• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 /*
2  * Copyright (c) 2014-2015 Muhammad Faiz <mfcc64@gmail.com>
3  *
4  * This file is part of FFmpeg.
5  *
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.
10  *
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.
15  *
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
19  */
20 
21 #include "config.h"
22 #include "libavutil/tx.h"
23 #include "libavutil/channel_layout.h"
24 #include "libavutil/opt.h"
25 #include "libavutil/xga_font_data.h"
26 #include "libavutil/eval.h"
27 #include "libavutil/pixdesc.h"
28 #include "libavutil/time.h"
29 #include "avfilter.h"
30 #include "internal.h"
31 #include "lavfutils.h"
32 #include "lswsutils.h"
33 
34 #if CONFIG_LIBFREETYPE
35 #include <ft2build.h>
36 #include FT_FREETYPE_H
37 #endif
38 
39 #if CONFIG_LIBFONTCONFIG
40 #include <fontconfig/fontconfig.h>
41 #endif
42 
43 #include "avf_showcqt.h"
44 
45 #define BASEFREQ        20.01523126408007475
46 #define ENDFREQ         20495.59681441799654
47 #define TLENGTH         "384*tc/(384+tc*f)"
48 #define TLENGTH_MIN     0.001
49 #define VOLUME_MAX      100.0
50 #define FONTCOLOR       "st(0, (midi(f)-59.5)/12);" \
51     "st(1, if(between(ld(0),0,1), 0.5-0.5*cos(2*PI*ld(0)), 0));" \
52     "r(1-ld(1)) + b(ld(1))"
53 #define CSCHEME         "1|0.5|0|0|0.5|1"
54 #define PTS_STEP 10
55 #define PTS_TOLERANCE 1
56 
57 #define OFFSET(x) offsetof(ShowCQTContext, x)
58 #define FLAGS (AV_OPT_FLAG_FILTERING_PARAM|AV_OPT_FLAG_VIDEO_PARAM)
59 
60 static const AVOption showcqt_options[] = {
61     { "size",         "set video size", OFFSET(width), AV_OPT_TYPE_IMAGE_SIZE, { .str = "1920x1080" },      0, 0,        FLAGS },
62     { "s",            "set video size", OFFSET(width), AV_OPT_TYPE_IMAGE_SIZE, { .str = "1920x1080" },      0, 0,        FLAGS },
63     { "fps",          "set video rate", OFFSET(rate),  AV_OPT_TYPE_VIDEO_RATE, { .str = "25" },             1, 1000,     FLAGS },
64     { "rate",         "set video rate", OFFSET(rate),  AV_OPT_TYPE_VIDEO_RATE, { .str = "25" },             1, 1000,     FLAGS },
65     { "r",            "set video rate", OFFSET(rate),  AV_OPT_TYPE_VIDEO_RATE, { .str = "25" },             1, 1000,     FLAGS },
66     { "bar_h",   "set bargraph height", OFFSET(bar_h),        AV_OPT_TYPE_INT, { .i64 = -1 },              -1, INT_MAX,  FLAGS },
67     { "axis_h",      "set axis height", OFFSET(axis_h),       AV_OPT_TYPE_INT, { .i64 = -1 },              -1, INT_MAX,  FLAGS },
68     { "sono_h",  "set sonogram height", OFFSET(sono_h),       AV_OPT_TYPE_INT, { .i64 = -1 },              -1, INT_MAX,  FLAGS },
69     { "fullhd",      "set fullhd size", OFFSET(fullhd),      AV_OPT_TYPE_BOOL, { .i64 = 1 },                0, 1,        FLAGS },
70     { "sono_v",  "set sonogram volume", OFFSET(sono_v),    AV_OPT_TYPE_STRING, { .str = "16" },      0, 0, FLAGS },
71     { "volume",  "set sonogram volume", OFFSET(sono_v),    AV_OPT_TYPE_STRING, { .str = "16" },      0, 0, FLAGS },
72     { "bar_v",   "set bargraph volume", OFFSET(bar_v),     AV_OPT_TYPE_STRING, { .str = "sono_v" },  0, 0, FLAGS },
73     { "volume2", "set bargraph volume", OFFSET(bar_v),     AV_OPT_TYPE_STRING, { .str = "sono_v" },  0, 0, FLAGS },
74     { "sono_g",   "set sonogram gamma", OFFSET(sono_g),     AV_OPT_TYPE_FLOAT, { .dbl = 3.0 },            1.0, 7.0,      FLAGS },
75     { "gamma",    "set sonogram gamma", OFFSET(sono_g),     AV_OPT_TYPE_FLOAT, { .dbl = 3.0 },            1.0, 7.0,      FLAGS },
76     { "bar_g",    "set bargraph gamma", OFFSET(bar_g),      AV_OPT_TYPE_FLOAT, { .dbl = 1.0 },            1.0, 7.0,      FLAGS },
77     { "gamma2",   "set bargraph gamma", OFFSET(bar_g),      AV_OPT_TYPE_FLOAT, { .dbl = 1.0 },            1.0, 7.0,      FLAGS },
78     { "bar_t",  "set bar transparency", OFFSET(bar_t),      AV_OPT_TYPE_FLOAT, { .dbl = 1.0 },            0.0, 1.0,      FLAGS },
79     { "timeclamp",     "set timeclamp", OFFSET(timeclamp), AV_OPT_TYPE_DOUBLE, { .dbl = 0.17 },         0.002, 1.0,      FLAGS },
80     { "tc",            "set timeclamp", OFFSET(timeclamp), AV_OPT_TYPE_DOUBLE, { .dbl = 0.17 },         0.002, 1.0,      FLAGS },
81     { "attack",      "set attack time", OFFSET(attack),    AV_OPT_TYPE_DOUBLE, { .dbl = 0 },              0.0, 1.0,      FLAGS },
82     { "basefreq", "set base frequency", OFFSET(basefreq),  AV_OPT_TYPE_DOUBLE, { .dbl = BASEFREQ },      10.0, 100000.0, FLAGS },
83     { "endfreq",   "set end frequency", OFFSET(endfreq),   AV_OPT_TYPE_DOUBLE, { .dbl = ENDFREQ },       10.0, 100000.0, FLAGS },
84     { "coeffclamp",   "set coeffclamp", OFFSET(coeffclamp), AV_OPT_TYPE_FLOAT, { .dbl = 1.0 },            0.1, 10.0,     FLAGS },
85     { "tlength",         "set tlength", OFFSET(tlength),   AV_OPT_TYPE_STRING, { .str = TLENGTH },   0, 0, FLAGS },
86     { "count",   "set transform count", OFFSET(count),        AV_OPT_TYPE_INT, { .i64 = 6 },                1, 30,       FLAGS },
87     { "fcount",  "set frequency count", OFFSET(fcount),       AV_OPT_TYPE_INT, { .i64 = 0 },                0, 10,       FLAGS },
88     { "fontfile", "set axis font file", OFFSET(fontfile),  AV_OPT_TYPE_STRING, { .str = NULL },      0, 0, FLAGS },
89     { "font",          "set axis font", OFFSET(font),      AV_OPT_TYPE_STRING, { .str = NULL },      0, 0, FLAGS },
90     { "fontcolor",    "set font color", OFFSET(fontcolor), AV_OPT_TYPE_STRING, { .str = FONTCOLOR }, 0, 0, FLAGS },
91     { "axisfile",     "set axis image", OFFSET(axisfile),  AV_OPT_TYPE_STRING, { .str = NULL },      0, 0, FLAGS },
92     { "axis",              "draw axis", OFFSET(axis),        AV_OPT_TYPE_BOOL, { .i64 = 1 },                0, 1,        FLAGS },
93     { "text",              "draw axis", OFFSET(axis),        AV_OPT_TYPE_BOOL, { .i64 = 1 },                0, 1,        FLAGS },
94     { "csp",         "set color space", OFFSET(csp),          AV_OPT_TYPE_INT, { .i64 = AVCOL_SPC_UNSPECIFIED }, 0, INT_MAX, FLAGS, "csp" },
95         { "unspecified", "unspecified", 0,                  AV_OPT_TYPE_CONST, { .i64 = AVCOL_SPC_UNSPECIFIED }, 0, 0, FLAGS, "csp" },
96         { "bt709",             "bt709", 0,                  AV_OPT_TYPE_CONST, { .i64 = AVCOL_SPC_BT709 },       0, 0, FLAGS, "csp" },
97         { "fcc",                 "fcc", 0,                  AV_OPT_TYPE_CONST, { .i64 = AVCOL_SPC_FCC },         0, 0, FLAGS, "csp" },
98         { "bt470bg",         "bt470bg", 0,                  AV_OPT_TYPE_CONST, { .i64 = AVCOL_SPC_BT470BG },     0, 0, FLAGS, "csp" },
99         { "smpte170m",     "smpte170m", 0,                  AV_OPT_TYPE_CONST, { .i64 = AVCOL_SPC_SMPTE170M },   0, 0, FLAGS, "csp" },
100         { "smpte240m",     "smpte240m", 0,                  AV_OPT_TYPE_CONST, { .i64 = AVCOL_SPC_SMPTE240M },   0, 0, FLAGS, "csp" },
101         { "bt2020ncl",     "bt2020ncl", 0,                  AV_OPT_TYPE_CONST, { .i64 = AVCOL_SPC_BT2020_NCL },  0, 0, FLAGS, "csp" },
102     { "cscheme",    "set color scheme", OFFSET(cscheme),   AV_OPT_TYPE_STRING, { .str = CSCHEME },   0, 0, FLAGS },
103     { NULL }
104 };
105 
106 AVFILTER_DEFINE_CLASS(showcqt);
107 
common_uninit(ShowCQTContext * s)108 static void common_uninit(ShowCQTContext *s)
109 {
110     int k;
111     int level = AV_LOG_DEBUG;
112     int64_t plot_time;
113 
114     if (s->fft_time)
115         av_log(s->ctx, level, "fft_time         = %16.3f s.\n", s->fft_time * 1e-6);
116     if (s->cqt_time)
117         av_log(s->ctx, level, "cqt_time         = %16.3f s.\n", s->cqt_time * 1e-6);
118     if (s->process_cqt_time)
119         av_log(s->ctx, level, "process_cqt_time = %16.3f s.\n", s->process_cqt_time * 1e-6);
120     if (s->update_sono_time)
121         av_log(s->ctx, level, "update_sono_time = %16.3f s.\n", s->update_sono_time * 1e-6);
122     if (s->alloc_time)
123         av_log(s->ctx, level, "alloc_time       = %16.3f s.\n", s->alloc_time * 1e-6);
124     if (s->bar_time)
125         av_log(s->ctx, level, "bar_time         = %16.3f s.\n", s->bar_time * 1e-6);
126     if (s->axis_time)
127         av_log(s->ctx, level, "axis_time        = %16.3f s.\n", s->axis_time * 1e-6);
128     if (s->sono_time)
129         av_log(s->ctx, level, "sono_time        = %16.3f s.\n", s->sono_time * 1e-6);
130 
131     plot_time = s->fft_time + s->cqt_time + s->process_cqt_time + s->update_sono_time
132               + s->alloc_time + s->bar_time + s->axis_time + s->sono_time;
133     if (plot_time)
134         av_log(s->ctx, level, "plot_time        = %16.3f s.\n", plot_time * 1e-6);
135 
136     s->fft_time = s->cqt_time = s->process_cqt_time = s->update_sono_time
137                 = s->alloc_time = s->bar_time = s->axis_time = s->sono_time = 0;
138     /* axis_frame may be non reference counted frame */
139     if (s->axis_frame && !s->axis_frame->buf[0]) {
140         av_freep(s->axis_frame->data);
141         for (k = 0; k < 4; k++)
142             s->axis_frame->data[k] = NULL;
143     }
144 
145     av_frame_free(&s->axis_frame);
146     av_frame_free(&s->sono_frame);
147     av_tx_uninit(&s->fft_ctx);
148     if (s->coeffs)
149         for (k = 0; k < s->cqt_len; k++)
150             av_freep(&s->coeffs[k].val);
151     av_freep(&s->coeffs);
152     av_freep(&s->fft_data);
153     av_freep(&s->fft_input);
154     av_freep(&s->fft_result);
155     av_freep(&s->cqt_result);
156     av_freep(&s->attack_data);
157     av_freep(&s->c_buf);
158     av_freep(&s->h_buf);
159     av_freep(&s->rcp_h_buf);
160     av_freep(&s->freq);
161     av_freep(&s->sono_v_buf);
162     av_freep(&s->bar_v_buf);
163 }
164 
create_freq_table(double base,double end,int n)165 static double *create_freq_table(double base, double end, int n)
166 {
167     double log_base, log_end;
168     double rcp_n = 1.0 / n;
169     double *freq;
170     int x;
171 
172     freq = av_malloc_array(n, sizeof(*freq));
173     if (!freq)
174         return NULL;
175 
176     log_base = log(base);
177     log_end  = log(end);
178     for (x = 0; x < n; x++) {
179         double log_freq = log_base + (x + 0.5) * (log_end - log_base) * rcp_n;
180         freq[x] = exp(log_freq);
181     }
182     return freq;
183 }
184 
clip_with_log(void * log_ctx,const char * name,double val,double min,double max,double nan_replace,int idx)185 static double clip_with_log(void *log_ctx, const char *name,
186                             double val, double min, double max,
187                             double nan_replace, int idx)
188 {
189     int level = AV_LOG_WARNING;
190     if (isnan(val)) {
191         av_log(log_ctx, level, "[%d] %s is nan, setting it to %g.\n",
192                idx, name, nan_replace);
193         val = nan_replace;
194     } else if (val < min) {
195         av_log(log_ctx, level, "[%d] %s is too low (%g), setting it to %g.\n",
196                idx, name, val, min);
197         val = min;
198     } else if (val > max) {
199         av_log(log_ctx, level, "[%d] %s it too high (%g), setting it to %g.\n",
200                idx, name, val, max);
201         val = max;
202     }
203     return val;
204 }
205 
a_weighting(void * p,double f)206 static double a_weighting(void *p, double f)
207 {
208     double ret = 12200.0*12200.0 * (f*f*f*f);
209     ret /= (f*f + 20.6*20.6) * (f*f + 12200.0*12200.0) *
210            sqrt((f*f + 107.7*107.7) * (f*f + 737.9*737.9));
211     return ret;
212 }
213 
b_weighting(void * p,double f)214 static double b_weighting(void *p, double f)
215 {
216     double ret = 12200.0*12200.0 * (f*f*f);
217     ret /= (f*f + 20.6*20.6) * (f*f + 12200.0*12200.0) * sqrt(f*f + 158.5*158.5);
218     return ret;
219 }
220 
c_weighting(void * p,double f)221 static double c_weighting(void *p, double f)
222 {
223     double ret = 12200.0*12200.0 * (f*f);
224     ret /= (f*f + 20.6*20.6) * (f*f + 12200.0*12200.0);
225     return ret;
226 }
227 
init_volume(ShowCQTContext * s)228 static int init_volume(ShowCQTContext *s)
229 {
230     const char *func_names[] = { "a_weighting", "b_weighting", "c_weighting", NULL };
231     const char *sono_names[] = { "timeclamp", "tc", "frequency", "freq", "f", "bar_v", NULL };
232     const char *bar_names[] = { "timeclamp", "tc", "frequency", "freq", "f", "sono_v", NULL };
233     double (*funcs[])(void *, double) = { a_weighting, b_weighting, c_weighting };
234     AVExpr *sono = NULL, *bar = NULL;
235     int x, ret = AVERROR(ENOMEM);
236 
237     s->sono_v_buf = av_malloc_array(s->cqt_len, sizeof(*s->sono_v_buf));
238     s->bar_v_buf = av_malloc_array(s->cqt_len, sizeof(*s->bar_v_buf));
239     if (!s->sono_v_buf || !s->bar_v_buf)
240         goto error;
241 
242     if ((ret = av_expr_parse(&sono, s->sono_v, sono_names, func_names, funcs, NULL, NULL, 0, s->ctx)) < 0)
243         goto error;
244 
245     if ((ret = av_expr_parse(&bar, s->bar_v, bar_names, func_names, funcs, NULL, NULL, 0, s->ctx)) < 0)
246         goto error;
247 
248     for (x = 0; x < s->cqt_len; x++) {
249         double vars[] = { s->timeclamp, s->timeclamp, s->freq[x], s->freq[x], s->freq[x], 0.0 };
250         double vol = clip_with_log(s->ctx, "sono_v", av_expr_eval(sono, vars, NULL), 0.0, VOLUME_MAX, 0.0, x);
251         vars[5] = vol;
252         vol = clip_with_log(s->ctx, "bar_v", av_expr_eval(bar, vars, NULL), 0.0, VOLUME_MAX, 0.0, x);
253         s->bar_v_buf[x] = vol * vol;
254         vars[5] = vol;
255         vol = clip_with_log(s->ctx, "sono_v", av_expr_eval(sono, vars, NULL), 0.0, VOLUME_MAX, 0.0, x);
256         s->sono_v_buf[x] = vol * vol;
257     }
258     av_expr_free(sono);
259     av_expr_free(bar);
260     return 0;
261 
262 error:
263     av_freep(&s->sono_v_buf);
264     av_freep(&s->bar_v_buf);
265     av_expr_free(sono);
266     av_expr_free(bar);
267     return ret;
268 }
269 
cqt_calc(AVComplexFloat * dst,const AVComplexFloat * src,const Coeffs * coeffs,int len,int fft_len)270 static void cqt_calc(AVComplexFloat *dst, const AVComplexFloat *src, const Coeffs *coeffs,
271                      int len, int fft_len)
272 {
273     int k, x, i, j;
274     for (k = 0; k < len; k++) {
275         AVComplexFloat l, r, a = {0,0}, b = {0,0};
276 
277         for (x = 0; x < coeffs[k].len; x++) {
278             float u = coeffs[k].val[x];
279             i = coeffs[k].start + x;
280             j = fft_len - i;
281             a.re += u * src[i].re;
282             a.im += u * src[i].im;
283             b.re += u * src[j].re;
284             b.im += u * src[j].im;
285         }
286 
287         /* separate left and right, (and multiply by 2.0) */
288         l.re = a.re + b.re;
289         l.im = a.im - b.im;
290         r.re = b.im + a.im;
291         r.im = b.re - a.re;
292         dst[k].re = l.re * l.re + l.im * l.im;
293         dst[k].im = r.re * r.re + r.im * r.im;
294     }
295 }
296 
init_cqt(ShowCQTContext * s)297 static int init_cqt(ShowCQTContext *s)
298 {
299     const char *var_names[] = { "timeclamp", "tc", "frequency", "freq", "f", NULL };
300     AVExpr *expr = NULL;
301     int rate = s->ctx->inputs[0]->sample_rate;
302     int nb_cqt_coeffs = 0;
303     int k, x, ret;
304 
305     if ((ret = av_expr_parse(&expr, s->tlength, var_names, NULL, NULL, NULL, NULL, 0, s->ctx)) < 0)
306         goto error;
307 
308     ret = AVERROR(ENOMEM);
309     if (!(s->coeffs = av_calloc(s->cqt_len, sizeof(*s->coeffs))))
310         goto error;
311 
312     for (k = 0; k < s->cqt_len; k++) {
313         double vars[] = { s->timeclamp, s->timeclamp, s->freq[k], s->freq[k], s->freq[k] };
314         double flen, center, tlength;
315         int start, end, m = k;
316 
317         if (s->freq[k] > 0.5 * rate)
318             continue;
319         tlength = clip_with_log(s->ctx, "tlength", av_expr_eval(expr, vars, NULL),
320                                 TLENGTH_MIN, s->timeclamp, s->timeclamp, k);
321 
322         flen = 8.0 * s->fft_len / (tlength * rate);
323         center = s->freq[k] * s->fft_len / rate;
324         start = FFMAX(0, ceil(center - 0.5 * flen));
325         end = FFMIN(s->fft_len, floor(center + 0.5 * flen));
326 
327         s->coeffs[m].start = start & ~(s->cqt_align - 1);
328         s->coeffs[m].len = (end | (s->cqt_align - 1)) + 1 - s->coeffs[m].start;
329         nb_cqt_coeffs += s->coeffs[m].len;
330         if (!(s->coeffs[m].val = av_calloc(s->coeffs[m].len, sizeof(*s->coeffs[m].val))))
331             goto error;
332 
333         for (x = start; x <= end; x++) {
334             int sign = (x & 1) ? (-1) : 1;
335             double y = 2.0 * M_PI * (x - center) * (1.0 / flen);
336             /* nuttall window */
337             double w = 0.355768 + 0.487396 * cos(y) + 0.144232 * cos(2*y) + 0.012604 * cos(3*y);
338             w *= sign * (1.0 / s->fft_len);
339             s->coeffs[m].val[x - s->coeffs[m].start] = w;
340         }
341 
342         if (s->permute_coeffs)
343             s->permute_coeffs(s->coeffs[m].val, s->coeffs[m].len);
344     }
345 
346     av_expr_free(expr);
347     av_log(s->ctx, AV_LOG_INFO, "nb_cqt_coeffs = %d.\n", nb_cqt_coeffs);
348     return 0;
349 
350 error:
351     av_expr_free(expr);
352     if (s->coeffs)
353         for (k = 0; k < s->cqt_len; k++)
354             av_freep(&s->coeffs[k].val);
355     av_freep(&s->coeffs);
356     return ret;
357 }
358 
alloc_frame_empty(enum AVPixelFormat format,int w,int h)359 static AVFrame *alloc_frame_empty(enum AVPixelFormat format, int w, int h)
360 {
361     AVFrame *out;
362     out = av_frame_alloc();
363     if (!out)
364         return NULL;
365     out->format = format;
366     out->width = w;
367     out->height = h;
368     if (av_frame_get_buffer(out, 0) < 0) {
369         av_frame_free(&out);
370         return NULL;
371     }
372     if (format == AV_PIX_FMT_RGB24 || format == AV_PIX_FMT_RGBA) {
373         memset(out->data[0], 0, out->linesize[0] * h);
374     } else {
375         int hh = (format == AV_PIX_FMT_YUV420P || format == AV_PIX_FMT_YUVA420P) ? h / 2 : h;
376         memset(out->data[0], 16, out->linesize[0] * h);
377         memset(out->data[1], 128, out->linesize[1] * hh);
378         memset(out->data[2], 128, out->linesize[2] * hh);
379         if (out->data[3])
380             memset(out->data[3], 0, out->linesize[3] * h);
381     }
382     return out;
383 }
384 
convert_axis_pixel_format(enum AVPixelFormat format)385 static enum AVPixelFormat convert_axis_pixel_format(enum AVPixelFormat format)
386 {
387     switch (format) {
388         case AV_PIX_FMT_RGB24:   format = AV_PIX_FMT_RGBA; break;
389         case AV_PIX_FMT_YUV444P:
390         case AV_PIX_FMT_YUV422P:
391         case AV_PIX_FMT_YUV420P: format = AV_PIX_FMT_YUVA444P; break;
392     }
393     return format;
394 }
395 
init_axis_empty(ShowCQTContext * s)396 static int init_axis_empty(ShowCQTContext *s)
397 {
398     if (!(s->axis_frame = alloc_frame_empty(convert_axis_pixel_format(s->format), s->width, s->axis_h)))
399         return AVERROR(ENOMEM);
400     return 0;
401 }
402 
init_axis_from_file(ShowCQTContext * s)403 static int init_axis_from_file(ShowCQTContext *s)
404 {
405     uint8_t *tmp_data[4] = { NULL };
406     int tmp_linesize[4];
407     enum AVPixelFormat tmp_format;
408     int tmp_w, tmp_h, ret;
409 
410     if ((ret = ff_load_image(tmp_data, tmp_linesize, &tmp_w, &tmp_h, &tmp_format,
411                              s->axisfile, s->ctx)) < 0)
412         goto error;
413 
414     ret = AVERROR(ENOMEM);
415     if (!(s->axis_frame = av_frame_alloc()))
416         goto error;
417 
418     if ((ret = ff_scale_image(s->axis_frame->data, s->axis_frame->linesize, s->width, s->axis_h,
419                               convert_axis_pixel_format(s->format), tmp_data, tmp_linesize, tmp_w, tmp_h,
420                               tmp_format, s->ctx)) < 0)
421         goto error;
422 
423     s->axis_frame->width = s->width;
424     s->axis_frame->height = s->axis_h;
425     s->axis_frame->format = convert_axis_pixel_format(s->format);
426     av_freep(tmp_data);
427     return 0;
428 
429 error:
430     av_frame_free(&s->axis_frame);
431     av_freep(tmp_data);
432     return ret;
433 }
434 
midi(void * p,double f)435 static double midi(void *p, double f)
436 {
437     return log2(f/440.0) * 12.0 + 69.0;
438 }
439 
r_func(void * p,double x)440 static double r_func(void *p, double x)
441 {
442     x = av_clipd(x, 0.0, 1.0);
443     return lrint(x*255.0) << 16;
444 }
445 
g_func(void * p,double x)446 static double g_func(void *p, double x)
447 {
448     x = av_clipd(x, 0.0, 1.0);
449     return lrint(x*255.0) << 8;
450 }
451 
b_func(void * p,double x)452 static double b_func(void *p, double x)
453 {
454     x = av_clipd(x, 0.0, 1.0);
455     return lrint(x*255.0);
456 }
457 
init_axis_color(ShowCQTContext * s,AVFrame * tmp,int half)458 static int init_axis_color(ShowCQTContext *s, AVFrame *tmp, int half)
459 {
460     const char *var_names[] = { "timeclamp", "tc", "frequency", "freq", "f", NULL };
461     const char *func_names[] = { "midi", "r", "g", "b", NULL };
462     double (*funcs[])(void *, double) = { midi, r_func, g_func, b_func };
463     AVExpr *expr = NULL;
464     double *freq = NULL;
465     int x, xs, y, ret;
466     int width = half ? 1920/2 : 1920, height = half ? 16 : 32;
467     int step = half ? 2 : 1;
468 
469     if (s->basefreq != (double) BASEFREQ || s->endfreq != (double) ENDFREQ) {
470         av_log(s->ctx, AV_LOG_WARNING, "font axis rendering is not implemented in non-default frequency range,"
471                " please use axisfile option instead.\n");
472         return AVERROR(EINVAL);
473     }
474 
475     if (s->cqt_len == 1920)
476         freq = s->freq;
477     else if (!(freq = create_freq_table(s->basefreq, s->endfreq, 1920)))
478         return AVERROR(ENOMEM);
479 
480     if ((ret = av_expr_parse(&expr, s->fontcolor, var_names, func_names, funcs, NULL, NULL, 0, s->ctx)) < 0) {
481         if (freq != s->freq)
482             av_freep(&freq);
483         return ret;
484     }
485 
486     for (x = 0, xs = 0; x < width; x++, xs += step) {
487         double vars[] = { s->timeclamp, s->timeclamp, freq[xs], freq[xs], freq[xs] };
488         int color = (int) av_expr_eval(expr, vars, NULL);
489         uint8_t r = (color >> 16) & 0xFF, g = (color >> 8) & 0xFF, b = color & 0xFF;
490         uint8_t *data = tmp->data[0];
491         int linesize = tmp->linesize[0];
492         for (y = 0; y < height; y++) {
493             data[linesize * y + 4 * x] = r;
494             data[linesize * y + 4 * x + 1] = g;
495             data[linesize * y + 4 * x + 2] = b;
496         }
497     }
498 
499     av_expr_free(expr);
500     if (freq != s->freq)
501         av_freep(&freq);
502     return 0;
503 }
504 
render_freetype(ShowCQTContext * s,AVFrame * tmp,char * fontfile)505 static int render_freetype(ShowCQTContext *s, AVFrame *tmp, char *fontfile)
506 {
507 #if CONFIG_LIBFREETYPE
508     const char *str = "EF G A BC D ";
509     uint8_t *data = tmp->data[0];
510     int linesize = tmp->linesize[0];
511     FT_Library lib = NULL;
512     FT_Face face = NULL;
513     int font_width = 16, font_height = 32;
514     int font_repeat = font_width * 12;
515     int linear_hori_advance = font_width * 65536;
516     int non_monospace_warning = 0;
517     int x;
518 
519     if (!fontfile)
520         return AVERROR(EINVAL);
521 
522     if (FT_Init_FreeType(&lib))
523         goto fail;
524 
525     if (FT_New_Face(lib, fontfile, 0, &face))
526         goto fail;
527 
528     if (FT_Set_Char_Size(face, 16*64, 0, 0, 0))
529         goto fail;
530 
531     if (FT_Load_Char(face, 'A', FT_LOAD_RENDER))
532         goto fail;
533 
534     if (FT_Set_Char_Size(face, 16*64 * linear_hori_advance / face->glyph->linearHoriAdvance, 0, 0, 0))
535         goto fail;
536 
537     for (x = 0; x < 12; x++) {
538         int sx, sy, rx, bx, by, dx, dy;
539 
540         if (str[x] == ' ')
541             continue;
542 
543         if (FT_Load_Char(face, str[x], FT_LOAD_RENDER))
544             goto fail;
545 
546         if (face->glyph->advance.x != font_width*64 && !non_monospace_warning) {
547             av_log(s->ctx, AV_LOG_WARNING, "font is not monospace.\n");
548             non_monospace_warning = 1;
549         }
550 
551         sy = font_height - 8 - face->glyph->bitmap_top;
552         for (rx = 0; rx < 10; rx++) {
553             sx = rx * font_repeat + x * font_width + face->glyph->bitmap_left;
554             for (by = 0; by < face->glyph->bitmap.rows; by++) {
555                 dy = by + sy;
556                 if (dy < 0)
557                     continue;
558                 if (dy >= font_height)
559                     break;
560 
561                 for (bx = 0; bx < face->glyph->bitmap.width; bx++) {
562                     dx = bx + sx;
563                     if (dx < 0)
564                         continue;
565                     if (dx >= 1920)
566                         break;
567                     data[dy*linesize+4*dx+3] = face->glyph->bitmap.buffer[by*face->glyph->bitmap.width+bx];
568                 }
569             }
570         }
571     }
572 
573     FT_Done_Face(face);
574     FT_Done_FreeType(lib);
575     return 0;
576 
577 fail:
578     av_log(s->ctx, AV_LOG_WARNING, "error while loading freetype font.\n");
579     FT_Done_Face(face);
580     FT_Done_FreeType(lib);
581     return AVERROR(EINVAL);
582 #else
583     if (fontfile)
584         av_log(s->ctx, AV_LOG_WARNING, "freetype is not available, ignoring fontfile option.\n");
585     return AVERROR(EINVAL);
586 #endif
587 }
588 
render_fontconfig(ShowCQTContext * s,AVFrame * tmp,char * font)589 static int render_fontconfig(ShowCQTContext *s, AVFrame *tmp, char* font)
590 {
591 #if CONFIG_LIBFONTCONFIG
592     FcConfig *fontconfig;
593     FcPattern *pat, *best;
594     FcResult result = FcResultMatch;
595     char *filename;
596     int i, ret;
597 
598     if (!font)
599         return AVERROR(EINVAL);
600 
601     for (i = 0; font[i]; i++) {
602         if (font[i] == '|')
603             font[i] = ':';
604     }
605 
606     if (!(fontconfig = FcInitLoadConfigAndFonts())) {
607         av_log(s->ctx, AV_LOG_ERROR, "impossible to init fontconfig.\n");
608         return AVERROR_UNKNOWN;
609     }
610 
611     if (!(pat = FcNameParse((uint8_t *)font))) {
612         av_log(s->ctx, AV_LOG_ERROR, "could not parse fontconfig pat.\n");
613         FcConfigDestroy(fontconfig);
614         return AVERROR(EINVAL);
615     }
616 
617     FcDefaultSubstitute(pat);
618 
619     if (!FcConfigSubstitute(fontconfig, pat, FcMatchPattern)) {
620         av_log(s->ctx, AV_LOG_ERROR, "could not substitue fontconfig options.\n");
621         FcPatternDestroy(pat);
622         FcConfigDestroy(fontconfig);
623         return AVERROR(ENOMEM);
624     }
625 
626     best = FcFontMatch(fontconfig, pat, &result);
627     FcPatternDestroy(pat);
628 
629     ret = AVERROR(EINVAL);
630     if (!best || result != FcResultMatch) {
631         av_log(s->ctx, AV_LOG_ERROR, "cannot find a valid font for %s.\n", font);
632         goto fail;
633     }
634 
635     if (FcPatternGetString(best, FC_FILE, 0, (FcChar8 **)&filename) != FcResultMatch) {
636         av_log(s->ctx, AV_LOG_ERROR, "no file path for %s\n", font);
637         goto fail;
638     }
639 
640     ret = render_freetype(s, tmp, filename);
641 
642 fail:
643     FcPatternDestroy(best);
644     FcConfigDestroy(fontconfig);
645     return ret;
646 #else
647     if (font)
648         av_log(s->ctx, AV_LOG_WARNING, "fontconfig is not available, ignoring font option.\n");
649     return AVERROR(EINVAL);
650 #endif
651 }
652 
render_default_font(AVFrame * tmp)653 static int render_default_font(AVFrame *tmp)
654 {
655     const char *str = "EF G A BC D ";
656     int x, u, v, mask;
657     uint8_t *data = tmp->data[0];
658     int linesize = tmp->linesize[0];
659     int width = 1920/2, height = 16;
660 
661     for (x = 0; x < width; x += width/10) {
662         uint8_t *startptr = data + 4 * x;
663         for (u = 0; u < 12; u++) {
664             for (v = 0; v < height; v++) {
665                 uint8_t *p = startptr + v * linesize + height/2 * 4 * u;
666                 for (mask = 0x80; mask; mask >>= 1, p += 4) {
667                     if (mask & avpriv_vga16_font[str[u] * 16 + v])
668                         p[3] = 255;
669                     else
670                         p[3] = 0;
671                 }
672             }
673         }
674     }
675 
676     return 0;
677 }
678 
init_axis_from_font(ShowCQTContext * s)679 static int init_axis_from_font(ShowCQTContext *s)
680 {
681     AVFrame *tmp = NULL;
682     int ret = AVERROR(ENOMEM);
683     int width = 1920, height = 32;
684     int default_font = 0;
685 
686     if (!(tmp = alloc_frame_empty(AV_PIX_FMT_RGBA, width, height)))
687         goto fail;
688 
689     if (!(s->axis_frame = av_frame_alloc()))
690         goto fail;
691 
692     if (render_freetype(s, tmp, s->fontfile) < 0 &&
693         render_fontconfig(s, tmp, s->font) < 0 &&
694         (default_font = 1, ret = render_default_font(tmp)) < 0)
695         goto fail;
696 
697     if (default_font)
698         width /= 2, height /= 2;
699 
700     if ((ret = init_axis_color(s, tmp, default_font)) < 0)
701         goto fail;
702 
703     if ((ret = ff_scale_image(s->axis_frame->data, s->axis_frame->linesize, s->width, s->axis_h,
704                               convert_axis_pixel_format(s->format), tmp->data, tmp->linesize,
705                               width, height, AV_PIX_FMT_RGBA, s->ctx)) < 0)
706         goto fail;
707 
708     av_frame_free(&tmp);
709     s->axis_frame->width = s->width;
710     s->axis_frame->height = s->axis_h;
711     s->axis_frame->format = convert_axis_pixel_format(s->format);
712     return 0;
713 
714 fail:
715     av_frame_free(&tmp);
716     av_frame_free(&s->axis_frame);
717     return ret;
718 }
719 
calculate_gamma(float v,float g)720 static float calculate_gamma(float v, float g)
721 {
722     if (g == 1.0f)
723         return v;
724     if (g == 2.0f)
725         return sqrtf(v);
726     if (g == 3.0f)
727         return cbrtf(v);
728     if (g == 4.0f)
729         return sqrtf(sqrtf(v));
730     return expf(logf(v) / g);
731 }
732 
rgb_from_cqt(ColorFloat * c,const AVComplexFloat * v,float g,int len,float cscheme[6])733 static void rgb_from_cqt(ColorFloat *c, const AVComplexFloat *v, float g, int len, float cscheme[6])
734 {
735     int x;
736     for (x = 0; x < len; x++) {
737         c[x].rgb.r = 255.0f * calculate_gamma(FFMIN(1.0f, cscheme[0] * v[x].re + cscheme[3] * v[x].im), g);
738         c[x].rgb.g = 255.0f * calculate_gamma(FFMIN(1.0f, cscheme[1] * v[x].re + cscheme[4] * v[x].im), g);
739         c[x].rgb.b = 255.0f * calculate_gamma(FFMIN(1.0f, cscheme[2] * v[x].re + cscheme[5] * v[x].im), g);
740     }
741 }
742 
yuv_from_cqt(ColorFloat * c,const AVComplexFloat * v,float gamma,int len,float cm[3][3],float cscheme[6])743 static void yuv_from_cqt(ColorFloat *c, const AVComplexFloat *v, float gamma, int len, float cm[3][3], float cscheme[6])
744 {
745     int x;
746     for (x = 0; x < len; x++) {
747         float r, g, b;
748         r = calculate_gamma(FFMIN(1.0f, cscheme[0] * v[x].re + cscheme[3] * v[x].im), gamma);
749         g = calculate_gamma(FFMIN(1.0f, cscheme[1] * v[x].re + cscheme[4] * v[x].im), gamma);
750         b = calculate_gamma(FFMIN(1.0f, cscheme[2] * v[x].re + cscheme[5] * v[x].im), gamma);
751         c[x].yuv.y = cm[0][0] * r + cm[0][1] * g + cm[0][2] * b;
752         c[x].yuv.u = cm[1][0] * r + cm[1][1] * g + cm[1][2] * b;
753         c[x].yuv.v = cm[2][0] * r + cm[2][1] * g + cm[2][2] * b;
754     }
755 }
756 
draw_bar_rgb(AVFrame * out,const float * h,const float * rcp_h,const ColorFloat * c,int bar_h,float bar_t)757 static void draw_bar_rgb(AVFrame *out, const float *h, const float *rcp_h,
758                          const ColorFloat *c, int bar_h, float bar_t)
759 {
760     int x, y, w = out->width;
761     float mul, ht, rcp_bar_h = 1.0f / bar_h, rcp_bar_t = 1.0f / bar_t;
762     uint8_t *v = out->data[0], *lp;
763     int ls = out->linesize[0];
764 
765     for (y = 0; y < bar_h; y++) {
766         ht = (bar_h - y) * rcp_bar_h;
767         lp = v + y * ls;
768         for (x = 0; x < w; x++) {
769             if (h[x] <= ht) {
770                 *lp++ = 0;
771                 *lp++ = 0;
772                 *lp++ = 0;
773             } else {
774                 mul = (h[x] - ht) * rcp_h[x];
775                 mul = (mul < bar_t) ? (mul * rcp_bar_t) : 1.0f;
776                 *lp++ = lrintf(mul * c[x].rgb.r);
777                 *lp++ = lrintf(mul * c[x].rgb.g);
778                 *lp++ = lrintf(mul * c[x].rgb.b);
779             }
780         }
781     }
782 }
783 
784 #define DRAW_BAR_WITH_CHROMA(x) \
785 do { \
786     if (h[x] <= ht) { \
787         *lpy++ = 16; \
788         *lpu++ = 128; \
789         *lpv++ = 128; \
790     } else { \
791         mul = (h[x] - ht) * rcp_h[x]; \
792         mul = (mul < bar_t) ? (mul * rcp_bar_t) : 1.0f; \
793         *lpy++ = lrintf(mul * c[x].yuv.y + 16.0f); \
794         *lpu++ = lrintf(mul * c[x].yuv.u + 128.0f); \
795         *lpv++ = lrintf(mul * c[x].yuv.v + 128.0f); \
796     } \
797 } while (0)
798 
799 #define DRAW_BAR_WITHOUT_CHROMA(x) \
800 do { \
801     if (h[x] <= ht) { \
802         *lpy++ = 16; \
803     } else { \
804         mul = (h[x] - ht) * rcp_h[x]; \
805         mul = (mul < bar_t) ? (mul * rcp_bar_t) : 1.0f; \
806         *lpy++ = lrintf(mul * c[x].yuv.y + 16.0f); \
807     } \
808 } while (0)
809 
draw_bar_yuv(AVFrame * out,const float * h,const float * rcp_h,const ColorFloat * c,int bar_h,float bar_t)810 static void draw_bar_yuv(AVFrame *out, const float *h, const float *rcp_h,
811                          const ColorFloat *c, int bar_h, float bar_t)
812 {
813     int x, y, yh, w = out->width;
814     float mul, ht, rcp_bar_h = 1.0f / bar_h, rcp_bar_t = 1.0f / bar_t;
815     uint8_t *vy = out->data[0], *vu = out->data[1], *vv = out->data[2];
816     uint8_t *lpy, *lpu, *lpv;
817     int lsy = out->linesize[0], lsu = out->linesize[1], lsv = out->linesize[2];
818     int fmt = out->format;
819 
820     for (y = 0; y < bar_h; y += 2) {
821         yh = (fmt == AV_PIX_FMT_YUV420P) ? y / 2 : y;
822         ht = (bar_h - y) * rcp_bar_h;
823         lpy = vy + y * lsy;
824         lpu = vu + yh * lsu;
825         lpv = vv + yh * lsv;
826         if (fmt == AV_PIX_FMT_YUV444P) {
827             for (x = 0; x < w; x += 2) {
828                 DRAW_BAR_WITH_CHROMA(x);
829                 DRAW_BAR_WITH_CHROMA(x+1);
830             }
831         } else {
832             for (x = 0; x < w; x += 2) {
833                 DRAW_BAR_WITH_CHROMA(x);
834                 DRAW_BAR_WITHOUT_CHROMA(x+1);
835             }
836         }
837 
838         ht = (bar_h - (y+1)) * rcp_bar_h;
839         lpy = vy + (y+1) * lsy;
840         lpu = vu + (y+1) * lsu;
841         lpv = vv + (y+1) * lsv;
842         if (fmt == AV_PIX_FMT_YUV444P) {
843             for (x = 0; x < w; x += 2) {
844                 DRAW_BAR_WITH_CHROMA(x);
845                 DRAW_BAR_WITH_CHROMA(x+1);
846             }
847         } else if (fmt == AV_PIX_FMT_YUV422P) {
848             for (x = 0; x < w; x += 2) {
849                 DRAW_BAR_WITH_CHROMA(x);
850                 DRAW_BAR_WITHOUT_CHROMA(x+1);
851             }
852         } else {
853             for (x = 0; x < w; x += 2) {
854                 DRAW_BAR_WITHOUT_CHROMA(x);
855                 DRAW_BAR_WITHOUT_CHROMA(x+1);
856             }
857         }
858     }
859 }
860 
draw_axis_rgb(AVFrame * out,AVFrame * axis,const ColorFloat * c,int off)861 static void draw_axis_rgb(AVFrame *out, AVFrame *axis, const ColorFloat *c, int off)
862 {
863     int x, y, w = axis->width, h = axis->height;
864     float a, rcp_255 = 1.0f / 255.0f;
865     uint8_t *lp, *lpa;
866 
867     for (y = 0; y < h; y++) {
868         lp = out->data[0] + (off + y) * out->linesize[0];
869         lpa = axis->data[0] + y * axis->linesize[0];
870         for (x = 0; x < w; x++) {
871             if (!lpa[3]) {
872                 *lp++ = lrintf(c[x].rgb.r);
873                 *lp++ = lrintf(c[x].rgb.g);
874                 *lp++ = lrintf(c[x].rgb.b);
875             } else if (lpa[3] == 255) {
876                 *lp++ = lpa[0];
877                 *lp++ = lpa[1];
878                 *lp++ = lpa[2];
879             } else {
880                 a = rcp_255 * lpa[3];
881                 *lp++ = lrintf(a * lpa[0] + (1.0f - a) * c[x].rgb.r);
882                 *lp++ = lrintf(a * lpa[1] + (1.0f - a) * c[x].rgb.g);
883                 *lp++ = lrintf(a * lpa[2] + (1.0f - a) * c[x].rgb.b);
884             }
885             lpa += 4;
886         }
887     }
888 }
889 
890 #define BLEND_WITH_CHROMA(c) \
891 do { \
892     if (!*lpaa) { \
893         *lpy = lrintf(c.yuv.y + 16.0f); \
894         *lpu = lrintf(c.yuv.u + 128.0f); \
895         *lpv = lrintf(c.yuv.v + 128.0f); \
896     } else if (255 == *lpaa) { \
897         *lpy = *lpay; \
898         *lpu = *lpau; \
899         *lpv = *lpav; \
900     } else { \
901         float a = (1.0f/255.0f) * (*lpaa); \
902         *lpy = lrintf(a * (*lpay) + (1.0f - a) * (c.yuv.y + 16.0f)); \
903         *lpu = lrintf(a * (*lpau) + (1.0f - a) * (c.yuv.u + 128.0f)); \
904         *lpv = lrintf(a * (*lpav) + (1.0f - a) * (c.yuv.v + 128.0f)); \
905     } \
906     lpy++; lpu++; lpv++; \
907     lpay++; lpau++; lpav++; lpaa++; \
908 } while (0)
909 
910 #define BLEND_WITHOUT_CHROMA(c, alpha_inc) \
911 do { \
912     if (!*lpaa) { \
913         *lpy = lrintf(c.yuv.y + 16.0f); \
914     } else if (255 == *lpaa) { \
915         *lpy = *lpay; \
916     } else { \
917         float a = (1.0f/255.0f) * (*lpaa); \
918         *lpy = lrintf(a * (*lpay) + (1.0f - a) * (c.yuv.y + 16.0f)); \
919     } \
920     lpy++; \
921     lpay++; lpaa += alpha_inc; \
922 } while (0)
923 
924 #define BLEND_CHROMA2(c) \
925 do { \
926     if (!lpaa[0] && !lpaa[1]) { \
927         *lpu = lrintf(c.yuv.u + 128.0f); \
928         *lpv = lrintf(c.yuv.v + 128.0f); \
929     } else if (255 == lpaa[0] && 255 == lpaa[1]) { \
930         *lpu = *lpau; *lpv = *lpav; \
931     } else { \
932         float a0 = (0.5f/255.0f) * lpaa[0]; \
933         float a1 = (0.5f/255.0f) * lpaa[1]; \
934         float b = 1.0f - a0 - a1; \
935         *lpu = lrintf(a0 * lpau[0] + a1 * lpau[1] + b * (c.yuv.u + 128.0f)); \
936         *lpv = lrintf(a0 * lpav[0] + a1 * lpav[1] + b * (c.yuv.v + 128.0f)); \
937     } \
938     lpau += 2; lpav += 2; lpaa++; lpu++; lpv++; \
939 } while (0)
940 
941 #define BLEND_CHROMA2x2(c) \
942 do { \
943     if (!lpaa[0] && !lpaa[1] && !lpaa[lsaa] && !lpaa[lsaa+1]) { \
944         *lpu = lrintf(c.yuv.u + 128.0f); \
945         *lpv = lrintf(c.yuv.v + 128.0f); \
946     } else if (255 == lpaa[0] && 255 == lpaa[1] && \
947                255 == lpaa[lsaa] && 255 == lpaa[lsaa+1]) { \
948         *lpu = *lpau; *lpv = *lpav; \
949     } else { \
950         float a0 = (0.25f/255.0f) * lpaa[0]; \
951         float a1 = (0.25f/255.0f) * lpaa[1]; \
952         float a2 = (0.25f/255.0f) * lpaa[lsaa]; \
953         float a3 = (0.25f/255.0f) * lpaa[lsaa+1]; \
954         float b = 1.0f - a0 - a1 - a2 - a3; \
955         *lpu = lrintf(a0 * lpau[0] + a1 * lpau[1] + a2 * lpau[lsau] + a3 * lpau[lsau+1] \
956                     + b * (c.yuv.u + 128.0f)); \
957         *lpv = lrintf(a0 * lpav[0] + a1 * lpav[1] + a2 * lpav[lsav] + a3 * lpav[lsav+1] \
958                     + b * (c.yuv.v + 128.0f)); \
959     } \
960     lpau += 2; lpav += 2; lpaa++; lpu++; lpv++; \
961 } while (0)
962 
draw_axis_yuv(AVFrame * out,AVFrame * axis,const ColorFloat * c,int off)963 static void draw_axis_yuv(AVFrame *out, AVFrame *axis, const ColorFloat *c, int off)
964 {
965     int fmt = out->format, x, y, yh, w = axis->width, h = axis->height;
966     int offh = (fmt == AV_PIX_FMT_YUV420P) ? off / 2 : off;
967     uint8_t *vy = out->data[0], *vu = out->data[1], *vv = out->data[2];
968     uint8_t *vay = axis->data[0], *vau = axis->data[1], *vav = axis->data[2], *vaa = axis->data[3];
969     int lsy = out->linesize[0], lsu = out->linesize[1], lsv = out->linesize[2];
970     int lsay = axis->linesize[0], lsau = axis->linesize[1], lsav = axis->linesize[2], lsaa = axis->linesize[3];
971     uint8_t *lpy, *lpu, *lpv, *lpay, *lpau, *lpav, *lpaa;
972 
973     for (y = 0; y < h; y += 2) {
974         yh = (fmt == AV_PIX_FMT_YUV420P) ? y / 2 : y;
975         lpy = vy + (off + y) * lsy;
976         lpu = vu + (offh + yh) * lsu;
977         lpv = vv + (offh + yh) * lsv;
978         lpay = vay + y * lsay;
979         lpau = vau + y * lsau;
980         lpav = vav + y * lsav;
981         lpaa = vaa + y * lsaa;
982         if (fmt == AV_PIX_FMT_YUV444P) {
983             for (x = 0; x < w; x += 2) {
984                 BLEND_WITH_CHROMA(c[x]);
985                 BLEND_WITH_CHROMA(c[x+1]);
986             }
987         } else if (fmt == AV_PIX_FMT_YUV422P) {
988             for (x = 0; x < w; x += 2) {
989                 BLEND_WITHOUT_CHROMA(c[x], 0);
990                 BLEND_CHROMA2(c[x]);
991                 BLEND_WITHOUT_CHROMA(c[x+1], 1);
992             }
993         } else {
994             for (x = 0; x < w; x += 2) {
995                 BLEND_WITHOUT_CHROMA(c[x], 0);
996                 BLEND_CHROMA2x2(c[x]);
997                 BLEND_WITHOUT_CHROMA(c[x+1], 1);
998             }
999         }
1000 
1001         lpy = vy + (off + y + 1) * lsy;
1002         lpu = vu + (off + y + 1) * lsu;
1003         lpv = vv + (off + y + 1) * lsv;
1004         lpay = vay + (y + 1) * lsay;
1005         lpau = vau + (y + 1) * lsau;
1006         lpav = vav + (y + 1) * lsav;
1007         lpaa = vaa + (y + 1) * lsaa;
1008         if (fmt == AV_PIX_FMT_YUV444P) {
1009             for (x = 0; x < w; x += 2) {
1010                 BLEND_WITH_CHROMA(c[x]);
1011                 BLEND_WITH_CHROMA(c[x+1]);
1012             }
1013         } else if (fmt == AV_PIX_FMT_YUV422P) {
1014             for (x = 0; x < w; x += 2) {
1015                 BLEND_WITHOUT_CHROMA(c[x], 0);
1016                 BLEND_CHROMA2(c[x]);
1017                 BLEND_WITHOUT_CHROMA(c[x+1], 1);
1018             }
1019         } else {
1020             for (x = 0; x < w; x += 2) {
1021                 BLEND_WITHOUT_CHROMA(c[x], 1);
1022                 BLEND_WITHOUT_CHROMA(c[x+1], 1);
1023             }
1024         }
1025     }
1026 }
1027 
draw_sono(AVFrame * out,AVFrame * sono,int off,int idx)1028 static void draw_sono(AVFrame *out, AVFrame *sono, int off, int idx)
1029 {
1030     int fmt = out->format, h = sono->height;
1031     int nb_planes = (fmt == AV_PIX_FMT_RGB24) ? 1 : 3;
1032     int offh = (fmt == AV_PIX_FMT_YUV420P) ? off / 2 : off;
1033     int inc = (fmt == AV_PIX_FMT_YUV420P) ? 2 : 1;
1034     int ls, i, y, yh;
1035 
1036     ls = FFMIN(out->linesize[0], sono->linesize[0]);
1037     for (y = 0; y < h; y++) {
1038         memcpy(out->data[0] + (off + y) * out->linesize[0],
1039                sono->data[0] + (idx + y) % h * sono->linesize[0], ls);
1040     }
1041 
1042     for (i = 1; i < nb_planes; i++) {
1043         ls = FFMIN(out->linesize[i], sono->linesize[i]);
1044         for (y = 0; y < h; y += inc) {
1045             yh = (fmt == AV_PIX_FMT_YUV420P) ? y / 2 : y;
1046             memcpy(out->data[i] + (offh + yh) * out->linesize[i],
1047                    sono->data[i] + (idx + y) % h * sono->linesize[i], ls);
1048         }
1049     }
1050 }
1051 
update_sono_rgb(AVFrame * sono,const ColorFloat * c,int idx)1052 static void update_sono_rgb(AVFrame *sono, const ColorFloat *c, int idx)
1053 {
1054     int x, w = sono->width;
1055     uint8_t *lp = sono->data[0] + idx * sono->linesize[0];
1056 
1057     for (x = 0; x < w; x++) {
1058         *lp++ = lrintf(c[x].rgb.r);
1059         *lp++ = lrintf(c[x].rgb.g);
1060         *lp++ = lrintf(c[x].rgb.b);
1061     }
1062 }
1063 
update_sono_yuv(AVFrame * sono,const ColorFloat * c,int idx)1064 static void update_sono_yuv(AVFrame *sono, const ColorFloat *c, int idx)
1065 {
1066     int x, fmt = sono->format, w = sono->width;
1067     uint8_t *lpy = sono->data[0] + idx * sono->linesize[0];
1068     uint8_t *lpu = sono->data[1] + idx * sono->linesize[1];
1069     uint8_t *lpv = sono->data[2] + idx * sono->linesize[2];
1070 
1071     for (x = 0; x < w; x += 2) {
1072         *lpy++ = lrintf(c[x].yuv.y + 16.0f);
1073         *lpu++ = lrintf(c[x].yuv.u + 128.0f);
1074         *lpv++ = lrintf(c[x].yuv.v + 128.0f);
1075         *lpy++ = lrintf(c[x+1].yuv.y + 16.0f);
1076         if (fmt == AV_PIX_FMT_YUV444P) {
1077             *lpu++ = lrintf(c[x+1].yuv.u + 128.0f);
1078             *lpv++ = lrintf(c[x+1].yuv.v + 128.0f);
1079         }
1080     }
1081 }
1082 
process_cqt(ShowCQTContext * s)1083 static void process_cqt(ShowCQTContext *s)
1084 {
1085     int x, i;
1086     if (!s->sono_count) {
1087         for (x = 0; x < s->cqt_len; x++) {
1088             s->h_buf[x] = s->bar_v_buf[x] * 0.5f * (s->cqt_result[x].re + s->cqt_result[x].im);
1089         }
1090         if (s->fcount > 1) {
1091             float rcp_fcount = 1.0f / s->fcount;
1092             for (x = 0; x < s->width; x++) {
1093                 float h = 0.0f;
1094                 for (i = 0; i < s->fcount; i++)
1095                     h += s->h_buf[s->fcount * x + i];
1096                 s->h_buf[x] = rcp_fcount * h;
1097             }
1098         }
1099         for (x = 0; x < s->width; x++) {
1100             s->h_buf[x] = calculate_gamma(s->h_buf[x], s->bar_g);
1101             s->rcp_h_buf[x] = 1.0f / (s->h_buf[x] + 0.0001f);
1102         }
1103     }
1104 
1105     for (x = 0; x < s->cqt_len; x++) {
1106         s->cqt_result[x].re *= s->sono_v_buf[x];
1107         s->cqt_result[x].im *= s->sono_v_buf[x];
1108     }
1109 
1110     if (s->fcount > 1) {
1111         float rcp_fcount = 1.0f / s->fcount;
1112         for (x = 0; x < s->width; x++) {
1113             AVComplexFloat result = {0.0f, 0.0f};
1114             for (i = 0; i < s->fcount; i++) {
1115                 result.re += s->cqt_result[s->fcount * x + i].re;
1116                 result.im += s->cqt_result[s->fcount * x + i].im;
1117             }
1118             s->cqt_result[x].re = rcp_fcount * result.re;
1119             s->cqt_result[x].im = rcp_fcount * result.im;
1120         }
1121     }
1122 
1123     if (s->format == AV_PIX_FMT_RGB24)
1124         rgb_from_cqt(s->c_buf, s->cqt_result, s->sono_g, s->width, s->cscheme_v);
1125     else
1126         yuv_from_cqt(s->c_buf, s->cqt_result, s->sono_g, s->width, s->cmatrix, s->cscheme_v);
1127 }
1128 
plot_cqt(AVFilterContext * ctx,AVFrame ** frameout)1129 static int plot_cqt(AVFilterContext *ctx, AVFrame **frameout)
1130 {
1131     AVFilterLink *outlink = ctx->outputs[0];
1132     ShowCQTContext *s = ctx->priv;
1133     int64_t last_time, cur_time;
1134 
1135 #define UPDATE_TIME(t) \
1136     cur_time = av_gettime_relative(); \
1137     t += cur_time - last_time; \
1138     last_time = cur_time
1139 
1140     last_time = av_gettime_relative();
1141 
1142     memcpy(s->fft_input, s->fft_data, s->fft_len * sizeof(*s->fft_data));
1143     if (s->attack_data) {
1144         int k;
1145         for (k = 0; k < s->remaining_fill_max; k++) {
1146             s->fft_input[s->fft_len/2+k].re *= s->attack_data[k];
1147             s->fft_input[s->fft_len/2+k].im *= s->attack_data[k];
1148         }
1149     }
1150 
1151     s->tx_fn(s->fft_ctx, s->fft_result, s->fft_input, sizeof(float));
1152     s->fft_result[s->fft_len] = s->fft_result[0];
1153     UPDATE_TIME(s->fft_time);
1154 
1155     s->cqt_calc(s->cqt_result, s->fft_result, s->coeffs, s->cqt_len, s->fft_len);
1156     UPDATE_TIME(s->cqt_time);
1157 
1158     process_cqt(s);
1159     UPDATE_TIME(s->process_cqt_time);
1160 
1161     if (s->sono_h) {
1162         s->update_sono(s->sono_frame, s->c_buf, s->sono_idx);
1163         UPDATE_TIME(s->update_sono_time);
1164     }
1165 
1166     if (!s->sono_count) {
1167         AVFrame *out = *frameout = ff_get_video_buffer(outlink, outlink->w, outlink->h);
1168         if (!out)
1169             return AVERROR(ENOMEM);
1170         out->sample_aspect_ratio = av_make_q(1, 1);
1171         out->color_range = AVCOL_RANGE_MPEG;
1172         out->colorspace = s->csp;
1173         UPDATE_TIME(s->alloc_time);
1174 
1175         if (s->bar_h) {
1176             s->draw_bar(out, s->h_buf, s->rcp_h_buf, s->c_buf, s->bar_h, s->bar_t);
1177             UPDATE_TIME(s->bar_time);
1178         }
1179 
1180         if (s->axis_h) {
1181             s->draw_axis(out, s->axis_frame, s->c_buf, s->bar_h);
1182             UPDATE_TIME(s->axis_time);
1183         }
1184 
1185         if (s->sono_h) {
1186             s->draw_sono(out, s->sono_frame, s->bar_h + s->axis_h, s->sono_idx);
1187             UPDATE_TIME(s->sono_time);
1188         }
1189         out->pts = s->next_pts;
1190         s->next_pts += PTS_STEP;
1191     }
1192     s->sono_count = (s->sono_count + 1) % s->count;
1193     if (s->sono_h)
1194         s->sono_idx = (s->sono_idx + s->sono_h - 1) % s->sono_h;
1195     return 0;
1196 }
1197 
init_colormatrix(ShowCQTContext * s)1198 static void init_colormatrix(ShowCQTContext *s)
1199 {
1200     double kr, kg, kb;
1201 
1202     /* from vf_colorspace.c */
1203     switch (s->csp) {
1204     default:
1205         av_log(s->ctx, AV_LOG_WARNING, "unsupported colorspace, setting it to unspecified.\n");
1206         s->csp = AVCOL_SPC_UNSPECIFIED;
1207     case AVCOL_SPC_UNSPECIFIED:
1208     case AVCOL_SPC_BT470BG:
1209     case AVCOL_SPC_SMPTE170M:
1210         kr = 0.299; kb = 0.114; break;
1211     case AVCOL_SPC_BT709:
1212         kr = 0.2126; kb = 0.0722; break;
1213     case AVCOL_SPC_FCC:
1214         kr = 0.30; kb = 0.11; break;
1215     case AVCOL_SPC_SMPTE240M:
1216         kr = 0.212; kb = 0.087; break;
1217     case AVCOL_SPC_BT2020_NCL:
1218         kr = 0.2627; kb = 0.0593; break;
1219     }
1220 
1221     kg = 1.0 - kr - kb;
1222     s->cmatrix[0][0] = 219.0 * kr;
1223     s->cmatrix[0][1] = 219.0 * kg;
1224     s->cmatrix[0][2] = 219.0 * kb;
1225     s->cmatrix[1][0] = -112.0 * kr / (1.0 - kb);
1226     s->cmatrix[1][1] = -112.0 * kg / (1.0 - kb);
1227     s->cmatrix[1][2] = 112.0;
1228     s->cmatrix[2][0] = 112.0;
1229     s->cmatrix[2][1] = -112.0 * kg / (1.0 - kr);
1230     s->cmatrix[2][2] = -112.0 * kb / (1.0 - kr);
1231 }
1232 
init_cscheme(ShowCQTContext * s)1233 static int init_cscheme(ShowCQTContext *s)
1234 {
1235     char tail[2];
1236     int k;
1237 
1238     if (sscanf(s->cscheme, " %f | %f | %f | %f | %f | %f %1s", &s->cscheme_v[0],
1239         &s->cscheme_v[1], &s->cscheme_v[2], &s->cscheme_v[3], &s->cscheme_v[4],
1240         &s->cscheme_v[5], tail) != 6)
1241         goto fail;
1242 
1243     for (k = 0; k < 6; k++)
1244         if (isnan(s->cscheme_v[k]) || s->cscheme_v[k] < 0.0f || s->cscheme_v[k] > 1.0f)
1245             goto fail;
1246 
1247     return 0;
1248 
1249 fail:
1250     av_log(s->ctx, AV_LOG_ERROR, "invalid cscheme.\n");
1251     return AVERROR(EINVAL);
1252 }
1253 
1254 /* main filter control */
init(AVFilterContext * ctx)1255 static av_cold int init(AVFilterContext *ctx)
1256 {
1257     ShowCQTContext *s = ctx->priv;
1258     s->ctx = ctx;
1259 
1260     if (!s->fullhd) {
1261         av_log(ctx, AV_LOG_WARNING, "fullhd option is deprecated, use size/s option instead.\n");
1262         if (s->width != 1920 || s->height != 1080) {
1263             av_log(ctx, AV_LOG_ERROR, "fullhd set to 0 but with custom dimension.\n");
1264             return AVERROR(EINVAL);
1265         }
1266         s->width /= 2;
1267         s->height /= 2;
1268         s->fullhd = 1;
1269     }
1270 
1271     if (s->axis_h < 0) {
1272         s->axis_h = s->width / 60;
1273         if (s->axis_h & 1)
1274             s->axis_h++;
1275         if (s->bar_h >= 0 && s->sono_h >= 0)
1276             s->axis_h = s->height - s->bar_h - s->sono_h;
1277         if (s->bar_h >= 0 && s->sono_h < 0)
1278             s->axis_h = FFMIN(s->axis_h, s->height - s->bar_h);
1279         if (s->bar_h < 0 && s->sono_h >= 0)
1280             s->axis_h = FFMIN(s->axis_h, s->height - s->sono_h);
1281     }
1282 
1283     if (s->bar_h < 0) {
1284         s->bar_h = (s->height - s->axis_h) / 2;
1285         if (s->bar_h & 1)
1286             s->bar_h--;
1287         if (s->sono_h >= 0)
1288             s->bar_h = s->height - s->sono_h - s->axis_h;
1289     }
1290 
1291     if (s->sono_h < 0)
1292         s->sono_h = s->height - s->axis_h - s->bar_h;
1293 
1294     if ((s->width & 1) || (s->height & 1) || (s->bar_h & 1) || (s->axis_h & 1) || (s->sono_h & 1) ||
1295         (s->bar_h < 0) || (s->axis_h < 0) || (s->sono_h < 0) || (s->bar_h > s->height) ||
1296         (s->axis_h > s->height) || (s->sono_h > s->height) || (s->bar_h + s->axis_h + s->sono_h != s->height)) {
1297         av_log(ctx, AV_LOG_ERROR, "invalid dimension.\n");
1298         return AVERROR(EINVAL);
1299     }
1300 
1301     if (!s->fcount) {
1302         do {
1303             s->fcount++;
1304         } while(s->fcount * s->width < 1920 && s->fcount < 10);
1305     }
1306 
1307     init_colormatrix(s);
1308 
1309     return init_cscheme(s);
1310 }
1311 
uninit(AVFilterContext * ctx)1312 static av_cold void uninit(AVFilterContext *ctx)
1313 {
1314     common_uninit(ctx->priv);
1315 }
1316 
query_formats(AVFilterContext * ctx)1317 static int query_formats(AVFilterContext *ctx)
1318 {
1319     AVFilterFormats *formats = NULL;
1320     AVFilterChannelLayouts *layouts = NULL;
1321     AVFilterLink *inlink = ctx->inputs[0];
1322     AVFilterLink *outlink = ctx->outputs[0];
1323     static const enum AVSampleFormat sample_fmts[] = { AV_SAMPLE_FMT_FLT, AV_SAMPLE_FMT_NONE };
1324     static const enum AVPixelFormat pix_fmts[] = {
1325         AV_PIX_FMT_YUV420P, AV_PIX_FMT_YUV422P,
1326         AV_PIX_FMT_YUV444P, AV_PIX_FMT_RGB24, AV_PIX_FMT_NONE
1327     };
1328     static const AVChannelLayout channel_layouts[] = { AV_CHANNEL_LAYOUT_STEREO,
1329                                                        AV_CHANNEL_LAYOUT_STEREO_DOWNMIX, { 0 } };
1330     int ret;
1331 
1332     /* set input audio formats */
1333     formats = ff_make_format_list(sample_fmts);
1334     if ((ret = ff_formats_ref(formats, &inlink->outcfg.formats)) < 0)
1335         return ret;
1336 
1337     layouts = ff_make_channel_layout_list(channel_layouts);
1338     if ((ret = ff_channel_layouts_ref(layouts, &inlink->outcfg.channel_layouts)) < 0)
1339         return ret;
1340 
1341     formats = ff_all_samplerates();
1342     if ((ret = ff_formats_ref(formats, &inlink->outcfg.samplerates)) < 0)
1343         return ret;
1344 
1345     /* set output video format */
1346     formats = ff_make_format_list(pix_fmts);
1347     if ((ret = ff_formats_ref(formats, &outlink->incfg.formats)) < 0)
1348         return ret;
1349 
1350     return 0;
1351 }
1352 
config_output(AVFilterLink * outlink)1353 static int config_output(AVFilterLink *outlink)
1354 {
1355     AVFilterContext *ctx = outlink->src;
1356     AVFilterLink *inlink = ctx->inputs[0];
1357     ShowCQTContext *s = ctx->priv;
1358     float scale;
1359     int ret;
1360 
1361     common_uninit(s);
1362 
1363     outlink->w = s->width;
1364     outlink->h = s->height;
1365     s->format = outlink->format;
1366     outlink->sample_aspect_ratio = av_make_q(1, 1);
1367     outlink->frame_rate = s->rate;
1368     outlink->time_base = av_mul_q(av_inv_q(s->rate), av_make_q(1, PTS_STEP));
1369     av_log(ctx, AV_LOG_INFO, "video: %dx%d %s %d/%d fps, bar_h = %d, axis_h = %d, sono_h = %d.\n",
1370            s->width, s->height, av_get_pix_fmt_name(s->format), s->rate.num, s->rate.den,
1371            s->bar_h, s->axis_h, s->sono_h);
1372 
1373     s->cqt_len = s->width * s->fcount;
1374     if (!(s->freq = create_freq_table(s->basefreq, s->endfreq, s->cqt_len)))
1375         return AVERROR(ENOMEM);
1376 
1377     if ((ret = init_volume(s)) < 0)
1378         return ret;
1379 
1380     s->fft_bits = FFMAX(ceil(log2(inlink->sample_rate * s->timeclamp)), 4);
1381     s->fft_len = 1 << s->fft_bits;
1382     av_log(ctx, AV_LOG_INFO, "fft_len = %d, cqt_len = %d.\n", s->fft_len, s->cqt_len);
1383 
1384     ret = av_tx_init(&s->fft_ctx, &s->tx_fn, AV_TX_FLOAT_FFT, 0, s->fft_len, &scale, 0);
1385     s->fft_data = av_calloc(s->fft_len, sizeof(*s->fft_data));
1386     s->fft_input = av_calloc(FFALIGN(s->fft_len + 64, 256), sizeof(*s->fft_input));
1387     s->fft_result = av_calloc(FFALIGN(s->fft_len + 64, 256), sizeof(*s->fft_result));
1388     s->cqt_result = av_malloc_array(s->cqt_len, sizeof(*s->cqt_result));
1389     if (!s->fft_ctx || !s->fft_data || !s->fft_result || !s->cqt_result)
1390         return AVERROR(ENOMEM);
1391 
1392     s->remaining_fill_max = s->fft_len / 2;
1393     if (s->attack > 0.0) {
1394         int k;
1395 
1396         s->remaining_fill_max = FFMIN(s->remaining_fill_max, ceil(inlink->sample_rate * s->attack));
1397         s->attack_data = av_malloc_array(s->remaining_fill_max, sizeof(*s->attack_data));
1398         if (!s->attack_data)
1399             return AVERROR(ENOMEM);
1400 
1401         for (k = 0; k < s->remaining_fill_max; k++) {
1402             double y = M_PI * k / (inlink->sample_rate * s->attack);
1403             s->attack_data[k] = 0.355768 + 0.487396 * cos(y) + 0.144232 * cos(2*y) + 0.012604 * cos(3*y);
1404         }
1405     }
1406 
1407     s->cqt_align = 1;
1408     s->cqt_calc = cqt_calc;
1409     s->permute_coeffs = NULL;
1410     s->draw_sono = draw_sono;
1411     if (s->format == AV_PIX_FMT_RGB24) {
1412         s->draw_bar = draw_bar_rgb;
1413         s->draw_axis = draw_axis_rgb;
1414         s->update_sono = update_sono_rgb;
1415     } else {
1416         s->draw_bar = draw_bar_yuv;
1417         s->draw_axis = draw_axis_yuv;
1418         s->update_sono = update_sono_yuv;
1419     }
1420 
1421 #if ARCH_X86
1422     ff_showcqt_init_x86(s);
1423 #endif
1424 
1425     if ((ret = init_cqt(s)) < 0)
1426         return ret;
1427 
1428     if (s->axis_h) {
1429         if (!s->axis) {
1430             if ((ret = init_axis_empty(s)) < 0)
1431                 return ret;
1432         } else if (s->axisfile) {
1433             if (init_axis_from_file(s) < 0) {
1434                 av_log(ctx, AV_LOG_WARNING, "loading axis image failed, fallback to font rendering.\n");
1435                 if (init_axis_from_font(s) < 0) {
1436                     av_log(ctx, AV_LOG_WARNING, "loading axis font failed, disable text drawing.\n");
1437                     if ((ret = init_axis_empty(s)) < 0)
1438                         return ret;
1439                 }
1440             }
1441         } else {
1442             if (init_axis_from_font(s) < 0) {
1443                 av_log(ctx, AV_LOG_WARNING, "loading axis font failed, disable text drawing.\n");
1444                 if ((ret = init_axis_empty(s)) < 0)
1445                     return ret;
1446             }
1447         }
1448     }
1449 
1450     if (s->sono_h) {
1451         s->sono_frame = alloc_frame_empty((outlink->format == AV_PIX_FMT_YUV420P) ?
1452                         AV_PIX_FMT_YUV422P : outlink->format, s->width, s->sono_h);
1453         if (!s->sono_frame)
1454             return AVERROR(ENOMEM);
1455     }
1456 
1457     s->h_buf = av_malloc_array(s->cqt_len, sizeof (*s->h_buf));
1458     s->rcp_h_buf = av_malloc_array(s->width, sizeof(*s->rcp_h_buf));
1459     s->c_buf = av_malloc_array(s->width, sizeof(*s->c_buf));
1460     if (!s->h_buf || !s->rcp_h_buf || !s->c_buf)
1461         return AVERROR(ENOMEM);
1462 
1463     s->sono_count = 0;
1464     s->next_pts = 0;
1465     s->sono_idx = 0;
1466     s->remaining_fill = s->remaining_fill_max;
1467     s->remaining_frac = 0;
1468     s->step_frac = av_div_q(av_make_q(inlink->sample_rate, s->count) , s->rate);
1469     s->step = (int)(s->step_frac.num / s->step_frac.den);
1470     s->step_frac.num %= s->step_frac.den;
1471     if (s->step_frac.num) {
1472         av_log(ctx, AV_LOG_INFO, "audio: %d Hz, step = %d + %d/%d.\n",
1473                inlink->sample_rate, s->step, s->step_frac.num, s->step_frac.den);
1474         av_log(ctx, AV_LOG_WARNING, "fractional step.\n");
1475     } else {
1476         av_log(ctx, AV_LOG_INFO, "audio: %d Hz, step = %d.\n",
1477                inlink->sample_rate, s->step);
1478     }
1479 
1480     return 0;
1481 }
1482 
1483 
filter_frame(AVFilterLink * inlink,AVFrame * insamples)1484 static int filter_frame(AVFilterLink *inlink, AVFrame *insamples)
1485 {
1486     AVFilterContext *ctx = inlink->dst;
1487     AVFilterLink *outlink = ctx->outputs[0];
1488     ShowCQTContext *s = ctx->priv;
1489     int remaining, step, ret, x, i, j, m;
1490     float *audio_data;
1491     AVFrame *out = NULL;
1492 
1493     if (!insamples) {
1494         while (s->remaining_fill < s->remaining_fill_max) {
1495             memset(&s->fft_data[s->fft_len/2 + s->remaining_fill_max - s->remaining_fill], 0, sizeof(*s->fft_data) * s->remaining_fill);
1496             ret = plot_cqt(ctx, &out);
1497             if (ret < 0)
1498                 return ret;
1499 
1500             step = s->step + (s->step_frac.num + s->remaining_frac) / s->step_frac.den;
1501             s->remaining_frac = (s->step_frac.num + s->remaining_frac) % s->step_frac.den;
1502             for (x = 0; x < (s->fft_len/2 + s->remaining_fill_max - step); x++)
1503                 s->fft_data[x] = s->fft_data[x+step];
1504             s->remaining_fill += step;
1505 
1506             if (out)
1507                 return ff_filter_frame(outlink, out);
1508         }
1509         return AVERROR_EOF;
1510     }
1511 
1512     remaining = insamples->nb_samples;
1513     audio_data = (float*) insamples->data[0];
1514 
1515     while (remaining) {
1516         i = insamples->nb_samples - remaining;
1517         j = s->fft_len/2 + s->remaining_fill_max - s->remaining_fill;
1518         if (remaining >= s->remaining_fill) {
1519             for (m = 0; m < s->remaining_fill; m++) {
1520                 s->fft_data[j+m].re = audio_data[2*(i+m)];
1521                 s->fft_data[j+m].im = audio_data[2*(i+m)+1];
1522             }
1523             ret = plot_cqt(ctx, &out);
1524             if (ret < 0) {
1525                 av_frame_free(&insamples);
1526                 return ret;
1527             }
1528             remaining -= s->remaining_fill;
1529             if (out) {
1530                 int64_t pts = av_rescale_q(insamples->pts, inlink->time_base, av_make_q(1, inlink->sample_rate));
1531                 pts += insamples->nb_samples - remaining - s->remaining_fill_max;
1532                 pts = av_rescale_q(pts, av_make_q(1, inlink->sample_rate), outlink->time_base);
1533                 if (FFABS(pts - out->pts) > PTS_TOLERANCE) {
1534                     av_log(ctx, AV_LOG_DEBUG, "changing pts from %"PRId64" (%.3f) to %"PRId64" (%.3f).\n",
1535                            out->pts, out->pts * av_q2d(outlink->time_base),
1536                            pts, pts * av_q2d(outlink->time_base));
1537                     out->pts = pts;
1538                     s->next_pts = pts + PTS_STEP;
1539                 }
1540                 ret = ff_filter_frame(outlink, out);
1541                 if (ret < 0) {
1542                     av_frame_free(&insamples);
1543                     return ret;
1544                 }
1545                 out = NULL;
1546             }
1547             step = s->step + (s->step_frac.num + s->remaining_frac) / s->step_frac.den;
1548             s->remaining_frac = (s->step_frac.num + s->remaining_frac) % s->step_frac.den;
1549             for (m = 0; m < s->fft_len/2 + s->remaining_fill_max - step; m++)
1550                 s->fft_data[m] = s->fft_data[m+step];
1551             s->remaining_fill = step;
1552         } else {
1553             for (m = 0; m < remaining; m++) {
1554                 s->fft_data[j+m].re = audio_data[2*(i+m)];
1555                 s->fft_data[j+m].im = audio_data[2*(i+m)+1];
1556             }
1557             s->remaining_fill -= remaining;
1558             remaining = 0;
1559         }
1560     }
1561     av_frame_free(&insamples);
1562     return 0;
1563 }
1564 
request_frame(AVFilterLink * outlink)1565 static int request_frame(AVFilterLink *outlink)
1566 {
1567     AVFilterLink *inlink = outlink->src->inputs[0];
1568     int ret;
1569 
1570     ret = ff_request_frame(inlink);
1571     if (ret == AVERROR_EOF)
1572         ret = filter_frame(inlink, NULL);
1573     return ret;
1574 }
1575 
1576 static const AVFilterPad showcqt_inputs[] = {
1577     {
1578         .name         = "default",
1579         .type         = AVMEDIA_TYPE_AUDIO,
1580         .filter_frame = filter_frame,
1581     },
1582 };
1583 
1584 static const AVFilterPad showcqt_outputs[] = {
1585     {
1586         .name          = "default",
1587         .type          = AVMEDIA_TYPE_VIDEO,
1588         .config_props  = config_output,
1589         .request_frame = request_frame,
1590     },
1591 };
1592 
1593 const AVFilter ff_avf_showcqt = {
1594     .name          = "showcqt",
1595     .description   = NULL_IF_CONFIG_SMALL("Convert input audio to a CQT (Constant/Clamped Q Transform) spectrum video output."),
1596     .init          = init,
1597     .uninit        = uninit,
1598     .priv_size     = sizeof(ShowCQTContext),
1599     FILTER_INPUTS(showcqt_inputs),
1600     FILTER_OUTPUTS(showcqt_outputs),
1601     FILTER_QUERY_FUNC(query_formats),
1602     .priv_class    = &showcqt_class,
1603 };
1604