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.
77597 Commits
32 Branches
637MB
Tree: 9dba3f8f09
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from '9dba3f8f09'
${ noResults }
FFmpeg/libavfilter/avf_showspectrum.c

680 lines
28KB
Raw Blame History 

  1. /*

  2.  * Copyright (c) 2012-2013 Clément Bœsch

  3.  * Copyright (c) 2013 Rudolf Polzer <divverent@xonotic.org>

  4.  * Copyright (c) 2015 Paul B Mahol

  5.  *

  6.  * This file is part of FFmpeg.

  7.  *

  8.  * FFmpeg is free software; you can redistribute it and/or

  9.  * modify it under the terms of the GNU Lesser General Public

  10.  * License as published by the Free Software Foundation; either

  11.  * version 2.1 of the License, or (at your option) any later version.

  12.  *

  13.  * FFmpeg is distributed in the hope that it will be useful,

  14.  * but WITHOUT ANY WARRANTY; without even the implied warranty of

  15.  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU

  16.  * Lesser General Public License for more details.

  17.  *

  18.  * You should have received a copy of the GNU Lesser General Public

  19.  * License along with FFmpeg; if not, write to the Free Software

  20.  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA

  21.  */

  22. 

  23. /**

  24.  * @file

  25.  * audio to spectrum (video) transmedia filter, based on ffplay rdft showmode

  26.  * (by Michael Niedermayer) and lavfi/avf_showwaves (by Stefano Sabatini).

  27.  */

  28. 

  29. #include <math.h>

  30. 

  31. #include "libavcodec/avfft.h"

  32. #include "libavutil/audio_fifo.h"

  33. #include "libavutil/avassert.h"

  34. #include "libavutil/channel_layout.h"

  35. #include "libavutil/opt.h"

  36. #include "audio.h"

  37. #include "video.h"

  38. #include "avfilter.h"

  39. #include "internal.h"

  40. #include "window_func.h"

  41. 

  42. enum DisplayMode  { COMBINED, SEPARATE, NB_MODES };

  43. enum DisplayScale { LINEAR, SQRT, CBRT, LOG, NB_SCALES };

  44. enum ColorMode    { CHANNEL, INTENSITY, RAINBOW, MORELAND, NEBULAE, FIRE, NB_CLMODES };

  45. enum SlideMode    { REPLACE, SCROLL, FULLFRAME, RSCROLL, NB_SLIDES };

  46. enum Orientation  { VERTICAL, HORIZONTAL, NB_ORIENTATIONS };

  47. 

  48. typedef struct {

  49.     const AVClass *class;

  50.     int w, h;

  51.     AVFrame *outpicref;

  52.     int nb_display_channels;

  53.     int orientation;

  54.     int channel_width;

  55.     int channel_height;

  56.     int sliding;                ///< 1 if sliding mode, 0 otherwise

  57.     int mode;                   ///< channel display mode

  58.     int color_mode;             ///< display color scheme

  59.     int scale;

  60.     float saturation;           ///< color saturation multiplier

  61.     int xpos;                   ///< x position (current column)

  62.     RDFTContext *rdft;          ///< Real Discrete Fourier Transform context

  63.     int rdft_bits;              ///< number of bits (RDFT window size = 1<<rdft_bits)

  64.     FFTSample **rdft_data;      ///< bins holder for each (displayed) channels

  65.     float *window_func_lut;     ///< Window function LUT

  66.     int win_func;

  67.     double win_scale;

  68.     float overlap;

  69.     int skip_samples;

  70.     float *combine_buffer;      ///< color combining buffer (3 * h items)

  71.     AVAudioFifo *fifo;

  72.     int64_t pts;

  73. } ShowSpectrumContext;

  74. 

  75. #define OFFSET(x) offsetof(ShowSpectrumContext, x)

  76. #define FLAGS AV_OPT_FLAG_FILTERING_PARAM|AV_OPT_FLAG_VIDEO_PARAM

  77. 

  78. static const AVOption showspectrum_options[] = {

  79.     { "size", "set video size", OFFSET(w), AV_OPT_TYPE_IMAGE_SIZE, {.str = "640x512"}, 0, 0, FLAGS },

  80.     { "s",    "set video size", OFFSET(w), AV_OPT_TYPE_IMAGE_SIZE, {.str = "640x512"}, 0, 0, FLAGS },

  81.     { "slide", "set sliding mode", OFFSET(sliding), AV_OPT_TYPE_INT, {.i64 = 0}, 0, NB_SLIDES-1, FLAGS, "slide" },

  82.         { "replace", "replace old columns with new", 0, AV_OPT_TYPE_CONST, {.i64=REPLACE}, 0, 0, FLAGS, "slide" },

  83.         { "scroll", "scroll from right to left", 0, AV_OPT_TYPE_CONST, {.i64=SCROLL}, 0, 0, FLAGS, "slide" },

  84.         { "rscroll", "scroll from left to right", 0, AV_OPT_TYPE_CONST, {.i64=RSCROLL}, 0, 0, FLAGS, "slide" },

  85.         { "fullframe", "return full frames", 0, AV_OPT_TYPE_CONST, {.i64=FULLFRAME}, 0, 0, FLAGS, "slide" },

  86.     { "mode", "set channel display mode", OFFSET(mode), AV_OPT_TYPE_INT, {.i64=COMBINED}, COMBINED, NB_MODES-1, FLAGS, "mode" },

  87.         { "combined", "combined mode", 0, AV_OPT_TYPE_CONST, {.i64=COMBINED}, 0, 0, FLAGS, "mode" },

  88.         { "separate", "separate mode", 0, AV_OPT_TYPE_CONST, {.i64=SEPARATE}, 0, 0, FLAGS, "mode" },

  89.     { "color", "set channel coloring", OFFSET(color_mode), AV_OPT_TYPE_INT, {.i64=CHANNEL}, CHANNEL, NB_CLMODES-1, FLAGS, "color" },

  90.         { "channel",   "separate color for each channel", 0, AV_OPT_TYPE_CONST, {.i64=CHANNEL},   0, 0, FLAGS, "color" },

  91.         { "intensity", "intensity based coloring",        0, AV_OPT_TYPE_CONST, {.i64=INTENSITY}, 0, 0, FLAGS, "color" },

  92.         { "rainbow",   "rainbow based coloring",          0, AV_OPT_TYPE_CONST, {.i64=RAINBOW},   0, 0, FLAGS, "color" },

  93.         { "moreland",  "moreland based coloring",         0, AV_OPT_TYPE_CONST, {.i64=MORELAND},  0, 0, FLAGS, "color" },

  94.         { "nebulae",   "nebulae based coloring",          0, AV_OPT_TYPE_CONST, {.i64=NEBULAE},   0, 0, FLAGS, "color" },

  95.         { "fire",      "fire based coloring",             0, AV_OPT_TYPE_CONST, {.i64=FIRE},      0, 0, FLAGS, "color" },

  96.     { "scale", "set display scale", OFFSET(scale), AV_OPT_TYPE_INT, {.i64=SQRT}, LINEAR, NB_SCALES-1, FLAGS, "scale" },

  97.         { "sqrt", "square root", 0, AV_OPT_TYPE_CONST, {.i64=SQRT},   0, 0, FLAGS, "scale" },

  98.         { "cbrt", "cubic root",  0, AV_OPT_TYPE_CONST, {.i64=CBRT},   0, 0, FLAGS, "scale" },

  99.         { "log",  "logarithmic", 0, AV_OPT_TYPE_CONST, {.i64=LOG},    0, 0, FLAGS, "scale" },

  100.         { "lin",  "linear",      0, AV_OPT_TYPE_CONST, {.i64=LINEAR}, 0, 0, FLAGS, "scale" },

  101.     { "saturation", "color saturation multiplier", OFFSET(saturation), AV_OPT_TYPE_FLOAT, {.dbl = 1}, -10, 10, FLAGS },

  102.     { "win_func", "set window function", OFFSET(win_func), AV_OPT_TYPE_INT, {.i64 = WFUNC_HANNING}, 0, NB_WFUNC-1, FLAGS, "win_func" },

  103.         { "rect",     "Rectangular",      0, AV_OPT_TYPE_CONST, {.i64=WFUNC_RECT},     0, 0, FLAGS, "win_func" },

  104.         { "bartlett", "Bartlett",         0, AV_OPT_TYPE_CONST, {.i64=WFUNC_BARTLETT}, 0, 0, FLAGS, "win_func" },

  105.         { "hann",     "Hann",             0, AV_OPT_TYPE_CONST, {.i64=WFUNC_HANNING},  0, 0, FLAGS, "win_func" },

  106.         { "hanning",  "Hanning",          0, AV_OPT_TYPE_CONST, {.i64=WFUNC_HANNING},  0, 0, FLAGS, "win_func" },

  107.         { "hamming",  "Hamming",          0, AV_OPT_TYPE_CONST, {.i64=WFUNC_HAMMING},  0, 0, FLAGS, "win_func" },

  108.         { "blackman", "Blackman",         0, AV_OPT_TYPE_CONST, {.i64=WFUNC_BLACKMAN}, 0, 0, FLAGS, "win_func" },

  109.         { "welch",    "Welch",            0, AV_OPT_TYPE_CONST, {.i64=WFUNC_WELCH},    0, 0, FLAGS, "win_func" },

  110.         { "flattop",  "Flat-top",         0, AV_OPT_TYPE_CONST, {.i64=WFUNC_FLATTOP},  0, 0, FLAGS, "win_func" },

  111.         { "bharris",  "Blackman-Harris",  0, AV_OPT_TYPE_CONST, {.i64=WFUNC_BHARRIS},  0, 0, FLAGS, "win_func" },

  112.         { "bnuttall", "Blackman-Nuttall", 0, AV_OPT_TYPE_CONST, {.i64=WFUNC_BNUTTALL}, 0, 0, FLAGS, "win_func" },

  113.         { "bhann",    "Bartlett-Hann",    0, AV_OPT_TYPE_CONST, {.i64=WFUNC_BHANN},    0, 0, FLAGS, "win_func" },

  114.         { "sine",     "Sine",             0, AV_OPT_TYPE_CONST, {.i64=WFUNC_SINE},     0, 0, FLAGS, "win_func" },

  115.         { "nuttall",  "Nuttall",          0, AV_OPT_TYPE_CONST, {.i64=WFUNC_NUTTALL},  0, 0, FLAGS, "win_func" },

  116.         { "lanczos",  "Lanczos",          0, AV_OPT_TYPE_CONST, {.i64=WFUNC_LANCZOS},  0, 0, FLAGS, "win_func" },

  117.         { "gauss",    "Gauss",            0, AV_OPT_TYPE_CONST, {.i64=WFUNC_GAUSS},    0, 0, FLAGS, "win_func" },

  118.     { "orientation", "set orientation", OFFSET(orientation), AV_OPT_TYPE_INT, {.i64=VERTICAL}, 0, NB_ORIENTATIONS-1, FLAGS, "orientation" },

  119.         { "vertical",   NULL, 0, AV_OPT_TYPE_CONST, {.i64=VERTICAL},   0, 0, FLAGS, "orientation" },

  120.         { "horizontal", NULL, 0, AV_OPT_TYPE_CONST, {.i64=HORIZONTAL}, 0, 0, FLAGS, "orientation" },

  121.     { "overlap", "set window overlap", OFFSET(overlap), AV_OPT_TYPE_FLOAT, {.dbl = 0}, 0, 1, FLAGS },

  122.     { NULL }

  123. };

  124. 

  125. AVFILTER_DEFINE_CLASS(showspectrum);

  126. 

  127. static const struct ColorTable {

  128.     float a, y, u, v;

  129. } color_table[][8] = {

  130.     [INTENSITY] = {

  131.     {    0,                  0,                  0,                   0 },

  132.     { 0.13, .03587126228984074,  .1573300977624594, -.02548747583751842 },

  133.     { 0.30, .18572281794568020,  .1772436246393981,  .17475554840414750 },

  134.     { 0.60, .28184980583656130, -.1593064119945782,  .47132074554608920 },

  135.     { 0.73, .65830621175547810, -.3716070802232764,  .24352759331252930 },

  136.     { 0.78, .76318535758242900, -.4307467689263783,  .16866496622310430 },

  137.     { 0.91, .95336363636363640, -.2045454545454546,  .03313636363636363 },

  138.     {    1,                  1,                  0,                   0 }},

  139.     [RAINBOW] = {

  140.     {    0,                  0,                  0,                   0 },

  141.     { 0.13,            44/256.,     (189-128)/256.,      (138-128)/256. },

  142.     { 0.25,            29/256.,     (186-128)/256.,      (119-128)/256. },

  143.     { 0.38,           119/256.,     (194-128)/256.,       (53-128)/256. },

  144.     { 0.60,           111/256.,      (73-128)/256.,       (59-128)/256. },

  145.     { 0.73,           205/256.,      (19-128)/256.,      (149-128)/256. },

  146.     { 0.86,           135/256.,      (83-128)/256.,      (200-128)/256. },

  147.     {    1,            73/256.,      (95-128)/256.,      (225-128)/256. }},

  148.     [MORELAND] = {

  149.     {    0,            44/256.,     (181-128)/256.,      (112-128)/256. },

  150.     { 0.13,           126/256.,     (177-128)/256.,      (106-128)/256. },

  151.     { 0.25,           164/256.,     (163-128)/256.,      (109-128)/256. },

  152.     { 0.38,           200/256.,     (140-128)/256.,      (120-128)/256. },

  153.     { 0.60,           201/256.,     (117-128)/256.,      (141-128)/256. },

  154.     { 0.73,           177/256.,     (103-128)/256.,      (165-128)/256. },

  155.     { 0.86,           136/256.,     (100-128)/256.,      (183-128)/256. },

  156.     {    1,            68/256.,     (117-128)/256.,      (203-128)/256. }},

  157.     [NEBULAE] = {

  158.     {    0,            10/256.,     (134-128)/256.,      (132-128)/256. },

  159.     { 0.23,            21/256.,     (137-128)/256.,      (130-128)/256. },

  160.     { 0.45,            35/256.,     (134-128)/256.,      (134-128)/256. },

  161.     { 0.57,            51/256.,     (130-128)/256.,      (139-128)/256. },

  162.     { 0.67,           104/256.,     (116-128)/256.,      (162-128)/256. },

  163.     { 0.77,           120/256.,     (105-128)/256.,      (188-128)/256. },

  164.     { 0.87,           140/256.,     (105-128)/256.,      (188-128)/256. },

  165.     {    1,                  1,                  0,                   0 }},

  166.     [FIRE] = {

  167.     {    0,                  0,                  0,                   0 },

  168.     { 0.23,            44/256.,     (132-128)/256.,      (127-128)/256. },

  169.     { 0.45,            62/256.,     (116-128)/256.,      (140-128)/256. },

  170.     { 0.57,            75/256.,     (105-128)/256.,      (152-128)/256. },

  171.     { 0.67,            95/256.,      (91-128)/256.,      (166-128)/256. },

  172.     { 0.77,           126/256.,      (74-128)/256.,      (172-128)/256. },

  173.     { 0.87,           164/256.,      (73-128)/256.,      (162-128)/256. },

  174.     {    1,                  1,                  0,                   0 }},

  175. };

  176. 

  177. static av_cold void uninit(AVFilterContext *ctx)

  178. {

  179.     ShowSpectrumContext *s = ctx->priv;

  180.     int i;

  181. 

  182.     av_freep(&s->combine_buffer);

  183.     av_rdft_end(s->rdft);

  184.     for (i = 0; i < s->nb_display_channels; i++)

  185.         av_freep(&s->rdft_data[i]);

  186.     av_freep(&s->rdft_data);

  187.     av_freep(&s->window_func_lut);

  188.     av_frame_free(&s->outpicref);

  189.     av_audio_fifo_free(s->fifo);

  190. }

  191. 

  192. static int query_formats(AVFilterContext *ctx)

  193. {

  194.     AVFilterFormats *formats = NULL;

  195.     AVFilterChannelLayouts *layouts = NULL;

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

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

  198.     static const enum AVSampleFormat sample_fmts[] = { AV_SAMPLE_FMT_S16P, AV_SAMPLE_FMT_NONE };

  199.     static const enum AVPixelFormat pix_fmts[] = { AV_PIX_FMT_YUVJ444P, AV_PIX_FMT_NONE };

  200.     int ret;

  201. 

  202.     /* set input audio formats */

  203.     formats = ff_make_format_list(sample_fmts);

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

  205.         return ret;

  206. 

  207.     layouts = ff_all_channel_layouts();

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

  209.         return ret;

  210. 

  211.     formats = ff_all_samplerates();

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

  213.         return ret;

  214. 

  215.     /* set output video format */

  216.     formats = ff_make_format_list(pix_fmts);

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

  218.         return ret;

  219. 

  220.     return 0;

  221. }

  222. 

  223. static int config_output(AVFilterLink *outlink)

  224. {

  225.     AVFilterContext *ctx = outlink->src;

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

  227.     ShowSpectrumContext *s = ctx->priv;

  228.     int i, rdft_bits, win_size, h, w;

  229.     float overlap;

  230. 

  231.     outlink->w = s->w;

  232.     outlink->h = s->h;

  233. 

  234.     h = (s->mode == COMBINED || s->orientation == HORIZONTAL) ? outlink->h : outlink->h / inlink->channels;

  235.     w = (s->mode == COMBINED || s->orientation == VERTICAL)   ? outlink->w : outlink->w / inlink->channels;

  236.     s->channel_height = h;

  237.     s->channel_width  = w;

  238. 

  239.     if (s->orientation == VERTICAL) {

  240.         /* RDFT window size (precision) according to the requested output frame height */

  241.         for (rdft_bits = 1; 1 << rdft_bits < 2 * h; rdft_bits++);

  242.     } else {

  243.         /* RDFT window size (precision) according to the requested output frame width */

  244.         for (rdft_bits = 1; 1 << rdft_bits < 2 * w; rdft_bits++);

  245.     }

  246.     win_size = 1 << rdft_bits;

  247. 

  248.     /* (re-)configuration if the video output changed (or first init) */

  249.     if (rdft_bits != s->rdft_bits) {

  250.         AVFrame *outpicref;

  251. 

  252.         av_rdft_end(s->rdft);

  253.         s->rdft = av_rdft_init(rdft_bits, DFT_R2C);

  254.         if (!s->rdft) {

  255.             av_log(ctx, AV_LOG_ERROR, "Unable to create RDFT context. "

  256.                    "The window size might be too high.\n");

  257.             return AVERROR(EINVAL);

  258.         }

  259.         s->rdft_bits = rdft_bits;

  260. 

  261.         /* RDFT buffers: x2 for each (display) channel buffer.

  262.          * Note: we use free and malloc instead of a realloc-like function to

  263.          * make sure the buffer is aligned in memory for the FFT functions. */

  264.         for (i = 0; i < s->nb_display_channels; i++)

  265.             av_freep(&s->rdft_data[i]);

  266.         av_freep(&s->rdft_data);

  267.         s->nb_display_channels = inlink->channels;

  268. 

  269.         s->rdft_data = av_calloc(s->nb_display_channels, sizeof(*s->rdft_data));

  270.         if (!s->rdft_data)

  271.             return AVERROR(ENOMEM);

  272.         for (i = 0; i < s->nb_display_channels; i++) {

  273.             s->rdft_data[i] = av_calloc(win_size, sizeof(**s->rdft_data));

  274.             if (!s->rdft_data[i])

  275.                 return AVERROR(ENOMEM);

  276.         }

  277. 

  278.         /* pre-calc windowing function */

  279.         s->window_func_lut =

  280.             av_realloc_f(s->window_func_lut, win_size,

  281.                          sizeof(*s->window_func_lut));

  282.         if (!s->window_func_lut)

  283.             return AVERROR(ENOMEM);

  284.         ff_generate_window_func(s->window_func_lut, win_size, s->win_func, &overlap);

  285.         if (s->overlap == 1)

  286.             s->overlap = overlap;

  287.         s->skip_samples = (1. - s->overlap) * win_size;

  288.         if (s->skip_samples < 1) {

  289.             av_log(ctx, AV_LOG_ERROR, "overlap %f too big\n", s->overlap);

  290.             return AVERROR(EINVAL);

  291.         }

  292. 

  293.         for (s->win_scale = 0, i = 0; i < win_size; i++) {

  294.             s->win_scale += s->window_func_lut[i] * s->window_func_lut[i];

  295.         }

  296.         s->win_scale = 1. / (sqrt(s->win_scale) * 32768.);

  297. 

  298.         /* prepare the initial picref buffer (black frame) */

  299.         av_frame_free(&s->outpicref);

  300.         s->outpicref = outpicref =

  301.             ff_get_video_buffer(outlink, outlink->w, outlink->h);

  302.         if (!outpicref)

  303.             return AVERROR(ENOMEM);

  304.         outlink->sample_aspect_ratio = (AVRational){1,1};

  305.         for (i = 0; i < outlink->h; i++) {

  306.             memset(outpicref->data[0] + i * outpicref->linesize[0],   0, outlink->w);

  307.             memset(outpicref->data[1] + i * outpicref->linesize[1], 128, outlink->w);

  308.             memset(outpicref->data[2] + i * outpicref->linesize[2], 128, outlink->w);

  309.         }

  310.     }

  311. 

  312.     if ((s->orientation == VERTICAL   && s->xpos >= outlink->w) ||

  313.         (s->orientation == HORIZONTAL && s->xpos >= outlink->h))

  314.         s->xpos = 0;

  315. 

  316.     outlink->frame_rate = av_make_q(inlink->sample_rate, win_size * (1.-s->overlap));

  317.     if (s->orientation == VERTICAL && s->sliding == FULLFRAME)

  318.         outlink->frame_rate.den *= outlink->w;

  319.     if (s->orientation == HORIZONTAL && s->sliding == FULLFRAME)

  320.         outlink->frame_rate.den *= outlink->h;

  321. 

  322.     if (s->orientation == VERTICAL) {

  323.         s->combine_buffer =

  324.             av_realloc_f(s->combine_buffer, outlink->h * 3,

  325.                          sizeof(*s->combine_buffer));

  326.     } else {

  327.         s->combine_buffer =

  328.             av_realloc_f(s->combine_buffer, outlink->w * 3,

  329.                          sizeof(*s->combine_buffer));

  330.     }

  331. 

  332.     av_log(ctx, AV_LOG_VERBOSE, "s:%dx%d RDFT window size:%d\n",

  333.            s->w, s->h, win_size);

  334. 

  335.     av_audio_fifo_free(s->fifo);

  336.     s->fifo = av_audio_fifo_alloc(inlink->format, inlink->channels, win_size);

  337.     if (!s->fifo)

  338.         return AVERROR(ENOMEM);

  339.     return 0;

  340. }

  341. 

  342. static int request_frame(AVFilterLink *outlink)

  343. {

  344.     ShowSpectrumContext *s = outlink->src->priv;

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

  346.     unsigned i;

  347.     int ret;

  348. 

  349.     ret = ff_request_frame(inlink);

  350.     if (ret == AVERROR_EOF && s->sliding == FULLFRAME && s->xpos > 0 &&

  351.         s->outpicref) {

  352.         if (s->orientation == VERTICAL) {

  353.             for (i = 0; i < outlink->h; i++) {

  354.                 memset(s->outpicref->data[0] + i * s->outpicref->linesize[0] + s->xpos,   0, outlink->w - s->xpos);

  355.                 memset(s->outpicref->data[1] + i * s->outpicref->linesize[1] + s->xpos, 128, outlink->w - s->xpos);

  356.                 memset(s->outpicref->data[2] + i * s->outpicref->linesize[2] + s->xpos, 128, outlink->w - s->xpos);

  357.             }

  358.         } else {

  359.             for (i = s->xpos; i < outlink->h; i++) {

  360.                 memset(s->outpicref->data[0] + i * s->outpicref->linesize[0],   0, outlink->w);

  361.                 memset(s->outpicref->data[1] + i * s->outpicref->linesize[1], 128, outlink->w);

  362.                 memset(s->outpicref->data[2] + i * s->outpicref->linesize[2], 128, outlink->w);

  363.             }

  364.         }

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

  366.         s->outpicref = NULL;

  367.     }

  368. 

  369.     return ret;

  370. }

  371. 

  372. static int plot_spectrum_column(AVFilterLink *inlink, AVFrame *insamples)

  373. {

  374.     int ret;

  375.     AVFilterContext *ctx = inlink->dst;

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

  377.     ShowSpectrumContext *s = ctx->priv;

  378.     AVFrame *outpicref = s->outpicref;

  379. 

  380.     /* nb_freq contains the power of two superior or equal to the output image

  381.      * height (or half the RDFT window size) */

  382.     const int nb_freq = 1 << (s->rdft_bits - 1);

  383.     const int win_size = nb_freq << 1;

  384.     const double w = s->win_scale;

  385.     int h = s->orientation == VERTICAL ? s->channel_height : s->channel_width;

  386. 

  387.     int ch, plane, n, x, y;

  388. 

  389.     av_assert0(insamples->nb_samples == win_size);

  390. 

  391.     /* fill RDFT input with the number of samples available */

  392.     for (ch = 0; ch < s->nb_display_channels; ch++) {

  393.         const int16_t *p = (int16_t *)insamples->extended_data[ch];

  394. 

  395.         for (n = 0; n < win_size; n++)

  396.             s->rdft_data[ch][n] = p[n] * s->window_func_lut[n];

  397.     }

  398. 

  399.     /* run RDFT on each samples set */

  400.     for (ch = 0; ch < s->nb_display_channels; ch++)

  401.         av_rdft_calc(s->rdft, s->rdft_data[ch]);

  402. 

  403.     /* fill a new spectrum column */

  404. #define RE(y, ch) s->rdft_data[ch][2 * (y) + 0]

  405. #define IM(y, ch) s->rdft_data[ch][2 * (y) + 1]

  406. #define MAGNITUDE(y, ch) hypot(RE(y, ch), IM(y, ch))

  407. 

  408.     /* initialize buffer for combining to black */

  409.     if (s->orientation == VERTICAL) {

  410.         for (y = 0; y < outlink->h; y++) {

  411.             s->combine_buffer[3 * y    ] = 0;

  412.             s->combine_buffer[3 * y + 1] = 127.5;

  413.             s->combine_buffer[3 * y + 2] = 127.5;

  414.         }

  415.     } else {

  416.         for (y = 0; y < outlink->w; y++) {

  417.             s->combine_buffer[3 * y    ] = 0;

  418.             s->combine_buffer[3 * y + 1] = 127.5;

  419.             s->combine_buffer[3 * y + 2] = 127.5;

  420.         }

  421.     }

  422. 

  423.     for (ch = 0; ch < s->nb_display_channels; ch++) {

  424.         float yf, uf, vf;

  425. 

  426.         /* decide color range */

  427.         switch (s->mode) {

  428.         case COMBINED:

  429.             // reduce range by channel count

  430.             yf = 256.0f / s->nb_display_channels;

  431.             switch (s->color_mode) {

  432.             case RAINBOW:

  433.             case MORELAND:

  434.             case NEBULAE:

  435.             case FIRE:

  436.             case INTENSITY:

  437.                 uf = yf;

  438.                 vf = yf;

  439.                 break;

  440.             case CHANNEL:

  441.                 /* adjust saturation for mixed UV coloring */

  442.                 /* this factor is correct for infinite channels, an approximation otherwise */

  443.                 uf = yf * M_PI;

  444.                 vf = yf * M_PI;

  445.                 break;

  446.             default:

  447.                 av_assert0(0);

  448.             }

  449.             break;

  450.         case SEPARATE:

  451.             // full range

  452.             yf = 256.0f;

  453.             uf = 256.0f;

  454.             vf = 256.0f;

  455.             break;

  456.         default:

  457.             av_assert0(0);

  458.         }

  459. 

  460.         if (s->color_mode == CHANNEL) {

  461.             if (s->nb_display_channels > 1) {

  462.                 uf *= 0.5 * sin((2 * M_PI * ch) / s->nb_display_channels);

  463.                 vf *= 0.5 * cos((2 * M_PI * ch) / s->nb_display_channels);

  464.             } else {

  465.                 uf = 0.0f;

  466.                 vf = 0.0f;

  467.             }

  468.         }

  469.         uf *= s->saturation;

  470.         vf *= s->saturation;

  471. 

  472.         /* draw the channel */

  473.         for (y = 0; y < h; y++) {

  474.             int row = (s->mode == COMBINED) ? y : ch * h + y;

  475.             float *out = &s->combine_buffer[3 * row];

  476. 

  477.             /* get magnitude */

  478.             float a = w * MAGNITUDE(y, ch);

  479. 

  480.             /* apply scale */

  481.             switch (s->scale) {

  482.             case LINEAR:

  483.                 break;

  484.             case SQRT:

  485.                 a = sqrt(a);

  486.                 break;

  487.             case CBRT:

  488.                 a = cbrt(a);

  489.                 break;

  490.             case LOG:

  491.                 a = 1 + log10(FFMAX(FFMIN(1, a), 1e-6)) / 5; // zero = -120dBFS

  492.                 break;

  493.             default:

  494.                 av_assert0(0);

  495.             }

  496. 

  497.             if (s->color_mode > CHANNEL) {

  498.                 const int cm = s->color_mode;

  499.                 float y, u, v;

  500.                 int i;

  501. 

  502.                 for (i = 1; i < FF_ARRAY_ELEMS(color_table[cm]) - 1; i++)

  503.                     if (color_table[cm][i].a >= a)

  504.                         break;

  505.                 // i now is the first item >= the color

  506.                 // now we know to interpolate between item i - 1 and i

  507.                 if (a <= color_table[cm][i - 1].a) {

  508.                     y = color_table[cm][i - 1].y;

  509.                     u = color_table[cm][i - 1].u;

  510.                     v = color_table[cm][i - 1].v;

  511.                 } else if (a >= color_table[cm][i].a) {

  512.                     y = color_table[cm][i].y;

  513.                     u = color_table[cm][i].u;

  514.                     v = color_table[cm][i].v;

  515.                 } else {

  516.                     float start = color_table[cm][i - 1].a;

  517.                     float end = color_table[cm][i].a;

  518.                     float lerpfrac = (a - start) / (end - start);

  519.                     y = color_table[cm][i - 1].y * (1.0f - lerpfrac)

  520.                       + color_table[cm][i].y * lerpfrac;

  521.                     u = color_table[cm][i - 1].u * (1.0f - lerpfrac)

  522.                       + color_table[cm][i].u * lerpfrac;

  523.                     v = color_table[cm][i - 1].v * (1.0f - lerpfrac)

  524.                       + color_table[cm][i].v * lerpfrac;

  525.                 }

  526. 

  527.                 out[0] += y * yf;

  528.                 out[1] += u * uf;

  529.                 out[2] += v * vf;

  530.             } else {

  531.                 out[0] += a * yf;

  532.                 out[1] += a * uf;

  533.                 out[2] += a * vf;

  534.             }

  535.         }

  536.     }

  537. 

  538.     av_frame_make_writable(s->outpicref);

  539.     /* copy to output */

  540.     if (s->orientation == VERTICAL) {

  541.         if (s->sliding == SCROLL) {

  542.             for (plane = 0; plane < 3; plane++) {

  543.                 for (y = 0; y < outlink->h; y++) {

  544.                     uint8_t *p = outpicref->data[plane] +

  545.                                  y * outpicref->linesize[plane];

  546.                     memmove(p, p + 1, outlink->w - 1);

  547.                 }

  548.             }

  549.             s->xpos = outlink->w - 1;

  550.         } else if (s->sliding == RSCROLL) {

  551.             for (plane = 0; plane < 3; plane++) {

  552.                 for (y = 0; y < outlink->h; y++) {

  553.                     uint8_t *p = outpicref->data[plane] +

  554.                                  y * outpicref->linesize[plane];

  555.                     memmove(p + 1, p, outlink->w - 1);

  556.                 }

  557.             }

  558.             s->xpos = 0;

  559.         }

  560.         for (plane = 0; plane < 3; plane++) {

  561.             uint8_t *p = outpicref->data[plane] +

  562.                          (outlink->h - 1) * outpicref->linesize[plane] +

  563.                          s->xpos;

  564.             for (y = 0; y < outlink->h; y++) {

  565.                 *p = lrint(FFMAX(0, FFMIN(s->combine_buffer[3 * y + plane], 255)));

  566.                 p -= outpicref->linesize[plane];

  567.             }

  568.         }

  569.     } else {

  570.         if (s->sliding == SCROLL) {

  571.             for (plane = 0; plane < 3; plane++) {

  572.                 for (y = 1; y < outlink->h; y++) {

  573.                     memmove(outpicref->data[plane] + (y-1) * outpicref->linesize[plane],

  574.                             outpicref->data[plane] + (y  ) * outpicref->linesize[plane],

  575.                             outlink->w);

  576.                 }

  577.             }

  578.             s->xpos = outlink->h - 1;

  579.         } else if (s->sliding == RSCROLL) {

  580.             for (plane = 0; plane < 3; plane++) {

  581.                 for (y = outlink->h - 1; y >= 1; y--) {

  582.                     memmove(outpicref->data[plane] + (y  ) * outpicref->linesize[plane],

  583.                             outpicref->data[plane] + (y-1) * outpicref->linesize[plane],

  584.                             outlink->w);

  585.                 }

  586.             }

  587.             s->xpos = 0;

  588.         }

  589.         for (plane = 0; plane < 3; plane++) {

  590.             uint8_t *p = outpicref->data[plane] +

  591.                          s->xpos * outpicref->linesize[plane];

  592.             for (x = 0; x < outlink->w; x++) {

  593.                 *p = lrint(FFMAX(0, FFMIN(s->combine_buffer[3 * x + plane], 255)));

  594.                 p++;

  595.             }

  596.         }

  597.     }

  598. 

  599.     if (s->sliding != FULLFRAME || s->xpos == 0)

  600.         outpicref->pts = insamples->pts;

  601. 

  602.     s->xpos++;

  603.     if (s->orientation == VERTICAL && s->xpos >= outlink->w)

  604.         s->xpos = 0;

  605.     if (s->orientation == HORIZONTAL && s->xpos >= outlink->h)

  606.         s->xpos = 0;

  607.     if (s->sliding != FULLFRAME || s->xpos == 0) {

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

  609.         if (ret < 0)

  610.             return ret;

  611.     }

  612. 

  613.     return win_size;

  614. }

  615. 

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

  617. {

  618.     AVFilterContext *ctx = inlink->dst;

  619.     ShowSpectrumContext *s = ctx->priv;

  620.     unsigned win_size = 1 << s->rdft_bits;

  621.     AVFrame *fin = NULL;

  622.     int ret = 0;

  623. 

  624.     av_audio_fifo_write(s->fifo, (void **)insamples->extended_data, insamples->nb_samples);

  625.     av_frame_free(&insamples);

  626.     while (av_audio_fifo_size(s->fifo) >= win_size) {

  627.         fin = ff_get_audio_buffer(inlink, win_size);

  628.         if (!fin) {

  629.             ret = AVERROR(ENOMEM);

  630.             goto fail;

  631.         }

  632. 

  633.         fin->pts = s->pts;

  634.         s->pts += s->skip_samples;

  635.         ret = av_audio_fifo_peek(s->fifo, (void **)fin->extended_data, win_size);

  636.         if (ret < 0)

  637.             goto fail;

  638. 

  639.         ret = plot_spectrum_column(inlink, fin);

  640.         av_frame_free(&fin);

  641.         av_audio_fifo_drain(s->fifo, s->skip_samples);

  642.         if (ret < 0)

  643.             goto fail;

  644.     }

  645. 

  646. fail:

  647.     av_frame_free(&fin);

  648.     return ret;

  649. }

  650. 

  651. static const AVFilterPad showspectrum_inputs[] = {

  652.     {

  653.         .name         = "default",

  654.         .type         = AVMEDIA_TYPE_AUDIO,

  655.         .filter_frame = filter_frame,

  656.     },

  657.     { NULL }

  658. };

  659. 

  660. static const AVFilterPad showspectrum_outputs[] = {

  661.     {

  662.         .name          = "default",

  663.         .type          = AVMEDIA_TYPE_VIDEO,

  664.         .config_props  = config_output,

  665.         .request_frame = request_frame,

  666.     },

  667.     { NULL }

  668. };

  669. 

  670. AVFilter ff_avf_showspectrum = {

  671.     .name          = "showspectrum",

  672.     .description   = NULL_IF_CONFIG_SMALL("Convert input audio to a spectrum video output."),

  673.     .uninit        = uninit,

  674.     .query_formats = query_formats,

  675.     .priv_size     = sizeof(ShowSpectrumContext),

  676.     .inputs        = showspectrum_inputs,

  677.     .outputs       = showspectrum_outputs,

  678.     .priv_class    = &showspectrum_class,

  679. };