
1280 return ret;
1281}
1282
1283static const AVFilterPad showspectrumpic_inputs[] = {
1284 {
1285 .name = "default",
1286 .type = AVMEDIA_TYPE_AUDIO,
1287 .filter_frame = showspectrumpic_filter_frame,
1288 },
1289 { NULL }
1290};
1291
1292static const AVFilterPad showspectrumpic_outputs[] = {
1293 {
1294 .name = "default",
1295 .type = AVMEDIA_TYPE_VIDEO,
1296 .config_props = config_output,
1297 .request_frame = showspectrumpic_request_frame,
1298 },
1299 { NULL }
1300};
1301
1302AVFilter ff_avf_showspectrumpic = {
1303 .name = "showspectrumpic",
1304 .description = NULL_IF_CONFIG_SMALL("Convert input audio to a spectrum video output single picture."),
1305 .uninit = uninit,
1306 .query_formats = query_formats,
1307 .priv_size = sizeof(ShowSpectrumContext),
1308 .inputs = showspectrumpic_inputs,
1309 .outputs = showspectrumpic_outputs,
1310 .priv_class = &showspectrumpic_class,
1311 .flags = AVFILTER_FLAG_SLICE_THREADS,
1312};
1313
1314#endif // CONFIG_SHOWSPECTRUMPIC_FILTER
1315
generated by cgit at 2024-05-17 04:44:39 (GMT)