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.
76083 Commits
32 Branches
312MB
Tree: 564db3e55d
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from '564db3e55d'
${ noResults }
FFmpeg/libavfilter/avf_showcqt.c

755 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 remaining_fill;

  79.     char *tlength;

  80.     char *volume;

  81.     char *fontcolor;

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

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

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

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

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

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

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

  89.     int draw_text;

  90. } ShowCQTContext;

  91. 

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

  93. #define FLAGS AV_OPT_FLAG_FILTERING_PARAM|AV_OPT_FLAG_VIDEO_PARAM

  94. 

  95. static const AVOption showcqt_options[] = {

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

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

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

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

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

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

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

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

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

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

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

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

  108.     { NULL }

  109. };

  110. 

  111. AVFILTER_DEFINE_CLASS(showcqt);

  112. 

  113. static av_cold void uninit(AVFilterContext *ctx)

  114. {

  115.     int k;

  116. 

  117.     ShowCQTContext *s = ctx->priv;

  118.     av_fft_end(s->fft_context);

  119.     s->fft_context = NULL;

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

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

  122.     av_freep(&s->fft_data);

  123.     av_freep(&s->fft_result);

  124.     av_freep(&s->spectogram);

  125.     av_freep(&s->font_alpha);

  126.     av_frame_free(&s->outpicref);

  127. }

  128. 

  129. static int query_formats(AVFilterContext *ctx)

  130. {

  131.     AVFilterFormats *formats = NULL;

  132.     AVFilterChannelLayouts *layouts = NULL;

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

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

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

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

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

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

  139.     int ret;

  140. 

  141.     /* set input audio formats */

  142.     formats = ff_make_format_list(sample_fmts);

  143.     if ((ret = ff_formats_ref(formats, &inlink->out_formats)) < 0)

  144.         return ret;

  145. 

  146.     layouts = avfilter_make_format64_list(channel_layouts);

  147.     if ((ret = ff_channel_layouts_ref(layouts, &inlink->out_channel_layouts)) < 0)

  148.         return ret;

  149. 

  150.     formats = ff_make_format_list(samplerates);

  151.     if ((ret = ff_formats_ref(formats, &inlink->out_samplerates)) < 0)

  152.         return ret;

  153. 

  154.     /* set output video format */

  155.     formats = ff_make_format_list(pix_fmts);

  156.     if ((ret = ff_formats_ref(formats, &outlink->in_formats)) < 0)

  157.         return ret;

  158. 

  159.     return 0;

  160. }

  161. 

  162. #if CONFIG_LIBFREETYPE

  163. static void load_freetype_font(AVFilterContext *ctx)

  164. {

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

  166.     ShowCQTContext *s = ctx->priv;

  167.     FT_Library lib = NULL;

  168.     FT_Face face = NULL;

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

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

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

  172.     int font_width = 8 * video_scale;

  173.     int font_repeat = font_width * 12;

  174.     int linear_hori_advance = font_width * 65536;

  175.     int non_monospace_warning = 0;

  176.     int x;

  177. 

  178.     s->font_alpha = NULL;

  179. 

  180.     if (!s->fontfile)

  181.         return;

  182. 

  183.     if (FT_Init_FreeType(&lib))

  184.         goto fail;

  185. 

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

  187.         goto fail;

  188. 

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

  190.         goto fail;

  191. 

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

  193.         goto fail;

  194. 

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

  196.         goto fail;

  197. 

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

  199.     if (!s->font_alpha)

  200.         goto fail;

  201. 

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

  203. 

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

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

  206. 

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

  208.             continue;

  209. 

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

  211.             goto fail;

  212. 

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

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

  215.             non_monospace_warning = 1;

  216.         }

  217. 

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

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

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

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

  222.                 dy = by + sy;

  223.                 if (dy < 0)

  224.                     continue;

  225.                 if (dy >= font_height)

  226.                     break;

  227. 

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

  229.                     dx = bx + sx;

  230.                     if (dx < 0)

  231.                         continue;

  232.                     if (dx >= video_width)

  233.                         break;

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

  235.                 }

  236.             }

  237.         }

  238.     }

  239. 

  240.     FT_Done_Face(face);

  241.     FT_Done_FreeType(lib);

  242.     return;

  243. 

  244.     fail:

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

  246.     FT_Done_Face(face);

  247.     FT_Done_FreeType(lib);

  248.     av_freep(&s->font_alpha);

  249.     return;

  250. }

  251. #endif

  252. 

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

  254. {

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

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

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

  258.     return ret;

  259. }

  260. 

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

  262. {

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

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

  265.     return ret;

  266. }

  267. 

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

  269. {

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

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

  272.     return ret;

  273. }

  274. 

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

  276. {

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

  278. }

  279. 

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

  281. {

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

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

  284. }

  285. 

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

  287. {

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

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

  290. }

  291. 

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

  293. {

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

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

  296. }

  297. 

  298. static int config_output(AVFilterLink *outlink)

  299. {

  300.     AVFilterContext *ctx = outlink->src;

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

  302.     ShowCQTContext *s = ctx->priv;

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

  304.     uint8_t *fontcolor_value = s->fontcolor_value;

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

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

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

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

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

  310.     int fft_len, k, x, ret;

  311.     int num_coeffs = 0;

  312.     int rate = inlink->sample_rate;

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

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

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

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

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

  318. 

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

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

  321. 

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

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

  324.         return AVERROR(EINVAL);

  325.     }

  326. 

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

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

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

  330. 

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

  332.         return AVERROR(ENOMEM);

  333. 

  334. #if CONFIG_LIBFREETYPE

  335.     load_freetype_font(ctx);

  336. #else

  337.     if (s->fontfile)

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

  339.     s->font_alpha = NULL;

  340. #endif

  341. 

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

  343.     if (ret < 0)

  344.         goto eval_error;

  345. 

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

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

  348.     if (ret < 0)

  349.         goto eval_error;

  350. 

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

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

  353.     if (ret < 0)

  354.         goto eval_error;

  355. 

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

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

  358.         double flen, center, tlength, volume;

  359.         int start, end;

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

  361. 

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

  363.         if (isnan(tlength)) {

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

  365.             tlength = s->timeclamp;

  366.         } else if (tlength < TLENGTH_MIN) {

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

  368.             tlength = TLENGTH_MIN;

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

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

  371.             tlength = s->timeclamp;

  372.         }

  373. 

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

  375.         if (isnan(volume)) {

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

  377.             volume = VOLUME_MIN;

  378.         } else if (volume < VOLUME_MIN) {

  379.             volume = VOLUME_MIN;

  380.         } else if (volume > VOLUME_MAX) {

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

  382.             volume = VOLUME_MAX;

  383.         }

  384. 

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

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

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

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

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

  390.             fontcolor_value += 3;

  391.         }

  392. 

  393.         /* direct frequency domain windowing */

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

  395.         center = freq * fft_len / rate;

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

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

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

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

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

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

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

  403.             ret = AVERROR(ENOMEM);

  404.             goto eval_error;

  405.         }

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

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

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

  409.             /* nuttall window */

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

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

  412.         }

  413.     }

  414.     av_expr_free(fontcolor_expr);

  415.     av_expr_free(volume_expr);

  416.     av_expr_free(tlength_expr);

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

  418. 

  419.     outlink->w = video_width;

  420.     outlink->h = video_height;

  421. 

  422.     s->spectogram_index = 0;

  423.     s->frame_count = 0;

  424.     s->spectogram_count = 0;

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

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

  427. 

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

  429.     if (!s->outpicref)

  430.         return AVERROR(ENOMEM);

  431. 

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

  433.     if (!s->spectogram)

  434.         return AVERROR(ENOMEM);

  435. 

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

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

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

  439.     return 0;

  440. 

  441. eval_error:

  442.     av_expr_free(fontcolor_expr);

  443.     av_expr_free(volume_expr);

  444.     av_expr_free(tlength_expr);

  445.     return ret;

  446. }

  447. 

  448. static int plot_cqt(AVFilterLink *inlink)

  449. {

  450.     AVFilterContext *ctx = inlink->dst;

  451.     ShowCQTContext *s = ctx->priv;

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

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

  454.     FFTSample result[VIDEO_WIDTH][4];

  455.     int x, y, ret = 0;

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

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

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

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

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

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

  462. 

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

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

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

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

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

  468. 

  469.     /* calculating cqt */

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

  471.         int u;

  472.         FFTComplex v = {0,0};

  473.         FFTComplex w = {0,0};

  474.         FFTComplex l, r;

  475. 

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

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

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

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

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

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

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

  483.         }

  484. 

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

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

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

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

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

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

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

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

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

  494. 

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

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

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

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

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

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

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

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

  503.         else

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

  505. 

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  525.         } else {

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

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

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

  529.         }

  530.     }

  531. 

  532.     if (!s->fullhd) {

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

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

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

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

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

  538.         }

  539.     }

  540. 

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

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

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

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

  545.     }

  546. 

  547.     /* drawing */

  548.     if (!s->spectogram_count) {

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

  550.         float rcp_result[VIDEO_WIDTH];

  551.         int total_length = linesize * spectogram_height;

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

  553. 

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

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

  556. 

  557.         /* drawing bar */

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

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

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

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

  562.                 float mul;

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

  564.                     *lineptr++ = 0;

  565.                     *lineptr++ = 0;

  566.                     *lineptr++ = 0;

  567.                 } else {

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

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

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

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

  572.                 }

  573.             }

  574.         }

  575. 

  576.         /* drawing font */

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

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

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

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

  581.                 uint8_t *fontcolor_value = s->fontcolor_value;

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

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

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

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

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

  587.                     fontcolor_value += 3;

  588.                 }

  589.             }

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

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

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

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

  594.             }

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

  596.                 int u;

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

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

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

  600.                     int v;

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

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

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

  604.                         int mask;

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

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

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

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

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

  610.                                 if (video_scale == 2) {

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

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

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

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

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

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

  617.                                 }

  618.                             }

  619.                             p  += 3 * video_scale;

  620.                             ux += video_scale;

  621.                         }

  622.                     }

  623.                 }

  624.             }

  625.         } else {

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

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

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

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

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

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

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

  633.                 }

  634.             }

  635.         }

  636. 

  637.         /* drawing spectogram/sonogram */

  638.         data += spectogram_start * linesize;

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

  640. 

  641.         data += total_length - back_length;

  642.         if (back_length)

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

  644. 

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

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

  647.         s->frame_count++;

  648.     }

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

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

  651.     return ret;

  652. }

  653. 

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

  655. {

  656.     AVFilterContext *ctx = inlink->dst;

  657.     ShowCQTContext *s = ctx->priv;

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

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

  660.     int remaining;

  661.     float *audio_data;

  662. 

  663.     if (!insamples) {

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

  665.             int ret, x;

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

  667.             ret = plot_cqt(inlink);

  668.             if (ret < 0)

  669.                 return ret;

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

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

  672.             s->remaining_fill += step;

  673.         }

  674.         return AVERROR_EOF;

  675.     }

  676. 

  677.     remaining = insamples->nb_samples;

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

  679. 

  680.     while (remaining) {

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

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

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

  684.             int m, ret;

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

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

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

  688.             }

  689.             ret = plot_cqt(inlink);

  690.             if (ret < 0) {

  691.                 av_frame_free(&insamples);

  692.                 return ret;

  693.             }

  694.             remaining -= s->remaining_fill;

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

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

  697.             s->remaining_fill = step;

  698.         } else {

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

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

  701.             int m;

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

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

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

  705.             }

  706.             s->remaining_fill -= remaining;

  707.             remaining = 0;

  708.         }

  709.     }

  710.     av_frame_free(&insamples);

  711.     return 0;

  712. }

  713. 

  714. static int request_frame(AVFilterLink *outlink)

  715. {

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

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

  718.     int ret;

  719. 

  720.     ret = ff_request_frame(inlink);

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

  722.         filter_frame(inlink, NULL);

  723.     return ret;

  724. }

  725. 

  726. static const AVFilterPad showcqt_inputs[] = {

  727.     {

  728.         .name         = "default",

  729.         .type         = AVMEDIA_TYPE_AUDIO,

  730.         .filter_frame = filter_frame,

  731.     },

  732.     { NULL }

  733. };

  734. 

  735. static const AVFilterPad showcqt_outputs[] = {

  736.     {

  737.         .name          = "default",

  738.         .type          = AVMEDIA_TYPE_VIDEO,

  739.         .config_props  = config_output,

  740.         .request_frame = request_frame,

  741.     },

  742.     { NULL }

  743. };

  744. 

  745. AVFilter ff_avf_showcqt = {

  746.     .name          = "showcqt",

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

  748.     .uninit        = uninit,

  749.     .query_formats = query_formats,

  750.     .priv_size     = sizeof(ShowCQTContext),

  751.     .inputs        = showcqt_inputs,

  752.     .outputs       = showcqt_outputs,

  753.     .priv_class    = &showcqt_class,

  754. };