2 * Copyright (c) 2019 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/channel_layout.h"
28 #include "libavutil/opt.h"
29 #include "libavutil/parseutils.h"
35 #include "window_func.h"
37 typedef struct ShowSpatialContext {
40 AVRational frame_rate;
41 FFTContext *fft[2]; ///< Fast Fourier Transform context
42 FFTContext *ifft[2]; ///< Inverse Fast Fourier Transform context
43 int fft_bits; ///< number of bits (FFT window size = 1<<fft_bits)
44 FFTComplex *fft_data[2]; ///< bins holder for each (displayed) channels
45 float *window_func_lut; ///< Window function LUT
56 #define OFFSET(x) offsetof(ShowSpatialContext, x)
57 #define FLAGS AV_OPT_FLAG_FILTERING_PARAM|AV_OPT_FLAG_VIDEO_PARAM
59 static const AVOption showspatial_options[] = {
60 { "size", "set video size", OFFSET(w), AV_OPT_TYPE_IMAGE_SIZE, {.str = "512x512"}, 0, 0, FLAGS },
61 { "s", "set video size", OFFSET(w), AV_OPT_TYPE_IMAGE_SIZE, {.str = "512x512"}, 0, 0, FLAGS },
62 { "win_size", "set window size", OFFSET(win_size), AV_OPT_TYPE_INT, {.i64 = 4096}, 1024, 65536, FLAGS },
63 { "win_func", "set window function", OFFSET(win_func), AV_OPT_TYPE_INT, {.i64 = WFUNC_HANNING}, 0, NB_WFUNC-1, FLAGS, "win_func" },
64 { "rect", "Rectangular", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_RECT}, 0, 0, FLAGS, "win_func" },
65 { "bartlett", "Bartlett", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_BARTLETT}, 0, 0, FLAGS, "win_func" },
66 { "hann", "Hann", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_HANNING}, 0, 0, FLAGS, "win_func" },
67 { "hanning", "Hanning", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_HANNING}, 0, 0, FLAGS, "win_func" },
68 { "hamming", "Hamming", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_HAMMING}, 0, 0, FLAGS, "win_func" },
69 { "blackman", "Blackman", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_BLACKMAN}, 0, 0, FLAGS, "win_func" },
70 { "welch", "Welch", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_WELCH}, 0, 0, FLAGS, "win_func" },
71 { "flattop", "Flat-top", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_FLATTOP}, 0, 0, FLAGS, "win_func" },
72 { "bharris", "Blackman-Harris", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_BHARRIS}, 0, 0, FLAGS, "win_func" },
73 { "bnuttall", "Blackman-Nuttall", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_BNUTTALL}, 0, 0, FLAGS, "win_func" },
74 { "bhann", "Bartlett-Hann", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_BHANN}, 0, 0, FLAGS, "win_func" },
75 { "sine", "Sine", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_SINE}, 0, 0, FLAGS, "win_func" },
76 { "nuttall", "Nuttall", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_NUTTALL}, 0, 0, FLAGS, "win_func" },
77 { "lanczos", "Lanczos", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_LANCZOS}, 0, 0, FLAGS, "win_func" },
78 { "gauss", "Gauss", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_GAUSS}, 0, 0, FLAGS, "win_func" },
79 { "tukey", "Tukey", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_TUKEY}, 0, 0, FLAGS, "win_func" },
80 { "dolph", "Dolph-Chebyshev", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_DOLPH}, 0, 0, FLAGS, "win_func" },
81 { "cauchy", "Cauchy", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_CAUCHY}, 0, 0, FLAGS, "win_func" },
82 { "parzen", "Parzen", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_PARZEN}, 0, 0, FLAGS, "win_func" },
83 { "poisson", "Poisson", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_POISSON}, 0, 0, FLAGS, "win_func" },
84 { "bohman", "Bohman", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_BOHMAN}, 0, 0, FLAGS, "win_func" },
85 { "overlap", "set window overlap", OFFSET(overlap), AV_OPT_TYPE_FLOAT, {.dbl=0.5}, 0, 1, FLAGS },
89 AVFILTER_DEFINE_CLASS(showspatial);
91 static av_cold void uninit(AVFilterContext *ctx)
93 ShowSpatialContext *s = ctx->priv;
96 for (i = 0; i < 2; i++)
97 av_fft_end(s->fft[i]);
98 for (i = 0; i < 2; i++)
99 av_fft_end(s->ifft[i]);
100 for (i = 0; i < 2; i++)
101 av_freep(&s->fft_data[i]);
102 av_freep(&s->window_func_lut);
103 av_audio_fifo_free(s->fifo);
106 static int query_formats(AVFilterContext *ctx)
108 AVFilterFormats *formats = NULL;
109 AVFilterChannelLayouts *layout = NULL;
110 AVFilterLink *inlink = ctx->inputs[0];
111 AVFilterLink *outlink = ctx->outputs[0];
112 static const enum AVSampleFormat sample_fmts[] = { AV_SAMPLE_FMT_FLTP, AV_SAMPLE_FMT_NONE };
113 static const enum AVPixelFormat pix_fmts[] = { AV_PIX_FMT_GBRP, AV_PIX_FMT_NONE };
116 formats = ff_make_format_list(sample_fmts);
117 if ((ret = ff_formats_ref (formats, &inlink->outcfg.formats )) < 0 ||
118 (ret = ff_add_channel_layout (&layout, AV_CH_LAYOUT_STEREO )) < 0 ||
119 (ret = ff_channel_layouts_ref (layout , &inlink->outcfg.channel_layouts)) < 0)
122 formats = ff_all_samplerates();
123 if ((ret = ff_formats_ref(formats, &inlink->outcfg.samplerates)) < 0)
126 formats = ff_make_format_list(pix_fmts);
127 if ((ret = ff_formats_ref(formats, &outlink->incfg.formats)) < 0)
133 static int run_channel_fft(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)
135 ShowSpatialContext *s = ctx->priv;
136 const float *window_func_lut = s->window_func_lut;
138 const int ch = jobnr;
139 const float *p = (float *)fin->extended_data[ch];
141 for (int n = 0; n < fin->nb_samples; n++) {
142 s->fft_data[ch][n].re = p[n] * window_func_lut[n];
143 s->fft_data[ch][n].im = 0;
146 av_fft_permute(s->fft[ch], s->fft_data[ch]);
147 av_fft_calc(s->fft[ch], s->fft_data[ch]);
152 static int config_output(AVFilterLink *outlink)
154 AVFilterContext *ctx = outlink->src;
155 AVFilterLink *inlink = ctx->inputs[0];
156 ShowSpatialContext *s = ctx->priv;
162 outlink->sample_aspect_ratio = (AVRational){1,1};
164 s->buf_size = 1 << av_log2(s->win_size);
165 s->win_size = s->buf_size;
166 fft_bits = av_log2(s->win_size);
168 /* (re-)configuration if the video output changed (or first init) */
169 if (fft_bits != s->fft_bits) {
170 s->fft_bits = fft_bits;
172 /* FFT buffers: x2 for each channel buffer.
173 * Note: we use free and malloc instead of a realloc-like function to
174 * make sure the buffer is aligned in memory for the FFT functions. */
175 for (i = 0; i < 2; i++) {
176 av_fft_end(s->fft[i]);
177 av_freep(&s->fft_data[i]);
179 for (i = 0; i < 2; i++) {
180 s->fft[i] = av_fft_init(fft_bits, 0);
182 av_log(ctx, AV_LOG_ERROR, "Unable to create FFT context. "
183 "The window size might be too high.\n");
184 return AVERROR(EINVAL);
188 for (i = 0; i < 2; i++) {
189 s->fft_data[i] = av_calloc(s->buf_size, sizeof(**s->fft_data));
191 return AVERROR(ENOMEM);
194 /* pre-calc windowing function */
196 av_realloc_f(s->window_func_lut, s->win_size,
197 sizeof(*s->window_func_lut));
198 if (!s->window_func_lut)
199 return AVERROR(ENOMEM);
200 generate_window_func(s->window_func_lut, s->win_size, s->win_func, &overlap);
202 s->overlap = overlap;
204 s->hop_size = (1.f - s->overlap) * s->win_size;
205 if (s->hop_size < 1) {
206 av_log(ctx, AV_LOG_ERROR, "overlap %f too big\n", s->overlap);
207 return AVERROR(EINVAL);
211 outlink->time_base = av_inv_q(outlink->frame_rate);
213 av_audio_fifo_free(s->fifo);
214 s->fifo = av_audio_fifo_alloc(inlink->format, inlink->channels, s->win_size);
216 return AVERROR(ENOMEM);
220 #define RE(y, ch) s->fft_data[ch][y].re
221 #define IM(y, ch) s->fft_data[ch][y].im
223 static void draw_dot(uint8_t *dst, int linesize, int value)
228 dst[linesize] = value;
229 dst[-linesize] = value;
232 static int draw_spatial(AVFilterLink *inlink, AVFrame *insamples)
234 AVFilterContext *ctx = inlink->dst;
235 AVFilterLink *outlink = ctx->outputs[0];
236 ShowSpatialContext *s = ctx->priv;
240 int z = s->win_size / 2;
242 outpicref = ff_get_video_buffer(outlink, outlink->w, outlink->h);
244 return AVERROR(ENOMEM);
246 outpicref->sample_aspect_ratio = (AVRational){1,1};
247 for (int i = 0; i < outlink->h; i++) {
248 memset(outpicref->data[0] + i * outpicref->linesize[0], 0, outlink->w);
249 memset(outpicref->data[1] + i * outpicref->linesize[1], 0, outlink->w);
250 memset(outpicref->data[2] + i * outpicref->linesize[2], 0, outlink->w);
253 for (int j = 0; j < z; j++) {
254 const int idx = z - 1 - j;
255 float l = hypotf(RE(idx, 0), IM(idx, 0));
256 float r = hypotf(RE(idx, 1), IM(idx, 1));
258 float lp = atan2f(IM(idx, 0), RE(idx, 0));
259 float rp = atan2f(IM(idx, 1), RE(idx, 1));
260 float diffp = ((rp - lp) / (2.f * M_PI) + 1.f) * 0.5f;
261 float diff = (sum < 0.000001f ? 0.f : (r - l) / sum) * 0.5f + 0.5f;
262 float cr = av_clipf(cbrtf(l / sum), 0, 1) * 255.f;
263 float cb = av_clipf(cbrtf(r / sum), 0, 1) * 255.f;
268 x = av_clip(w * diff, 0, w - 2) + 1;
269 y = av_clip(h * diffp, 0, h - 2) + 1;
271 draw_dot(outpicref->data[0] + outpicref->linesize[0] * y + x, outpicref->linesize[0], cg);
272 draw_dot(outpicref->data[1] + outpicref->linesize[1] * y + x, outpicref->linesize[1], cb);
273 draw_dot(outpicref->data[2] + outpicref->linesize[2] * y + x, outpicref->linesize[2], cr);
276 outpicref->pts = av_rescale_q(insamples->pts, inlink->time_base, outlink->time_base);
278 return ff_filter_frame(outlink, outpicref);
281 static int spatial_activate(AVFilterContext *ctx)
283 AVFilterLink *inlink = ctx->inputs[0];
284 AVFilterLink *outlink = ctx->outputs[0];
285 ShowSpatialContext *s = ctx->priv;
288 FF_FILTER_FORWARD_STATUS_BACK(outlink, inlink);
290 if (av_audio_fifo_size(s->fifo) < s->win_size) {
291 AVFrame *frame = NULL;
293 ret = ff_inlink_consume_frame(inlink, &frame);
300 av_audio_fifo_write(s->fifo, (void **)frame->extended_data, frame->nb_samples);
301 av_frame_free(&frame);
305 if (av_audio_fifo_size(s->fifo) >= s->win_size) {
306 AVFrame *fin = ff_get_audio_buffer(inlink, s->win_size);
308 return AVERROR(ENOMEM);
310 fin->pts = s->pts + s->consumed;
311 s->consumed += s->hop_size;
312 ret = av_audio_fifo_peek(s->fifo, (void **)fin->extended_data,
313 FFMIN(s->win_size, av_audio_fifo_size(s->fifo)));
319 av_assert0(fin->nb_samples == s->win_size);
321 ctx->internal->execute(ctx, run_channel_fft, fin, NULL, 2);
323 ret = draw_spatial(inlink, fin);
326 av_audio_fifo_drain(s->fifo, s->hop_size);
331 FF_FILTER_FORWARD_STATUS(inlink, outlink);
332 if (ff_outlink_frame_wanted(outlink) && av_audio_fifo_size(s->fifo) < s->win_size) {
333 ff_inlink_request_frame(inlink);
337 if (av_audio_fifo_size(s->fifo) >= s->win_size) {
338 ff_filter_set_ready(ctx, 10);
341 return FFERROR_NOT_READY;
344 static const AVFilterPad showspatial_inputs[] = {
347 .type = AVMEDIA_TYPE_AUDIO,
352 static const AVFilterPad showspatial_outputs[] = {
355 .type = AVMEDIA_TYPE_VIDEO,
356 .config_props = config_output,
361 const AVFilter ff_avf_showspatial = {
362 .name = "showspatial",
363 .description = NULL_IF_CONFIG_SMALL("Convert input audio to a spatial video output."),
365 .query_formats = query_formats,
366 .priv_size = sizeof(ShowSpatialContext),
367 .inputs = showspatial_inputs,
368 .outputs = showspatial_outputs,
369 .activate = spatial_activate,
370 .priv_class = &showspatial_class,
371 .flags = AVFILTER_FLAG_SLICE_THREADS,