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.
51196 Commits
32 Branches
444MB
Tree: 4257b804e2
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from '4257b804e2'
${ noResults }
FFmpeg/libavfilter/f_ebur128.c

823 lines
31KB
Raw Blame History 

  1. /*

  2.  * Copyright (c) 2012 Clément Bœsch

  3.  *

  4.  * This file is part of FFmpeg.

  5.  *

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

  7.  * it under the terms of the GNU General Public License as published by

  8.  * the Free Software Foundation; either version 2 of the License, or

  9.  * (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

  14.  * GNU General Public License for more details.

  15.  *

  16.  * You should have received a copy of the GNU General Public License along

  17.  * with FFmpeg; if not, write to the Free Software Foundation, Inc.,

  18.  * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.

  19.  */

  20. 

  21. /**

  22.  * @file

  23.  * EBU R.128 implementation

  24.  * @see http://tech.ebu.ch/loudness

  25.  * @see https://www.youtube.com/watch?v=iuEtQqC-Sqo "EBU R128 Introduction - Florian Camerer"

  26.  * @todo True Peak

  27.  * @todo implement start/stop/reset through filter command injection

  28.  * @todo support other frequencies to avoid resampling

  29.  */

  30. 

  31. #include <math.h>

  32. 

  33. #include "libavutil/avassert.h"

  34. #include "libavutil/avstring.h"

  35. #include "libavutil/channel_layout.h"

  36. #include "libavutil/dict.h"

  37. #include "libavutil/xga_font_data.h"

  38. #include "libavutil/opt.h"

  39. #include "libavutil/timestamp.h"

  40. #include "audio.h"

  41. #include "avfilter.h"

  42. #include "formats.h"

  43. #include "internal.h"

  44. 

  45. #define MAX_CHANNELS 63

  46. 

  47. /* pre-filter coefficients */

  48. #define PRE_B0  1.53512485958697

  49. #define PRE_B1 -2.69169618940638

  50. #define PRE_B2  1.19839281085285

  51. #define PRE_A1 -1.69065929318241

  52. #define PRE_A2  0.73248077421585

  53. 

  54. /* RLB-filter coefficients */

  55. #define RLB_B0  1.0

  56. #define RLB_B1 -2.0

  57. #define RLB_B2  1.0

  58. #define RLB_A1 -1.99004745483398

  59. #define RLB_A2  0.99007225036621

  60. 

  61. #define ABS_THRES    -70            ///< silence gate: we discard anything below this absolute (LUFS) threshold

  62. #define ABS_UP_THRES  10            ///< upper loud limit to consider (ABS_THRES being the minimum)

  63. #define HIST_GRAIN   100            ///< defines histogram precision

  64. #define HIST_SIZE  ((ABS_UP_THRES - ABS_THRES) * HIST_GRAIN + 1)

  65. 

  66. /**

  67.  * An histogram is an array of HIST_SIZE hist_entry storing all the energies

  68.  * recorded (with an accuracy of 1/HIST_GRAIN) of the loudnesses from ABS_THRES

  69.  * (at 0) to ABS_UP_THRES (at HIST_SIZE-1).

  70.  * This fixed-size system avoids the need of a list of energies growing

  71.  * infinitely over the time and is thus more scalable.

  72.  */

  73. struct hist_entry {

  74.     int count;                      ///< how many times the corresponding value occurred

  75.     double energy;                  ///< E = 10^((L + 0.691) / 10)

  76.     double loudness;                ///< L = -0.691 + 10 * log10(E)

  77. };

  78. 

  79. struct integrator {

  80.     double *cache[MAX_CHANNELS];    ///< window of filtered samples (N ms)

  81.     int cache_pos;                  ///< focus on the last added bin in the cache array

  82.     double sum[MAX_CHANNELS];       ///< sum of the last N ms filtered samples (cache content)

  83.     int filled;                     ///< 1 if the cache is completely filled, 0 otherwise

  84.     double rel_threshold;           ///< relative threshold

  85.     double sum_kept_powers;         ///< sum of the powers (weighted sums) above absolute threshold

  86.     int nb_kept_powers;             ///< number of sum above absolute threshold

  87.     struct hist_entry *histogram;   ///< histogram of the powers, used to compute LRA and I

  88. };

  89. 

  90. struct rect { int x, y, w, h; };

  91. 

  92. typedef struct {

  93.     const AVClass *class;           ///< AVClass context for log and options purpose

  94. 

  95.     /* video  */

  96.     int do_video;                   ///< 1 if video output enabled, 0 otherwise

  97.     int w, h;                       ///< size of the video output

  98.     struct rect text;               ///< rectangle for the LU legend on the left

  99.     struct rect graph;              ///< rectangle for the main graph in the center

  100.     struct rect gauge;              ///< rectangle for the gauge on the right

  101.     AVFrame *outpicref;             ///< output picture reference, updated regularly

  102.     int meter;                      ///< select a EBU mode between +9 and +18

  103.     int scale_range;                ///< the range of LU values according to the meter

  104.     int y_zero_lu;                  ///< the y value (pixel position) for 0 LU

  105.     int *y_line_ref;                ///< y reference values for drawing the LU lines in the graph and the gauge

  106. 

  107.     /* audio */

  108.     int nb_channels;                ///< number of channels in the input

  109.     double *ch_weighting;           ///< channel weighting mapping

  110.     int sample_count;               ///< sample count used for refresh frequency, reset at refresh

  111. 

  112.     /* Filter caches.

  113.      * The mult by 3 in the following is for X[i], X[i-1] and X[i-2] */

  114.     double x[MAX_CHANNELS * 3];     ///< 3 input samples cache for each channel

  115.     double y[MAX_CHANNELS * 3];     ///< 3 pre-filter samples cache for each channel

  116.     double z[MAX_CHANNELS * 3];     ///< 3 RLB-filter samples cache for each channel

  117. 

  118. #define I400_BINS  (48000 * 4 / 10)

  119. #define I3000_BINS (48000 * 3)

  120.     struct integrator i400;         ///< 400ms integrator, used for Momentary loudness  (M), and Integrated loudness (I)

  121.     struct integrator i3000;        ///<    3s integrator, used for Short term loudness (S), and Loudness Range      (LRA)

  122. 

  123.     /* I and LRA specific */

  124.     double integrated_loudness;     ///< integrated loudness in LUFS (I)

  125.     double loudness_range;          ///< loudness range in LU (LRA)

  126.     double lra_low, lra_high;       ///< low and high LRA values

  127. 

  128.     /* misc */

  129.     int loglevel;                   ///< log level for frame logging

  130.     int metadata;                   ///< whether or not to inject loudness results in frames

  131.     int request_fulfilled;          ///< 1 if some audio just got pushed, 0 otherwise. FIXME: remove me

  132. } EBUR128Context;

  133. 

  134. #define OFFSET(x) offsetof(EBUR128Context, x)

  135. #define A AV_OPT_FLAG_AUDIO_PARAM

  136. #define V AV_OPT_FLAG_VIDEO_PARAM

  137. #define F AV_OPT_FLAG_FILTERING_PARAM

  138. static const AVOption ebur128_options[] = {

  139.     { "video", "set video output", OFFSET(do_video), AV_OPT_TYPE_INT, {.i64 = 0}, 0, 1, V|F },

  140.     { "size",  "set video size",   OFFSET(w), AV_OPT_TYPE_IMAGE_SIZE, {.str = "640x480"}, 0, 0, V|F },

  141.     { "meter", "set scale meter (+9 to +18)",  OFFSET(meter), AV_OPT_TYPE_INT, {.i64 = 9}, 9, 18, V|F },

  142.     { "framelog", "force frame logging level", OFFSET(loglevel), AV_OPT_TYPE_INT, {.i64 = -1},   INT_MIN, INT_MAX, A|V|F, "level" },

  143.         { "info",    "information logging level", 0, AV_OPT_TYPE_CONST, {.i64 = AV_LOG_INFO},    INT_MIN, INT_MAX, A|V|F, "level" },

  144.         { "verbose", "verbose logging level",     0, AV_OPT_TYPE_CONST, {.i64 = AV_LOG_VERBOSE}, INT_MIN, INT_MAX, A|V|F, "level" },

  145.     { "metadata", "inject metadata in the filtergraph", OFFSET(metadata), AV_OPT_TYPE_INT, {.i64 = 0}, 0, 1, A|V|F },

  146.     { NULL },

  147. };

  148. 

  149. AVFILTER_DEFINE_CLASS(ebur128);

  150. 

  151. static const uint8_t graph_colors[] = {

  152.     0xdd, 0x66, 0x66,   // value above 0LU non reached

  153.     0x66, 0x66, 0xdd,   // value below 0LU non reached

  154.     0x96, 0x33, 0x33,   // value above 0LU reached

  155.     0x33, 0x33, 0x96,   // value below 0LU reached

  156.     0xdd, 0x96, 0x96,   // value above 0LU line non reached

  157.     0x96, 0x96, 0xdd,   // value below 0LU line non reached

  158.     0xdd, 0x33, 0x33,   // value above 0LU line reached

  159.     0x33, 0x33, 0xdd,   // value below 0LU line reached

  160. };

  161. 

  162. static const uint8_t *get_graph_color(const EBUR128Context *ebur128, int v, int y)

  163. {

  164.     const int below0  = y > ebur128->y_zero_lu;

  165.     const int reached = y >= v;

  166.     const int line    = ebur128->y_line_ref[y] || y == ebur128->y_zero_lu;

  167.     const int colorid = 4*line + 2*reached + below0;

  168.     return graph_colors + 3*colorid;

  169. }

  170. 

  171. static inline int lu_to_y(const EBUR128Context *ebur128, double v)

  172. {

  173.     v += 2 * ebur128->meter;                            // make it in range [0;...]

  174.     v  = av_clipf(v, 0, ebur128->scale_range);          // make sure it's in the graph scale

  175.     v  = ebur128->scale_range - v;                      // invert value (y=0 is on top)

  176.     return v * ebur128->graph.h / ebur128->scale_range; // rescale from scale range to px height

  177. }

  178. 

  179. #define FONT8   0

  180. #define FONT16  1

  181. 

  182. static const uint8_t font_colors[] = {

  183.     0xdd, 0xdd, 0x00,

  184.     0x00, 0x96, 0x96,

  185. };

  186. 

  187. static void drawtext(AVFrame *pic, int x, int y, int ftid, const uint8_t *color, const char *fmt, ...)

  188. {

  189.     int i;

  190.     char buf[128] = {0};

  191.     const uint8_t *font;

  192.     int font_height;

  193.     va_list vl;

  194. 

  195.     if      (ftid == FONT16) font = avpriv_vga16_font, font_height = 16;

  196.     else if (ftid == FONT8)  font = avpriv_cga_font,   font_height =  8;

  197.     else return;

  198. 

  199.     va_start(vl, fmt);

  200.     vsnprintf(buf, sizeof(buf), fmt, vl);

  201.     va_end(vl);

  202. 

  203.     for (i = 0; buf[i]; i++) {

  204.         int char_y, mask;

  205.         uint8_t *p = pic->data[0] + y*pic->linesize[0] + (x + i*8)*3;

  206. 

  207.         for (char_y = 0; char_y < font_height; char_y++) {

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

  209.                 if (font[buf[i] * font_height + char_y] & mask)

  210.                     memcpy(p, color, 3);

  211.                 else

  212.                     memcpy(p, "\x00\x00\x00", 3);

  213.                 p += 3;

  214.             }

  215.             p += pic->linesize[0] - 8*3;

  216.         }

  217.     }

  218. }

  219. 

  220. static void drawline(AVFrame *pic, int x, int y, int len, int step)

  221. {

  222.     int i;

  223.     uint8_t *p = pic->data[0] + y*pic->linesize[0] + x*3;

  224. 

  225.     for (i = 0; i < len; i++) {

  226.         memcpy(p, "\x00\xff\x00", 3);

  227.         p += step;

  228.     }

  229. }

  230. 

  231. static int config_video_output(AVFilterLink *outlink)

  232. {

  233.     int i, x, y;

  234.     uint8_t *p;

  235.     AVFilterContext *ctx = outlink->src;

  236.     EBUR128Context *ebur128 = ctx->priv;

  237.     AVFrame *outpicref;

  238. 

  239.     /* check if there is enough space to represent everything decently */

  240.     if (ebur128->w < 640 || ebur128->h < 480) {

  241.         av_log(ctx, AV_LOG_ERROR, "Video size %dx%d is too small, "

  242.                "minimum size is 640x480\n", ebur128->w, ebur128->h);

  243.         return AVERROR(EINVAL);

  244.     }

  245.     outlink->w = ebur128->w;

  246.     outlink->h = ebur128->h;

  247. 

  248. #define PAD 8

  249. 

  250.     /* configure text area position and size */

  251.     ebur128->text.x  = PAD;

  252.     ebur128->text.y  = 40;

  253.     ebur128->text.w  = 3 * 8;   // 3 characters

  254.     ebur128->text.h  = ebur128->h - PAD - ebur128->text.y;

  255. 

  256.     /* configure gauge position and size */

  257.     ebur128->gauge.w = 20;

  258.     ebur128->gauge.h = ebur128->text.h;

  259.     ebur128->gauge.x = ebur128->w - PAD - ebur128->gauge.w;

  260.     ebur128->gauge.y = ebur128->text.y;

  261. 

  262.     /* configure graph position and size */

  263.     ebur128->graph.x = ebur128->text.x + ebur128->text.w + PAD;

  264.     ebur128->graph.y = ebur128->gauge.y;

  265.     ebur128->graph.w = ebur128->gauge.x - ebur128->graph.x - PAD;

  266.     ebur128->graph.h = ebur128->gauge.h;

  267. 

  268.     /* graph and gauge share the LU-to-pixel code */

  269.     av_assert0(ebur128->graph.h == ebur128->gauge.h);

  270. 

  271.     /* prepare the initial picref buffer */

  272.     av_frame_free(&ebur128->outpicref);

  273.     ebur128->outpicref = outpicref =

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

  275.     if (!outpicref)

  276.         return AVERROR(ENOMEM);

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

  278. 

  279.     /* init y references values (to draw LU lines) */

  280.     ebur128->y_line_ref = av_calloc(ebur128->graph.h + 1, sizeof(*ebur128->y_line_ref));

  281.     if (!ebur128->y_line_ref)

  282.         return AVERROR(ENOMEM);

  283. 

  284.     /* black background */

  285.     memset(outpicref->data[0], 0, ebur128->h * outpicref->linesize[0]);

  286. 

  287.     /* draw LU legends */

  288.     drawtext(outpicref, PAD, PAD+16, FONT8, font_colors+3, " LU");

  289.     for (i = ebur128->meter; i >= -ebur128->meter * 2; i--) {

  290.         y = lu_to_y(ebur128, i);

  291.         x = PAD + (i < 10 && i > -10) * 8;

  292.         ebur128->y_line_ref[y] = i;

  293.         y -= 4; // -4 to center vertically

  294.         drawtext(outpicref, x, y + ebur128->graph.y, FONT8, font_colors+3,

  295.                  "%c%d", i < 0 ? '-' : i > 0 ? '+' : ' ', FFABS(i));

  296.     }

  297. 

  298.     /* draw graph */

  299.     ebur128->y_zero_lu = lu_to_y(ebur128, 0);

  300.     p = outpicref->data[0] + ebur128->graph.y * outpicref->linesize[0]

  301.                            + ebur128->graph.x * 3;

  302.     for (y = 0; y < ebur128->graph.h; y++) {

  303.         const uint8_t *c = get_graph_color(ebur128, INT_MAX, y);

  304. 

  305.         for (x = 0; x < ebur128->graph.w; x++)

  306.             memcpy(p + x*3, c, 3);

  307.         p += outpicref->linesize[0];

  308.     }

  309. 

  310.     /* draw fancy rectangles around the graph and the gauge */

  311. #define DRAW_RECT(r) do { \

  312.     drawline(outpicref, r.x,       r.y - 1,   r.w, 3); \

  313.     drawline(outpicref, r.x,       r.y + r.h, r.w, 3); \

  314.     drawline(outpicref, r.x - 1,   r.y,       r.h, outpicref->linesize[0]); \

  315.     drawline(outpicref, r.x + r.w, r.y,       r.h, outpicref->linesize[0]); \

  316. } while (0)

  317.     DRAW_RECT(ebur128->graph);

  318.     DRAW_RECT(ebur128->gauge);

  319. 

  320.     return 0;

  321. }

  322. 

  323. static int config_audio_input(AVFilterLink *inlink)

  324. {

  325.     AVFilterContext *ctx = inlink->dst;

  326.     EBUR128Context *ebur128 = ctx->priv;

  327. 

  328.     /* force 100ms framing in case of metadata injection: the frames must have

  329.      * a granularity of the window overlap to be accurately exploited */

  330.     if (ebur128->metadata)

  331.         inlink->min_samples =

  332.         inlink->max_samples =

  333.         inlink->partial_buf_size = inlink->sample_rate / 10;

  334.     return 0;

  335. }

  336. 

  337. static int config_audio_output(AVFilterLink *outlink)

  338. {

  339.     int i;

  340.     int idx_bitposn = 0;

  341.     AVFilterContext *ctx = outlink->src;

  342.     EBUR128Context *ebur128 = ctx->priv;

  343.     const int nb_channels = av_get_channel_layout_nb_channels(outlink->channel_layout);

  344. 

  345. #define BACK_MASK (AV_CH_BACK_LEFT    |AV_CH_BACK_CENTER    |AV_CH_BACK_RIGHT| \

  346.                    AV_CH_TOP_BACK_LEFT|AV_CH_TOP_BACK_CENTER|AV_CH_TOP_BACK_RIGHT| \

  347.                    AV_CH_SIDE_LEFT                          |AV_CH_SIDE_RIGHT| \

  348.                    AV_CH_SURROUND_DIRECT_LEFT               |AV_CH_SURROUND_DIRECT_RIGHT)

  349. 

  350.     ebur128->nb_channels  = nb_channels;

  351.     ebur128->ch_weighting = av_calloc(nb_channels, sizeof(*ebur128->ch_weighting));

  352.     if (!ebur128->ch_weighting)

  353.         return AVERROR(ENOMEM);

  354. 

  355.     for (i = 0; i < nb_channels; i++) {

  356. 

  357.         /* find the next bit that is set starting from the right */

  358.         while ((outlink->channel_layout & 1ULL<<idx_bitposn) == 0 && idx_bitposn < 63)

  359.             idx_bitposn++;

  360. 

  361.         /* channel weighting */

  362.         if ((1ULL<<idx_bitposn & AV_CH_LOW_FREQUENCY) ||

  363.             (1ULL<<idx_bitposn & AV_CH_LOW_FREQUENCY_2)) {

  364.             ebur128->ch_weighting[i] = 0;

  365.         } else if (1ULL<<idx_bitposn & BACK_MASK) {

  366.             ebur128->ch_weighting[i] = 1.41;

  367.         } else {

  368.             ebur128->ch_weighting[i] = 1.0;

  369.         }

  370. 

  371.         idx_bitposn++;

  372. 

  373.         if (!ebur128->ch_weighting[i])

  374.             continue;

  375. 

  376.         /* bins buffer for the two integration window (400ms and 3s) */

  377.         ebur128->i400.cache[i]  = av_calloc(I400_BINS,  sizeof(*ebur128->i400.cache[0]));

  378.         ebur128->i3000.cache[i] = av_calloc(I3000_BINS, sizeof(*ebur128->i3000.cache[0]));

  379.         if (!ebur128->i400.cache[i] || !ebur128->i3000.cache[i])

  380.             return AVERROR(ENOMEM);

  381.     }

  382. 

  383.     return 0;

  384. }

  385. 

  386. #define ENERGY(loudness) (pow(10, ((loudness) + 0.691) / 10.))

  387. #define LOUDNESS(energy) (-0.691 + 10 * log10(energy))

  388. 

  389. static struct hist_entry *get_histogram(void)

  390. {

  391.     int i;

  392.     struct hist_entry *h = av_calloc(HIST_SIZE, sizeof(*h));

  393. 

  394.     if (!h)

  395.         return NULL;

  396.     for (i = 0; i < HIST_SIZE; i++) {

  397.         h[i].loudness = i / (double)HIST_GRAIN + ABS_THRES;

  398.         h[i].energy   = ENERGY(h[i].loudness);

  399.     }

  400.     return h;

  401. }

  402. 

  403. /* This is currently necessary for the min/max samples to work properly.

  404.  * FIXME: remove me when possible */

  405. static int audio_request_frame(AVFilterLink *outlink)

  406. {

  407.     int ret;

  408.     AVFilterContext *ctx = outlink->src;

  409.     EBUR128Context *ebur128 = ctx->priv;

  410. 

  411.     ebur128->request_fulfilled = 0;

  412.     do {

  413.         ret = ff_request_frame(ctx->inputs[0]);

  414.     } while (!ebur128->request_fulfilled && ret >= 0);

  415.     return ret;

  416. }

  417. 

  418. static av_cold int init(AVFilterContext *ctx, const char *args)

  419. {

  420.     int ret;

  421.     EBUR128Context *ebur128 = ctx->priv;

  422.     AVFilterPad pad;

  423. 

  424.     ebur128->class = &ebur128_class;

  425.     av_opt_set_defaults(ebur128);

  426. 

  427.     if ((ret = av_set_options_string(ebur128, args, "=", ":")) < 0)

  428.         return ret;

  429. 

  430.     if (ebur128->loglevel != AV_LOG_INFO &&

  431.         ebur128->loglevel != AV_LOG_VERBOSE) {

  432.         if (ebur128->do_video || ebur128->metadata)

  433.             ebur128->loglevel = AV_LOG_VERBOSE;

  434.         else

  435.             ebur128->loglevel = AV_LOG_INFO;

  436.     }

  437. 

  438.     // if meter is  +9 scale, scale range is from -18 LU to  +9 LU (or 3*9)

  439.     // if meter is +18 scale, scale range is from -36 LU to +18 LU (or 3*18)

  440.     ebur128->scale_range = 3 * ebur128->meter;

  441. 

  442.     ebur128->i400.histogram  = get_histogram();

  443.     ebur128->i3000.histogram = get_histogram();

  444.     if (!ebur128->i400.histogram || !ebur128->i3000.histogram)

  445.         return AVERROR(ENOMEM);

  446. 

  447.     ebur128->integrated_loudness = ABS_THRES;

  448.     ebur128->loudness_range = 0;

  449. 

  450.     /* insert output pads */

  451.     if (ebur128->do_video) {

  452.         pad = (AVFilterPad){

  453.             .name         = av_strdup("out0"),

  454.             .type         = AVMEDIA_TYPE_VIDEO,

  455.             .config_props = config_video_output,

  456.         };

  457.         if (!pad.name)

  458.             return AVERROR(ENOMEM);

  459.         ff_insert_outpad(ctx, 0, &pad);

  460.     }

  461.     pad = (AVFilterPad){

  462.         .name         = av_asprintf("out%d", ebur128->do_video),

  463.         .type         = AVMEDIA_TYPE_AUDIO,

  464.         .config_props = config_audio_output,

  465.     };

  466.     if (ebur128->metadata)

  467.         pad.request_frame = audio_request_frame;

  468.     if (!pad.name)

  469.         return AVERROR(ENOMEM);

  470.     ff_insert_outpad(ctx, ebur128->do_video, &pad);

  471. 

  472.     /* summary */

  473.     av_log(ctx, AV_LOG_VERBOSE, "EBU +%d scale\n", ebur128->meter);

  474. 

  475.     return 0;

  476. }

  477. 

  478. #define HIST_POS(power) (int)(((power) - ABS_THRES) * HIST_GRAIN)

  479. 

  480. /* loudness and power should be set such as loudness = -0.691 +

  481.  * 10*log10(power), we just avoid doing that calculus two times */

  482. static int gate_update(struct integrator *integ, double power,

  483.                        double loudness, int gate_thres)

  484. {

  485.     int ipower;

  486.     double relative_threshold;

  487.     int gate_hist_pos;

  488. 

  489.     /* update powers histograms by incrementing current power count */

  490.     ipower = av_clip(HIST_POS(loudness), 0, HIST_SIZE - 1);

  491.     integ->histogram[ipower].count++;

  492. 

  493.     /* compute relative threshold and get its position in the histogram */

  494.     integ->sum_kept_powers += power;

  495.     integ->nb_kept_powers++;

  496.     relative_threshold = integ->sum_kept_powers / integ->nb_kept_powers;

  497.     if (!relative_threshold)

  498.         relative_threshold = 1e-12;

  499.     integ->rel_threshold = LOUDNESS(relative_threshold) + gate_thres;

  500.     gate_hist_pos = av_clip(HIST_POS(integ->rel_threshold), 0, HIST_SIZE - 1);

  501. 

  502.     return gate_hist_pos;

  503. }

  504. 

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

  506. {

  507.     int i, ch, idx_insample;

  508.     AVFilterContext *ctx = inlink->dst;

  509.     EBUR128Context *ebur128 = ctx->priv;

  510.     const int nb_channels = ebur128->nb_channels;

  511.     const int nb_samples  = insamples->nb_samples;

  512.     const double *samples = (double *)insamples->data[0];

  513.     AVFrame *pic = ebur128->outpicref;

  514. 

  515.     for (idx_insample = 0; idx_insample < nb_samples; idx_insample++) {

  516.         const int bin_id_400  = ebur128->i400.cache_pos;

  517.         const int bin_id_3000 = ebur128->i3000.cache_pos;

  518. 

  519. #define MOVE_TO_NEXT_CACHED_ENTRY(time) do {                \

  520.     ebur128->i##time.cache_pos++;                           \

  521.     if (ebur128->i##time.cache_pos == I##time##_BINS) {     \

  522.         ebur128->i##time.filled    = 1;                     \

  523.         ebur128->i##time.cache_pos = 0;                     \

  524.     }                                                       \

  525. } while (0)

  526. 

  527.         MOVE_TO_NEXT_CACHED_ENTRY(400);

  528.         MOVE_TO_NEXT_CACHED_ENTRY(3000);

  529. 

  530.         for (ch = 0; ch < nb_channels; ch++) {

  531.             double bin;

  532. 

  533.             ebur128->x[ch * 3] = *samples++; // set X[i]

  534. 

  535.             if (!ebur128->ch_weighting[ch])

  536.                 continue;

  537. 

  538.             /* Y[i] = X[i]*b0 + X[i-1]*b1 + X[i-2]*b2 - Y[i-1]*a1 - Y[i-2]*a2 */

  539. #define FILTER(Y, X, name) do {                                                 \

  540.             double *dst = ebur128->Y + ch*3;                                    \

  541.             double *src = ebur128->X + ch*3;                                    \

  542.             dst[2] = dst[1];                                                    \

  543.             dst[1] = dst[0];                                                    \

  544.             dst[0] = src[0]*name##_B0 + src[1]*name##_B1 + src[2]*name##_B2     \

  545.                                       - dst[1]*name##_A1 - dst[2]*name##_A2;    \

  546. } while (0)

  547. 

  548.             // TODO: merge both filters in one?

  549.             FILTER(y, x, PRE);  // apply pre-filter

  550.             ebur128->x[ch * 3 + 2] = ebur128->x[ch * 3 + 1];

  551.             ebur128->x[ch * 3 + 1] = ebur128->x[ch * 3    ];

  552.             FILTER(z, y, RLB);  // apply RLB-filter

  553. 

  554.             bin = ebur128->z[ch * 3] * ebur128->z[ch * 3];

  555. 

  556.             /* add the new value, and limit the sum to the cache size (400ms or 3s)

  557.              * by removing the oldest one */

  558.             ebur128->i400.sum [ch] = ebur128->i400.sum [ch] + bin - ebur128->i400.cache [ch][bin_id_400];

  559.             ebur128->i3000.sum[ch] = ebur128->i3000.sum[ch] + bin - ebur128->i3000.cache[ch][bin_id_3000];

  560. 

  561.             /* override old cache entry with the new value */

  562.             ebur128->i400.cache [ch][bin_id_400 ] = bin;

  563.             ebur128->i3000.cache[ch][bin_id_3000] = bin;

  564.         }

  565. 

  566.         /* For integrated loudness, gating blocks are 400ms long with 75%

  567.          * overlap (see BS.1770-2 p5), so a re-computation is needed each 100ms

  568.          * (4800 samples at 48kHz). */

  569.         if (++ebur128->sample_count == 4800) {

  570.             double loudness_400, loudness_3000;

  571.             double power_400 = 1e-12, power_3000 = 1e-12;

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

  573.             const int64_t pts = insamples->pts +

  574.                 av_rescale_q(idx_insample, (AVRational){ 1, inlink->sample_rate },

  575.                              outlink->time_base);

  576. 

  577.             ebur128->sample_count = 0;

  578. 

  579. #define COMPUTE_LOUDNESS(m, time) do {                                              \

  580.     if (ebur128->i##time.filled) {                                                  \

  581.         /* weighting sum of the last <time> ms */                                   \

  582.         for (ch = 0; ch < nb_channels; ch++)                                        \

  583.             power_##time += ebur128->ch_weighting[ch] * ebur128->i##time.sum[ch];   \

  584.         power_##time /= I##time##_BINS;                                             \

  585.     }                                                                               \

  586.     loudness_##time = LOUDNESS(power_##time);                                       \

  587. } while (0)

  588. 

  589.             COMPUTE_LOUDNESS(M,  400);

  590.             COMPUTE_LOUDNESS(S, 3000);

  591. 

  592.             /* Integrated loudness */

  593. #define I_GATE_THRES -10  // initially defined to -8 LU in the first EBU standard

  594. 

  595.             if (loudness_400 >= ABS_THRES) {

  596.                 double integrated_sum = 0;

  597.                 int nb_integrated = 0;

  598.                 int gate_hist_pos = gate_update(&ebur128->i400, power_400,

  599.                                                 loudness_400, I_GATE_THRES);

  600. 

  601.                 /* compute integrated loudness by summing the histogram values

  602.                  * above the relative threshold */

  603.                 for (i = gate_hist_pos; i < HIST_SIZE; i++) {

  604.                     const int nb_v = ebur128->i400.histogram[i].count;

  605.                     nb_integrated  += nb_v;

  606.                     integrated_sum += nb_v * ebur128->i400.histogram[i].energy;

  607.                 }

  608.                 if (nb_integrated)

  609.                     ebur128->integrated_loudness = LOUDNESS(integrated_sum / nb_integrated);

  610.             }

  611. 

  612.             /* LRA */

  613. #define LRA_GATE_THRES -20

  614. #define LRA_LOWER_PRC   10

  615. #define LRA_HIGHER_PRC  95

  616. 

  617.             /* XXX: example code in EBU 3342 is ">=" but formula in BS.1770

  618.              * specs is ">" */

  619.             if (loudness_3000 >= ABS_THRES) {

  620.                 int nb_powers = 0;

  621.                 int gate_hist_pos = gate_update(&ebur128->i3000, power_3000,

  622.                                                 loudness_3000, LRA_GATE_THRES);

  623. 

  624.                 for (i = gate_hist_pos; i < HIST_SIZE; i++)

  625.                     nb_powers += ebur128->i3000.histogram[i].count;

  626.                 if (nb_powers) {

  627.                     int n, nb_pow;

  628. 

  629.                     /* get lower loudness to consider */

  630.                     n = 0;

  631.                     nb_pow = LRA_LOWER_PRC  * nb_powers / 100. + 0.5;

  632.                     for (i = gate_hist_pos; i < HIST_SIZE; i++) {

  633.                         n += ebur128->i3000.histogram[i].count;

  634.                         if (n >= nb_pow) {

  635.                             ebur128->lra_low = ebur128->i3000.histogram[i].loudness;

  636.                             break;

  637.                         }

  638.                     }

  639. 

  640.                     /* get higher loudness to consider */

  641.                     n = nb_powers;

  642.                     nb_pow = LRA_HIGHER_PRC * nb_powers / 100. + 0.5;

  643.                     for (i = HIST_SIZE - 1; i >= 0; i--) {

  644.                         n -= ebur128->i3000.histogram[i].count;

  645.                         if (n < nb_pow) {

  646.                             ebur128->lra_high = ebur128->i3000.histogram[i].loudness;

  647.                             break;

  648.                         }

  649.                     }

  650. 

  651.                     // XXX: show low & high on the graph?

  652.                     ebur128->loudness_range = ebur128->lra_high - ebur128->lra_low;

  653.                 }

  654.             }

  655. 

  656. #define LOG_FMT "M:%6.1f S:%6.1f     I:%6.1f LUFS     LRA:%6.1f LU"

  657. 

  658.             /* push one video frame */

  659.             if (ebur128->do_video) {

  660.                 int x, y, ret;

  661.                 uint8_t *p;

  662. 

  663.                 const int y_loudness_lu_graph = lu_to_y(ebur128, loudness_3000 + 23);

  664.                 const int y_loudness_lu_gauge = lu_to_y(ebur128, loudness_400  + 23);

  665. 

  666.                 /* draw the graph using the short-term loudness */

  667.                 p = pic->data[0] + ebur128->graph.y*pic->linesize[0] + ebur128->graph.x*3;

  668.                 for (y = 0; y < ebur128->graph.h; y++) {

  669.                     const uint8_t *c = get_graph_color(ebur128, y_loudness_lu_graph, y);

  670. 

  671.                     memmove(p, p + 3, (ebur128->graph.w - 1) * 3);

  672.                     memcpy(p + (ebur128->graph.w - 1) * 3, c, 3);

  673.                     p += pic->linesize[0];

  674.                 }

  675. 

  676.                 /* draw the gauge using the momentary loudness */

  677.                 p = pic->data[0] + ebur128->gauge.y*pic->linesize[0] + ebur128->gauge.x*3;

  678.                 for (y = 0; y < ebur128->gauge.h; y++) {

  679.                     const uint8_t *c = get_graph_color(ebur128, y_loudness_lu_gauge, y);

  680. 

  681.                     for (x = 0; x < ebur128->gauge.w; x++)

  682.                         memcpy(p + x*3, c, 3);

  683.                     p += pic->linesize[0];

  684.                 }

  685. 

  686.                 /* draw textual info */

  687.                 drawtext(pic, PAD, PAD - PAD/2, FONT16, font_colors,

  688.                          LOG_FMT "     ", // padding to erase trailing characters

  689.                          loudness_400, loudness_3000,

  690.                          ebur128->integrated_loudness, ebur128->loudness_range);

  691. 

  692.                 /* set pts and push frame */

  693.                 pic->pts = pts;

  694.                 ret = ff_filter_frame(outlink, av_frame_clone(pic));

  695.                 if (ret < 0)

  696.                     return ret;

  697.             }

  698. 

  699.             if (ebur128->metadata) { /* happens only once per filter_frame call */

  700.                 char metabuf[128];

  701. #define SET_META(name, var) do {                                            \

  702.     snprintf(metabuf, sizeof(metabuf), "%.3f", var);                        \

  703.     av_dict_set(&insamples->metadata, "lavfi.r128." name, metabuf, 0);      \

  704. } while (0)

  705.                 SET_META("M",        loudness_400);

  706.                 SET_META("S",        loudness_3000);

  707.                 SET_META("I",        ebur128->integrated_loudness);

  708.                 SET_META("LRA",      ebur128->loudness_range);

  709.                 SET_META("LRA.low",  ebur128->lra_low);

  710.                 SET_META("LRA.high", ebur128->lra_high);

  711.             }

  712. 

  713.             av_log(ctx, ebur128->loglevel, "t: %-10s " LOG_FMT "\n",

  714.                    av_ts2timestr(pts, &outlink->time_base),

  715.                    loudness_400, loudness_3000,

  716.                    ebur128->integrated_loudness, ebur128->loudness_range);

  717.         }

  718.     }

  719. 

  720.     ebur128->request_fulfilled = 1;

  721.     return ff_filter_frame(ctx->outputs[ebur128->do_video], insamples);

  722. }

  723. 

  724. static int query_formats(AVFilterContext *ctx)

  725. {

  726.     EBUR128Context *ebur128 = ctx->priv;

  727.     AVFilterFormats *formats;

  728.     AVFilterChannelLayouts *layouts;

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

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

  731. 

  732.     static const enum AVSampleFormat sample_fmts[] = { AV_SAMPLE_FMT_DBL, AV_SAMPLE_FMT_NONE };

  733.     static const int input_srate[] = {48000, -1}; // ITU-R BS.1770 provides coeff only for 48kHz

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

  735. 

  736.     /* set optional output video format */

  737.     if (ebur128->do_video) {

  738.         formats = ff_make_format_list(pix_fmts);

  739.         if (!formats)

  740.             return AVERROR(ENOMEM);

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

  742.         outlink = ctx->outputs[1];

  743.     }

  744. 

  745.     /* set input and output audio formats

  746.      * Note: ff_set_common_* functions are not used because they affect all the

  747.      * links, and thus break the video format negociation */

  748.     formats = ff_make_format_list(sample_fmts);

  749.     if (!formats)

  750.         return AVERROR(ENOMEM);

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

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

  753. 

  754.     layouts = ff_all_channel_layouts();

  755.     if (!layouts)

  756.         return AVERROR(ENOMEM);

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

  758.     ff_channel_layouts_ref(layouts, &outlink->in_channel_layouts);

  759. 

  760.     formats = ff_make_format_list(input_srate);

  761.     if (!formats)

  762.         return AVERROR(ENOMEM);

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

  764.     ff_formats_ref(formats, &outlink->in_samplerates);

  765. 

  766.     return 0;

  767. }

  768. 

  769. static av_cold void uninit(AVFilterContext *ctx)

  770. {

  771.     int i;

  772.     EBUR128Context *ebur128 = ctx->priv;

  773. 

  774.     av_log(ctx, AV_LOG_INFO, "Summary:\n\n"

  775.            "  Integrated loudness:\n"

  776.            "    I:         %5.1f LUFS\n"

  777.            "    Threshold: %5.1f LUFS\n\n"

  778.            "  Loudness range:\n"

  779.            "    LRA:       %5.1f LU\n"

  780.            "    Threshold: %5.1f LUFS\n"

  781.            "    LRA low:   %5.1f LUFS\n"

  782.            "    LRA high:  %5.1f LUFS\n",

  783.            ebur128->integrated_loudness, ebur128->i400.rel_threshold,

  784.            ebur128->loudness_range,      ebur128->i3000.rel_threshold,

  785.            ebur128->lra_low, ebur128->lra_high);

  786. 

  787.     av_freep(&ebur128->y_line_ref);

  788.     av_freep(&ebur128->ch_weighting);

  789.     av_freep(&ebur128->i400.histogram);

  790.     av_freep(&ebur128->i3000.histogram);

  791.     for (i = 0; i < ebur128->nb_channels; i++) {

  792.         av_freep(&ebur128->i400.cache[i]);

  793.         av_freep(&ebur128->i3000.cache[i]);

  794.     }

  795.     for (i = 0; i < ctx->nb_outputs; i++)

  796.         av_freep(&ctx->output_pads[i].name);

  797.     av_frame_free(&ebur128->outpicref);

  798.     av_opt_free(ebur128);

  799. }

  800. 

  801. static const AVFilterPad ebur128_inputs[] = {

  802.     {

  803.         .name             = "default",

  804.         .type             = AVMEDIA_TYPE_AUDIO,

  805.         .get_audio_buffer = ff_null_get_audio_buffer,

  806.         .filter_frame     = filter_frame,

  807.         .config_props     = config_audio_input,

  808.     },

  809.     { NULL }

  810. };

  811. 

  812. AVFilter avfilter_af_ebur128 = {

  813.     .name          = "ebur128",

  814.     .description   = NULL_IF_CONFIG_SMALL("EBU R128 scanner."),

  815.     .priv_size     = sizeof(EBUR128Context),

  816.     .init          = init,

  817.     .uninit        = uninit,

  818.     .query_formats = query_formats,

  819.     .inputs        = ebur128_inputs,

  820.     .outputs       = NULL,

  821.     .priv_class    = &ebur128_class,

  822. };