falkTX
/
FFmpeg
mirror of https://github.com/falkTX/FFmpeg.git
1
0
Fork 0
Code Issues 0 Releases 338 Wiki Activity
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
75730 Commits
32 Branches
2.4GB
Tree: 513fcd4167
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from '513fcd4167'
${ noResults }
FFmpeg/libavfilter/avf_showcqt.c

765 lines
28KB
Raw Blame History 

  1. /*

  2.  * Copyright (c) 2014 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 "libavcodec/avfft.h"

  23. #include "libavutil/avassert.h"

  24. #include "libavutil/channel_layout.h"

  25. #include "libavutil/opt.h"

  26. #include "libavutil/xga_font_data.h"

  27. #include "libavutil/eval.h"

  28. #include "avfilter.h"

  29. #include "internal.h"

  30. 

  31. #include <math.h>

  32. #include <stdlib.h>

  33. 

  34. #if CONFIG_LIBFREETYPE

  35. #include <ft2build.h>

  36. #include FT_FREETYPE_H

  37. #endif

  38. 

  39. /* this filter is designed to do 16 bins/semitones constant Q transform with Brown-Puckette algorithm

  40.  * start from E0 to D#10 (10 octaves)

  41.  * so there are 16 bins/semitones * 12 semitones/octaves * 10 octaves = 1920 bins

  42.  * match with full HD resolution */

  43. 

  44. #define VIDEO_WIDTH 1920

  45. #define VIDEO_HEIGHT 1080

  46. #define FONT_HEIGHT 32

  47. #define SPECTOGRAM_HEIGHT ((VIDEO_HEIGHT-FONT_HEIGHT)/2)

  48. #define SPECTOGRAM_START (VIDEO_HEIGHT-SPECTOGRAM_HEIGHT)

  49. #define BASE_FREQ 20.051392800492

  50. #define TLENGTH_MIN 0.001

  51. #define TLENGTH_DEFAULT "384/f*tc/(384/f+tc)"

  52. #define VOLUME_MIN 1e-10

  53. #define VOLUME_MAX 100.0

  54. #define FONTCOLOR_DEFAULT "st(0, (midi(f)-59.5)/12);" \

  55.     "st(1, if(between(ld(0),0,1), 0.5-0.5*cos(2*PI*ld(0)), 0));" \

  56.     "r(1-ld(1)) + b(ld(1))"

  57. 

  58. typedef struct {

  59.     FFTSample *values;

  60.     int start, len;

  61. } Coeffs;

  62. 

  63. typedef struct {

  64.     const AVClass *class;

  65.     AVFrame *outpicref;

  66.     FFTContext *fft_context;

  67.     FFTComplex *fft_data;

  68.     FFTComplex *fft_result;

  69.     uint8_t *spectogram;

  70.     Coeffs coeffs[VIDEO_WIDTH];

  71.     uint8_t *font_alpha;

  72.     char *fontfile;     /* using freetype */

  73.     uint8_t fontcolor_value[VIDEO_WIDTH*3];  /* result of fontcolor option */

  74.     int64_t frame_count;

  75.     int spectogram_count;

  76.     int spectogram_index;

  77.     int fft_bits;

  78.     int req_fullfilled;

  79.     int remaining_fill;

  80.     char *tlength;

  81.     char *volume;

  82.     char *fontcolor;

  83.     double timeclamp;   /* lower timeclamp, time-accurate, higher timeclamp, freq-accurate (at low freq)*/

  84.     float coeffclamp;   /* lower coeffclamp, more precise, higher coeffclamp, faster */

  85.     int fullhd;         /* if true, output video is at full HD resolution, otherwise it will be halved */

  86.     float gamma;        /* lower gamma, more contrast, higher gamma, more range */

  87.     float gamma2;       /* gamma of bargraph */

  88.     int fps;            /* the required fps is so strict, so it's enough to be int, but 24000/1001 etc cannot be encoded */

  89.     int count;          /* fps * count = transform rate */

  90.     int draw_text;

  91. } ShowCQTContext;

  92. 

  93. #define OFFSET(x) offsetof(ShowCQTContext, x)

  94. #define FLAGS AV_OPT_FLAG_FILTERING_PARAM|AV_OPT_FLAG_VIDEO_PARAM

  95. 

  96. static const AVOption showcqt_options[] = {

  97.     { "volume", "set volume", OFFSET(volume), AV_OPT_TYPE_STRING, { .str = "16" }, CHAR_MIN, CHAR_MAX, FLAGS },

  98.     { "tlength", "set transform length", OFFSET(tlength), AV_OPT_TYPE_STRING, { .str = TLENGTH_DEFAULT }, CHAR_MIN, CHAR_MAX, FLAGS },

  99.     { "timeclamp", "set timeclamp", OFFSET(timeclamp), AV_OPT_TYPE_DOUBLE, { .dbl = 0.17 }, 0.1, 1.0, FLAGS },

  100.     { "coeffclamp", "set coeffclamp", OFFSET(coeffclamp), AV_OPT_TYPE_FLOAT, { .dbl = 1 }, 0.1, 10, FLAGS },

  101.     { "gamma", "set gamma", OFFSET(gamma), AV_OPT_TYPE_FLOAT, { .dbl = 3 }, 1, 7, FLAGS },

  102.     { "gamma2", "set gamma of bargraph", OFFSET(gamma2), AV_OPT_TYPE_FLOAT, { .dbl = 1 }, 1, 7, FLAGS },

  103.     { "fullhd", "set full HD resolution", OFFSET(fullhd), AV_OPT_TYPE_BOOL, { .i64 = 1 }, 0, 1, FLAGS },

  104.     { "fps", "set video fps", OFFSET(fps), AV_OPT_TYPE_INT, { .i64 = 25 }, 10, 100, FLAGS },

  105.     { "count", "set number of transform per frame", OFFSET(count), AV_OPT_TYPE_INT, { .i64 = 6 }, 1, 30, FLAGS },

  106.     { "fontfile", "set font file", OFFSET(fontfile), AV_OPT_TYPE_STRING, { .str = NULL }, CHAR_MIN, CHAR_MAX, FLAGS },

  107.     { "fontcolor", "set font color", OFFSET(fontcolor), AV_OPT_TYPE_STRING, { .str = FONTCOLOR_DEFAULT }, CHAR_MIN, CHAR_MAX, FLAGS },

  108.     { "text", "draw text", OFFSET(draw_text), AV_OPT_TYPE_BOOL, { .i64 = 1 }, 0, 1, FLAGS },

  109.     { NULL }

  110. };

  111. 

  112. AVFILTER_DEFINE_CLASS(showcqt);

  113. 

  114. static av_cold void uninit(AVFilterContext *ctx)

  115. {

  116.     int k;

  117. 

  118.     ShowCQTContext *s = ctx->priv;

  119.     av_fft_end(s->fft_context);

  120.     s->fft_context = NULL;

  121.     for (k = 0; k < VIDEO_WIDTH; k++)

  122.         av_freep(&s->coeffs[k].values);

  123.     av_freep(&s->fft_data);

  124.     av_freep(&s->fft_result);

  125.     av_freep(&s->spectogram);

  126.     av_freep(&s->font_alpha);

  127.     av_frame_free(&s->outpicref);

  128. }

  129. 

  130. static int query_formats(AVFilterContext *ctx)

  131. {

  132.     AVFilterFormats *formats = NULL;

  133.     AVFilterChannelLayouts *layouts = NULL;

  134.     AVFilterLink *inlink = ctx->inputs[0];

  135.     AVFilterLink *outlink = ctx->outputs[0];

  136.     static const enum AVSampleFormat sample_fmts[] = { AV_SAMPLE_FMT_FLT, AV_SAMPLE_FMT_NONE };

  137.     static const enum AVPixelFormat pix_fmts[] = { AV_PIX_FMT_RGB24, AV_PIX_FMT_NONE };

  138.     static const int64_t channel_layouts[] = { AV_CH_LAYOUT_STEREO, AV_CH_LAYOUT_STEREO_DOWNMIX, -1 };

  139.     static const int samplerates[] = { 44100, 48000, -1 };

  140. 

  141.     /* set input audio formats */

  142.     formats = ff_make_format_list(sample_fmts);

  143.     if (!formats)

  144.         return AVERROR(ENOMEM);

  145.     ff_formats_ref(formats, &inlink->out_formats);

  146. 

  147.     layouts = avfilter_make_format64_list(channel_layouts);

  148.     if (!layouts)

  149.         return AVERROR(ENOMEM);

  150.     ff_channel_layouts_ref(layouts, &inlink->out_channel_layouts);

  151. 

  152.     formats = ff_make_format_list(samplerates);

  153.     if (!formats)

  154.         return AVERROR(ENOMEM);

  155.     ff_formats_ref(formats, &inlink->out_samplerates);

  156. 

  157.     /* set output video format */

  158.     formats = ff_make_format_list(pix_fmts);

  159.     if (!formats)

  160.         return AVERROR(ENOMEM);

  161.     ff_formats_ref(formats, &outlink->in_formats);

  162. 

  163.     return 0;

  164. }

  165. 

  166. #if CONFIG_LIBFREETYPE

  167. static void load_freetype_font(AVFilterContext *ctx)

  168. {

  169.     static const char str[] = "EF G A BC D ";

  170.     ShowCQTContext *s = ctx->priv;

  171.     FT_Library lib = NULL;

  172.     FT_Face face = NULL;

  173.     int video_scale = s->fullhd ? 2 : 1;

  174.     int video_width = (VIDEO_WIDTH/2) * video_scale;

  175.     int font_height = (FONT_HEIGHT/2) * video_scale;

  176.     int font_width = 8 * video_scale;

  177.     int font_repeat = font_width * 12;

  178.     int linear_hori_advance = font_width * 65536;

  179.     int non_monospace_warning = 0;

  180.     int x;

  181. 

  182.     s->font_alpha = NULL;

  183. 

  184.     if (!s->fontfile)

  185.         return;

  186. 

  187.     if (FT_Init_FreeType(&lib))

  188.         goto fail;

  189. 

  190.     if (FT_New_Face(lib, s->fontfile, 0, &face))

  191.         goto fail;

  192. 

  193.     if (FT_Set_Char_Size(face, 16*64, 0, 0, 0))

  194.         goto fail;

  195. 

  196.     if (FT_Load_Char(face, 'A', FT_LOAD_RENDER))

  197.         goto fail;

  198. 

  199.     if (FT_Set_Char_Size(face, 16*64 * linear_hori_advance / face->glyph->linearHoriAdvance, 0, 0, 0))

  200.         goto fail;

  201. 

  202.     s->font_alpha = av_malloc_array(font_height, video_width);

  203.     if (!s->font_alpha)

  204.         goto fail;

  205. 

  206.     memset(s->font_alpha, 0, font_height * video_width);

  207. 

  208.     for (x = 0; x < 12; x++) {

  209.         int sx, sy, rx, bx, by, dx, dy;

  210. 

  211.         if (str[x] == ' ')

  212.             continue;

  213. 

  214.         if (FT_Load_Char(face, str[x], FT_LOAD_RENDER))

  215.             goto fail;

  216. 

  217.         if (face->glyph->advance.x != font_width*64 && !non_monospace_warning) {

  218.             av_log(ctx, AV_LOG_WARNING, "Font is not monospace\n");

  219.             non_monospace_warning = 1;

  220.         }

  221. 

  222.         sy = font_height - 4*video_scale - face->glyph->bitmap_top;

  223.         for (rx = 0; rx < 10; rx++) {

  224.             sx = rx * font_repeat + x * font_width + face->glyph->bitmap_left;

  225.             for (by = 0; by < face->glyph->bitmap.rows; by++) {

  226.                 dy = by + sy;

  227.                 if (dy < 0)

  228.                     continue;

  229.                 if (dy >= font_height)

  230.                     break;

  231. 

  232.                 for (bx = 0; bx < face->glyph->bitmap.width; bx++) {

  233.                     dx = bx + sx;

  234.                     if (dx < 0)

  235.                         continue;

  236.                     if (dx >= video_width)

  237.                         break;

  238.                     s->font_alpha[dy*video_width+dx] = face->glyph->bitmap.buffer[by*face->glyph->bitmap.width+bx];

  239.                 }

  240.             }

  241.         }

  242.     }

  243. 

  244.     FT_Done_Face(face);

  245.     FT_Done_FreeType(lib);

  246.     return;

  247. 

  248.     fail:

  249.     av_log(ctx, AV_LOG_WARNING, "Error while loading freetype font, using default font instead\n");

  250.     FT_Done_Face(face);

  251.     FT_Done_FreeType(lib);

  252.     av_freep(&s->font_alpha);

  253.     return;

  254. }

  255. #endif

  256. 

  257. static double a_weighting(void *p, double f)

  258. {

  259.     double ret = 12200.0*12200.0 * (f*f*f*f);

  260.     ret /= (f*f + 20.6*20.6) * (f*f + 12200.0*12200.0) *

  261.            sqrt((f*f + 107.7*107.7) * (f*f + 737.9*737.9));

  262.     return ret;

  263. }

  264. 

  265. static double b_weighting(void *p, double f)

  266. {

  267.     double ret = 12200.0*12200.0 * (f*f*f);

  268.     ret /= (f*f + 20.6*20.6) * (f*f + 12200.0*12200.0) * sqrt(f*f + 158.5*158.5);

  269.     return ret;

  270. }

  271. 

  272. static double c_weighting(void *p, double f)

  273. {

  274.     double ret = 12200.0*12200.0 * (f*f);

  275.     ret /= (f*f + 20.6*20.6) * (f*f + 12200.0*12200.0);

  276.     return ret;

  277. }

  278. 

  279. static double midi(void *p, double f)

  280. {

  281.     return log2(f/440.0) * 12.0 + 69.0;

  282. }

  283. 

  284. static double r_func(void *p, double x)

  285. {

  286.     x = av_clipd(x, 0.0, 1.0);

  287.     return (int)(x*255.0+0.5) << 16;

  288. }

  289. 

  290. static double g_func(void *p, double x)

  291. {

  292.     x = av_clipd(x, 0.0, 1.0);

  293.     return (int)(x*255.0+0.5) << 8;

  294. }

  295. 

  296. static double b_func(void *p, double x)

  297. {

  298.     x = av_clipd(x, 0.0, 1.0);

  299.     return (int)(x*255.0+0.5);

  300. }

  301. 

  302. static int config_output(AVFilterLink *outlink)

  303. {

  304.     AVFilterContext *ctx = outlink->src;

  305.     AVFilterLink *inlink = ctx->inputs[0];

  306.     ShowCQTContext *s = ctx->priv;

  307.     AVExpr *tlength_expr = NULL, *volume_expr = NULL, *fontcolor_expr = NULL;

  308.     uint8_t *fontcolor_value = s->fontcolor_value;

  309.     static const char * const expr_vars[] = { "timeclamp", "tc", "frequency", "freq", "f", NULL };

  310.     static const char * const expr_func_names[] = { "a_weighting", "b_weighting", "c_weighting", NULL };

  311.     static const char * const expr_fontcolor_func_names[] = { "midi", "r", "g", "b", NULL };

  312.     static double (* const expr_funcs[])(void *, double) = { a_weighting, b_weighting, c_weighting, NULL };

  313.     static double (* const expr_fontcolor_funcs[])(void *, double) = { midi, r_func, g_func, b_func, NULL };

  314.     int fft_len, k, x, ret;

  315.     int num_coeffs = 0;

  316.     int rate = inlink->sample_rate;

  317.     double max_len = rate * (double) s->timeclamp;

  318.     int video_scale = s->fullhd ? 2 : 1;

  319.     int video_width = (VIDEO_WIDTH/2) * video_scale;

  320.     int video_height = (VIDEO_HEIGHT/2) * video_scale;

  321.     int spectogram_height = (SPECTOGRAM_HEIGHT/2) * video_scale;

  322. 

  323.     s->fft_bits = ceil(log2(max_len));

  324.     fft_len = 1 << s->fft_bits;

  325. 

  326.     if (rate % (s->fps * s->count)) {

  327.         av_log(ctx, AV_LOG_ERROR, "Rate (%u) is not divisible by fps*count (%u*%u)\n", rate, s->fps, s->count);

  328.         return AVERROR(EINVAL);

  329.     }

  330. 

  331.     s->fft_data         = av_malloc_array(fft_len, sizeof(*s->fft_data));

  332.     s->fft_result       = av_malloc_array(fft_len + 1, sizeof(*s->fft_result));

  333.     s->fft_context      = av_fft_init(s->fft_bits, 0);

  334. 

  335.     if (!s->fft_data || !s->fft_result || !s->fft_context)

  336.         return AVERROR(ENOMEM);

  337. 

  338. #if CONFIG_LIBFREETYPE

  339.     load_freetype_font(ctx);

  340. #else

  341.     if (s->fontfile)

  342.         av_log(ctx, AV_LOG_WARNING, "Freetype is not available, ignoring fontfile option\n");

  343.     s->font_alpha = NULL;

  344. #endif

  345. 

  346.     ret = av_expr_parse(&tlength_expr, s->tlength, expr_vars, NULL, NULL, NULL, NULL, 0, ctx);

  347.     if (ret < 0)

  348.         goto eval_error;

  349. 

  350.     ret = av_expr_parse(&volume_expr, s->volume, expr_vars, expr_func_names,

  351.                         expr_funcs, NULL, NULL, 0, ctx);

  352.     if (ret < 0)

  353.         goto eval_error;

  354. 

  355.     ret = av_expr_parse(&fontcolor_expr, s->fontcolor, expr_vars, expr_fontcolor_func_names,

  356.                         expr_fontcolor_funcs, NULL, NULL, 0, ctx);

  357.     if (ret < 0)

  358.         goto eval_error;

  359. 

  360.     for (k = 0; k < VIDEO_WIDTH; k++) {

  361.         double freq = BASE_FREQ * exp2(k * (1.0/192.0));

  362.         double flen, center, tlength, volume;

  363.         int start, end;

  364.         double expr_vars_val[] = { s->timeclamp, s->timeclamp, freq, freq, freq, 0 };

  365. 

  366.         tlength = av_expr_eval(tlength_expr, expr_vars_val, NULL);

  367.         if (isnan(tlength)) {

  368.             av_log(ctx, AV_LOG_WARNING, "at freq %g: tlength is nan, setting it to %g\n", freq, s->timeclamp);

  369.             tlength = s->timeclamp;

  370.         } else if (tlength < TLENGTH_MIN) {

  371.             av_log(ctx, AV_LOG_WARNING, "at freq %g: tlength is %g, setting it to %g\n", freq, tlength, TLENGTH_MIN);

  372.             tlength = TLENGTH_MIN;

  373.         } else if (tlength > s->timeclamp) {

  374.             av_log(ctx, AV_LOG_WARNING, "at freq %g: tlength is %g, setting it to %g\n", freq, tlength, s->timeclamp);

  375.             tlength = s->timeclamp;

  376.         }

  377. 

  378.         volume = FFABS(av_expr_eval(volume_expr, expr_vars_val, NULL));

  379.         if (isnan(volume)) {

  380.             av_log(ctx, AV_LOG_WARNING, "at freq %g: volume is nan, setting it to 0\n", freq);

  381.             volume = VOLUME_MIN;

  382.         } else if (volume < VOLUME_MIN) {

  383.             volume = VOLUME_MIN;

  384.         } else if (volume > VOLUME_MAX) {

  385.             av_log(ctx, AV_LOG_WARNING, "at freq %g: volume is %g, setting it to %g\n", freq, volume, VOLUME_MAX);

  386.             volume = VOLUME_MAX;

  387.         }

  388. 

  389.         if (s->fullhd || !(k & 1)) {

  390.             int fontcolor = av_expr_eval(fontcolor_expr, expr_vars_val, NULL);

  391.             fontcolor_value[0] = (fontcolor >> 16) & 0xFF;

  392.             fontcolor_value[1] = (fontcolor >> 8) & 0xFF;

  393.             fontcolor_value[2] = fontcolor & 0xFF;

  394.             fontcolor_value += 3;

  395.         }

  396. 

  397.         /* direct frequency domain windowing */

  398.         flen = 8.0 * fft_len / (tlength * rate);

  399.         center = freq * fft_len / rate;

  400.         start = FFMAX(0, ceil(center - 0.5 * flen));

  401.         end = FFMIN(fft_len, floor(center + 0.5 * flen));

  402.         s->coeffs[k].len = end - start + 1;

  403.         s->coeffs[k].start = start;

  404.         num_coeffs += s->coeffs[k].len;

  405.         s->coeffs[k].values = av_malloc_array(s->coeffs[k].len, sizeof(*s->coeffs[k].values));

  406.         if (!s->coeffs[k].values) {

  407.             ret = AVERROR(ENOMEM);

  408.             goto eval_error;

  409.         }

  410.         for (x = start; x <= end; x++) {

  411.             int sign = (x & 1) ? (-1) : 1;

  412.             double u = 2.0 * M_PI * (x - center) * (1.0/flen);

  413.             /* nuttall window */

  414.             double w = 0.355768 + 0.487396 * cos(u) + 0.144232 * cos(2*u) + 0.012604 * cos(3*u);

  415.             s->coeffs[k].values[x-start] = sign * volume * (1.0/fft_len) * w;

  416.         }

  417.     }

  418.     av_expr_free(fontcolor_expr);

  419.     av_expr_free(volume_expr);

  420.     av_expr_free(tlength_expr);

  421.     av_log(ctx, AV_LOG_INFO, "fft_len=%u, num_coeffs=%u\n", fft_len, num_coeffs);

  422. 

  423.     outlink->w = video_width;

  424.     outlink->h = video_height;

  425. 

  426.     s->req_fullfilled = 0;

  427.     s->spectogram_index = 0;

  428.     s->frame_count = 0;

  429.     s->spectogram_count = 0;

  430.     s->remaining_fill = fft_len >> 1;

  431.     memset(s->fft_data, 0, fft_len * sizeof(*s->fft_data));

  432. 

  433.     s->outpicref = ff_get_video_buffer(outlink, outlink->w, outlink->h);

  434.     if (!s->outpicref)

  435.         return AVERROR(ENOMEM);

  436. 

  437.     s->spectogram = av_calloc(spectogram_height, s->outpicref->linesize[0]);

  438.     if (!s->spectogram)

  439.         return AVERROR(ENOMEM);

  440. 

  441.     outlink->sample_aspect_ratio = av_make_q(1, 1);

  442.     outlink->time_base = av_make_q(1, s->fps);

  443.     outlink->frame_rate = av_make_q(s->fps, 1);

  444.     return 0;

  445. 

  446. eval_error:

  447.     av_expr_free(fontcolor_expr);

  448.     av_expr_free(volume_expr);

  449.     av_expr_free(tlength_expr);

  450.     return ret;

  451. }

  452. 

  453. static int plot_cqt(AVFilterLink *inlink)

  454. {

  455.     AVFilterContext *ctx = inlink->dst;

  456.     ShowCQTContext *s = ctx->priv;

  457.     AVFilterLink *outlink = ctx->outputs[0];

  458.     int fft_len = 1 << s->fft_bits;

  459.     FFTSample result[VIDEO_WIDTH][4];

  460.     int x, y, ret = 0;

  461.     int linesize = s->outpicref->linesize[0];

  462.     int video_scale = s->fullhd ? 2 : 1;

  463.     int video_width = (VIDEO_WIDTH/2) * video_scale;

  464.     int spectogram_height = (SPECTOGRAM_HEIGHT/2) * video_scale;

  465.     int spectogram_start = (SPECTOGRAM_START/2) * video_scale;

  466.     int font_height = (FONT_HEIGHT/2) * video_scale;

  467. 

  468.     /* real part contains left samples, imaginary part contains right samples */

  469.     memcpy(s->fft_result, s->fft_data, fft_len * sizeof(*s->fft_data));

  470.     av_fft_permute(s->fft_context, s->fft_result);

  471.     av_fft_calc(s->fft_context, s->fft_result);

  472.     s->fft_result[fft_len] = s->fft_result[0];

  473. 

  474.     /* calculating cqt */

  475.     for (x = 0; x < VIDEO_WIDTH; x++) {

  476.         int u;

  477.         FFTComplex v = {0,0};

  478.         FFTComplex w = {0,0};

  479.         FFTComplex l, r;

  480. 

  481.         for (u = 0; u < s->coeffs[x].len; u++) {

  482.             FFTSample value = s->coeffs[x].values[u];

  483.             int index = s->coeffs[x].start + u;

  484.             v.re += value * s->fft_result[index].re;

  485.             v.im += value * s->fft_result[index].im;

  486.             w.re += value * s->fft_result[fft_len - index].re;

  487.             w.im += value * s->fft_result[fft_len - index].im;

  488.         }

  489. 

  490.         /* separate left and right, (and multiply by 2.0) */

  491.         l.re = v.re + w.re;

  492.         l.im = v.im - w.im;

  493.         r.re = w.im + v.im;

  494.         r.im = w.re - v.re;

  495.         /* result is power, not amplitude */

  496.         result[x][0] = l.re * l.re + l.im * l.im;

  497.         result[x][2] = r.re * r.re + r.im * r.im;

  498.         result[x][1] = 0.5f * (result[x][0] + result[x][2]);

  499. 

  500.         if (s->gamma2 == 1.0f)

  501.             result[x][3] = result[x][1];

  502.         else if (s->gamma2 == 2.0f)

  503.             result[x][3] = sqrtf(result[x][1]);

  504.         else if (s->gamma2 == 3.0f)

  505.             result[x][3] = cbrtf(result[x][1]);

  506.         else if (s->gamma2 == 4.0f)

  507.             result[x][3] = sqrtf(sqrtf(result[x][1]));

  508.         else

  509.             result[x][3] = expf(logf(result[x][1]) * (1.0f / s->gamma2));

  510. 

  511.         result[x][0] = FFMIN(1.0f, result[x][0]);

  512.         result[x][1] = FFMIN(1.0f, result[x][1]);

  513.         result[x][2] = FFMIN(1.0f, result[x][2]);

  514.         if (s->gamma == 1.0f) {

  515.             result[x][0] = 255.0f * result[x][0];

  516.             result[x][1] = 255.0f * result[x][1];

  517.             result[x][2] = 255.0f * result[x][2];

  518.         } else if (s->gamma == 2.0f) {

  519.             result[x][0] = 255.0f * sqrtf(result[x][0]);

  520.             result[x][1] = 255.0f * sqrtf(result[x][1]);

  521.             result[x][2] = 255.0f * sqrtf(result[x][2]);

  522.         } else if (s->gamma == 3.0f) {

  523.             result[x][0] = 255.0f * cbrtf(result[x][0]);

  524.             result[x][1] = 255.0f * cbrtf(result[x][1]);

  525.             result[x][2] = 255.0f * cbrtf(result[x][2]);

  526.         } else if (s->gamma == 4.0f) {

  527.             result[x][0] = 255.0f * sqrtf(sqrtf(result[x][0]));

  528.             result[x][1] = 255.0f * sqrtf(sqrtf(result[x][1]));

  529.             result[x][2] = 255.0f * sqrtf(sqrtf(result[x][2]));

  530.         } else {

  531.             result[x][0] = 255.0f * expf(logf(result[x][0]) * (1.0f / s->gamma));

  532.             result[x][1] = 255.0f * expf(logf(result[x][1]) * (1.0f / s->gamma));

  533.             result[x][2] = 255.0f * expf(logf(result[x][2]) * (1.0f / s->gamma));

  534.         }

  535.     }

  536. 

  537.     if (!s->fullhd) {

  538.         for (x = 0; x < video_width; x++) {

  539.             result[x][0] = 0.5f * (result[2*x][0] + result[2*x+1][0]);

  540.             result[x][1] = 0.5f * (result[2*x][1] + result[2*x+1][1]);

  541.             result[x][2] = 0.5f * (result[2*x][2] + result[2*x+1][2]);

  542.             result[x][3] = 0.5f * (result[2*x][3] + result[2*x+1][3]);

  543.         }

  544.     }

  545. 

  546.     for (x = 0; x < video_width; x++) {

  547.         s->spectogram[s->spectogram_index*linesize + 3*x] = result[x][0] + 0.5f;

  548.         s->spectogram[s->spectogram_index*linesize + 3*x + 1] = result[x][1] + 0.5f;

  549.         s->spectogram[s->spectogram_index*linesize + 3*x + 2] = result[x][2] + 0.5f;

  550.     }

  551. 

  552.     /* drawing */

  553.     if (!s->spectogram_count) {

  554.         uint8_t *data = (uint8_t*) s->outpicref->data[0];

  555.         float rcp_result[VIDEO_WIDTH];

  556.         int total_length = linesize * spectogram_height;

  557.         int back_length = linesize * s->spectogram_index;

  558. 

  559.         for (x = 0; x < video_width; x++)

  560.             rcp_result[x] = 1.0f / (result[x][3]+0.0001f);

  561. 

  562.         /* drawing bar */

  563.         for (y = 0; y < spectogram_height; y++) {

  564.             float height = (spectogram_height - y) * (1.0f/spectogram_height);

  565.             uint8_t *lineptr = data + y * linesize;

  566.             for (x = 0; x < video_width; x++) {

  567.                 float mul;

  568.                 if (result[x][3] <= height) {

  569.                     *lineptr++ = 0;

  570.                     *lineptr++ = 0;

  571.                     *lineptr++ = 0;

  572.                 } else {

  573.                     mul = (result[x][3] - height) * rcp_result[x];

  574.                     *lineptr++ = mul * result[x][0] + 0.5f;

  575.                     *lineptr++ = mul * result[x][1] + 0.5f;

  576.                     *lineptr++ = mul * result[x][2] + 0.5f;

  577.                 }

  578.             }

  579.         }

  580. 

  581.         /* drawing font */

  582.         if (s->font_alpha && s->draw_text) {

  583.             for (y = 0; y < font_height; y++) {

  584.                 uint8_t *lineptr = data + (spectogram_height + y) * linesize;

  585.                 uint8_t *spectogram_src = s->spectogram + s->spectogram_index * linesize;

  586.                 uint8_t *fontcolor_value = s->fontcolor_value;

  587.                 for (x = 0; x < video_width; x++) {

  588.                     uint8_t alpha = s->font_alpha[y*video_width+x];

  589.                     lineptr[3*x] = (spectogram_src[3*x] * (255-alpha) + fontcolor_value[0] * alpha + 255) >> 8;

  590.                     lineptr[3*x+1] = (spectogram_src[3*x+1] * (255-alpha) + fontcolor_value[1] * alpha + 255) >> 8;

  591.                     lineptr[3*x+2] = (spectogram_src[3*x+2] * (255-alpha) + fontcolor_value[2] * alpha + 255) >> 8;

  592.                     fontcolor_value += 3;

  593.                 }

  594.             }

  595.         } else if (s->draw_text) {

  596.             for (y = 0; y < font_height; y++) {

  597.                 uint8_t *lineptr = data + (spectogram_height + y) * linesize;

  598.                 memcpy(lineptr, s->spectogram + s->spectogram_index * linesize, video_width*3);

  599.             }

  600.             for (x = 0; x < video_width; x += video_width/10) {

  601.                 int u;

  602.                 static const char str[] = "EF G A BC D ";

  603.                 uint8_t *startptr = data + spectogram_height * linesize + x * 3;

  604.                 for (u = 0; str[u]; u++) {

  605.                     int v;

  606.                     for (v = 0; v < 16; v++) {

  607.                         uint8_t *p = startptr + v * linesize * video_scale + 8 * 3 * u * video_scale;

  608.                         int ux = x + 8 * u * video_scale;

  609.                         int mask;

  610.                         for (mask = 0x80; mask; mask >>= 1) {

  611.                             if (mask & avpriv_vga16_font[str[u] * 16 + v]) {

  612.                                 p[0] = s->fontcolor_value[3*ux];

  613.                                 p[1] = s->fontcolor_value[3*ux+1];

  614.                                 p[2] = s->fontcolor_value[3*ux+2];

  615.                                 if (video_scale == 2) {

  616.                                     p[linesize] = p[0];

  617.                                     p[linesize+1] = p[1];

  618.                                     p[linesize+2] = p[2];

  619.                                     p[3] = p[linesize+3] = s->fontcolor_value[3*ux+3];

  620.                                     p[4] = p[linesize+4] = s->fontcolor_value[3*ux+4];

  621.                                     p[5] = p[linesize+5] = s->fontcolor_value[3*ux+5];

  622.                                 }

  623.                             }

  624.                             p  += 3 * video_scale;

  625.                             ux += video_scale;

  626.                         }

  627.                     }

  628.                 }

  629.             }

  630.         } else {

  631.             for (y = 0; y < font_height; y++) {

  632.                 uint8_t *lineptr = data + (spectogram_height + y) * linesize;

  633.                 uint8_t *spectogram_src = s->spectogram + s->spectogram_index * linesize;

  634.                 for (x = 0; x < video_width; x++) {

  635.                     lineptr[3*x] = spectogram_src[3*x];

  636.                     lineptr[3*x+1] = spectogram_src[3*x+1];

  637.                     lineptr[3*x+2] = spectogram_src[3*x+2];

  638.                 }

  639.             }

  640.         }

  641. 

  642.         /* drawing spectogram/sonogram */

  643.         data += spectogram_start * linesize;

  644.         memcpy(data, s->spectogram + s->spectogram_index*linesize, total_length - back_length);

  645. 

  646.         data += total_length - back_length;

  647.         if (back_length)

  648.             memcpy(data, s->spectogram, back_length);

  649. 

  650.         s->outpicref->pts = s->frame_count;

  651.         ret = ff_filter_frame(outlink, av_frame_clone(s->outpicref));

  652.         s->req_fullfilled = 1;

  653.         s->frame_count++;

  654.     }

  655.     s->spectogram_count = (s->spectogram_count + 1) % s->count;

  656.     s->spectogram_index = (s->spectogram_index + spectogram_height - 1) % spectogram_height;

  657.     return ret;

  658. }

  659. 

  660. static int filter_frame(AVFilterLink *inlink, AVFrame *insamples)

  661. {

  662.     AVFilterContext *ctx = inlink->dst;

  663.     ShowCQTContext *s = ctx->priv;

  664.     int step = inlink->sample_rate / (s->fps * s->count);

  665.     int fft_len = 1 << s->fft_bits;

  666.     int remaining;

  667.     float *audio_data;

  668. 

  669.     if (!insamples) {

  670.         while (s->remaining_fill < (fft_len >> 1)) {

  671.             int ret, x;

  672.             memset(&s->fft_data[fft_len - s->remaining_fill], 0, sizeof(*s->fft_data) * s->remaining_fill);

  673.             ret = plot_cqt(inlink);

  674.             if (ret < 0)

  675.                 return ret;

  676.             for (x = 0; x < (fft_len-step); x++)

  677.                 s->fft_data[x] = s->fft_data[x+step];

  678.             s->remaining_fill += step;

  679.         }

  680.         return AVERROR_EOF;

  681.     }

  682. 

  683.     remaining = insamples->nb_samples;

  684.     audio_data = (float*) insamples->data[0];

  685. 

  686.     while (remaining) {

  687.         if (remaining >= s->remaining_fill) {

  688.             int i = insamples->nb_samples - remaining;

  689.             int j = fft_len - s->remaining_fill;

  690.             int m, ret;

  691.             for (m = 0; m < s->remaining_fill; m++) {

  692.                 s->fft_data[j+m].re = audio_data[2*(i+m)];

  693.                 s->fft_data[j+m].im = audio_data[2*(i+m)+1];

  694.             }

  695.             ret = plot_cqt(inlink);

  696.             if (ret < 0) {

  697.                 av_frame_free(&insamples);

  698.                 return ret;

  699.             }

  700.             remaining -= s->remaining_fill;

  701.             for (m = 0; m < fft_len-step; m++)

  702.                 s->fft_data[m] = s->fft_data[m+step];

  703.             s->remaining_fill = step;

  704.         } else {

  705.             int i = insamples->nb_samples - remaining;

  706.             int j = fft_len - s->remaining_fill;

  707.             int m;

  708.             for (m = 0; m < remaining; m++) {

  709.                 s->fft_data[m+j].re = audio_data[2*(i+m)];

  710.                 s->fft_data[m+j].im = audio_data[2*(i+m)+1];

  711.             }

  712.             s->remaining_fill -= remaining;

  713.             remaining = 0;

  714.         }

  715.     }

  716.     av_frame_free(&insamples);

  717.     return 0;

  718. }

  719. 

  720. static int request_frame(AVFilterLink *outlink)

  721. {

  722.     ShowCQTContext *s = outlink->src->priv;

  723.     AVFilterLink *inlink = outlink->src->inputs[0];

  724.     int ret;

  725. 

  726.     s->req_fullfilled = 0;

  727.     do {

  728.         ret = ff_request_frame(inlink);

  729.     } while (!s->req_fullfilled && ret >= 0);

  730. 

  731.     if (ret == AVERROR_EOF && s->outpicref)

  732.         filter_frame(inlink, NULL);

  733.     return ret;

  734. }

  735. 

  736. static const AVFilterPad showcqt_inputs[] = {

  737.     {

  738.         .name         = "default",

  739.         .type         = AVMEDIA_TYPE_AUDIO,

  740.         .filter_frame = filter_frame,

  741.     },

  742.     { NULL }

  743. };

  744. 

  745. static const AVFilterPad showcqt_outputs[] = {

  746.     {

  747.         .name          = "default",

  748.         .type          = AVMEDIA_TYPE_VIDEO,

  749.         .config_props  = config_output,

  750.         .request_frame = request_frame,

  751.     },

  752.     { NULL }

  753. };

  754. 

  755. AVFilter ff_avf_showcqt = {

  756.     .name          = "showcqt",

  757.     .description   = NULL_IF_CONFIG_SMALL("Convert input audio to a CQT (Constant Q Transform) spectrum video output."),

  758.     .uninit        = uninit,

  759.     .query_formats = query_formats,

  760.     .priv_size     = sizeof(ShowCQTContext),

  761.     .inputs        = showcqt_inputs,

  762.     .outputs       = showcqt_outputs,

  763.     .priv_class    = &showcqt_class,

  764. };