1 /*
2 * Copyright (c) 2017 Paul B Mahol
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 /**
22 * @file
23 * Filter for reading closed captioning data (EIA-608).
24 * See also https://en.wikipedia.org/wiki/EIA-608
25 */
26
27 #include <string.h>
28
29 #include "libavutil/internal.h"
30 #include "libavutil/opt.h"
31 #include "libavutil/pixdesc.h"
32 #include "libavutil/timestamp.h"
33
34 #include "avfilter.h"
35 #include "formats.h"
36 #include "internal.h"
37 #include "video.h"
38
39 #define LAG 25
40 #define CLOCK_BITSIZE_MIN 0.2f
41 #define CLOCK_BITSIZE_MAX 1.5f
42 #define SYNC_BITSIZE_MIN 12.f
43 #define SYNC_BITSIZE_MAX 15.f
44
45 typedef struct LineItem {
46 int input;
47 int output;
48
49 float unfiltered;
50 float filtered;
51 float average;
52 float deviation;
53 } LineItem;
54
55 typedef struct CodeItem {
56 uint8_t bit;
57 int size;
58 } CodeItem;
59
60 typedef struct ScanItem {
61 int nb_line;
62 int found;
63 int white;
64 int black;
65 uint64_t *histogram;
66 uint8_t byte[2];
67
68 CodeItem *code;
69 LineItem *line;
70 } ScanItem;
71
72 typedef struct ReadEIA608Context {
73 const AVClass *class;
74
75 int start, end;
76 float spw;
77 int chp;
78 int lp;
79
80 int depth;
81 int max;
82 int nb_allocated;
83 ScanItem *scan;
84
85 void (*read_line[2])(AVFrame *in, int nb_line,
86 LineItem *line, int lp, int w);
87 } ReadEIA608Context;
88
89 #define OFFSET(x) offsetof(ReadEIA608Context, x)
90 #define FLAGS AV_OPT_FLAG_VIDEO_PARAM|AV_OPT_FLAG_FILTERING_PARAM|AV_OPT_FLAG_RUNTIME_PARAM
91
92 static const AVOption readeia608_options[] = {
93 { "scan_min", "set from which line to scan for codes", OFFSET(start), AV_OPT_TYPE_INT, {.i64=0}, 0, INT_MAX, FLAGS },
94 { "scan_max", "set to which line to scan for codes", OFFSET(end), AV_OPT_TYPE_INT, {.i64=29}, 0, INT_MAX, FLAGS },
95 { "spw", "set ratio of width reserved for sync code detection", OFFSET(spw), AV_OPT_TYPE_FLOAT, {.dbl=.27}, 0.1, 0.7, FLAGS },
96 { "chp", "check and apply parity bit", OFFSET(chp), AV_OPT_TYPE_BOOL, {.i64= 0}, 0, 1, FLAGS },
97 { "lp", "lowpass line prior to processing", OFFSET(lp), AV_OPT_TYPE_BOOL, {.i64= 1}, 0, 1, FLAGS },
98 { NULL }
99 };
100
101 AVFILTER_DEFINE_CLASS(readeia608);
102
103 static const enum AVPixelFormat pixel_fmts[] = {
104 AV_PIX_FMT_GRAY8, AV_PIX_FMT_GRAY9,
105 AV_PIX_FMT_GRAY10, AV_PIX_FMT_GRAY12, AV_PIX_FMT_GRAY14,
106 AV_PIX_FMT_GRAY16,
107 AV_PIX_FMT_YUV410P, AV_PIX_FMT_YUV411P,
108 AV_PIX_FMT_YUV420P, AV_PIX_FMT_YUV422P,
109 AV_PIX_FMT_YUV440P, AV_PIX_FMT_YUV444P,
110 AV_PIX_FMT_YUVJ420P, AV_PIX_FMT_YUVJ422P,
111 AV_PIX_FMT_YUVJ440P, AV_PIX_FMT_YUVJ444P,
112 AV_PIX_FMT_YUVJ411P,
113 AV_PIX_FMT_YUV420P9, AV_PIX_FMT_YUV422P9, AV_PIX_FMT_YUV444P9,
114 AV_PIX_FMT_YUV420P10, AV_PIX_FMT_YUV422P10, AV_PIX_FMT_YUV444P10,
115 AV_PIX_FMT_YUV440P10,
116 AV_PIX_FMT_YUV444P12, AV_PIX_FMT_YUV422P12, AV_PIX_FMT_YUV420P12,
117 AV_PIX_FMT_YUV440P12,
118 AV_PIX_FMT_YUV444P14, AV_PIX_FMT_YUV422P14, AV_PIX_FMT_YUV420P14,
119 AV_PIX_FMT_YUV420P16, AV_PIX_FMT_YUV422P16, AV_PIX_FMT_YUV444P16,
120 AV_PIX_FMT_YUVA420P, AV_PIX_FMT_YUVA422P, AV_PIX_FMT_YUVA444P,
121 AV_PIX_FMT_YUVA444P9, AV_PIX_FMT_YUVA444P10, AV_PIX_FMT_YUVA444P12, AV_PIX_FMT_YUVA444P16,
122 AV_PIX_FMT_YUVA422P9, AV_PIX_FMT_YUVA422P10, AV_PIX_FMT_YUVA422P12, AV_PIX_FMT_YUVA422P16,
123 AV_PIX_FMT_YUVA420P9, AV_PIX_FMT_YUVA420P10, AV_PIX_FMT_YUVA420P16,
124 AV_PIX_FMT_NONE
125 };
126
config_filter(AVFilterContext * ctx,int start,int end)127 static int config_filter(AVFilterContext *ctx, int start, int end)
128 {
129 ReadEIA608Context *s = ctx->priv;
130 AVFilterLink *inlink = ctx->inputs[0];
131 int size = inlink->w + LAG;
132
133 if (end >= inlink->h) {
134 av_log(ctx, AV_LOG_WARNING, "Last line to scan too large, clipping.\n");
135 end = inlink->h - 1;
136 }
137
138 if (start > end) {
139 av_log(ctx, AV_LOG_ERROR, "Invalid range.\n");
140 return AVERROR(EINVAL);
141 }
142
143 if (s->nb_allocated < end - start + 1) {
144 const int diff = end - start + 1 - s->nb_allocated;
145
146 s->scan = av_realloc_f(s->scan, end - start + 1, sizeof(*s->scan));
147 if (!s->scan)
148 return AVERROR(ENOMEM);
149 memset(&s->scan[s->nb_allocated], 0, diff * sizeof(*s->scan));
150 s->nb_allocated = end - start + 1;
151 }
152
153 for (int i = 0; i < s->nb_allocated; i++) {
154 ScanItem *scan = &s->scan[i];
155
156 if (!scan->histogram)
157 scan->histogram = av_calloc(s->max + 1, sizeof(*scan->histogram));
158 if (!scan->line)
159 scan->line = av_calloc(size, sizeof(*scan->line));
160 if (!scan->code)
161 scan->code = av_calloc(size, sizeof(*scan->code));
162 if (!scan->line || !scan->code || !scan->histogram)
163 return AVERROR(ENOMEM);
164 }
165
166 s->start = start;
167 s->end = end;
168
169 return 0;
170 }
171
build_histogram(ReadEIA608Context * s,ScanItem * scan,const LineItem * line,int len)172 static void build_histogram(ReadEIA608Context *s, ScanItem *scan, const LineItem *line, int len)
173 {
174 memset(scan->histogram, 0, (s->max + 1) * sizeof(*scan->histogram));
175
176 for (int i = LAG; i < len + LAG; i++)
177 scan->histogram[line[i].input]++;
178 }
179
find_black_and_white(ReadEIA608Context * s,ScanItem * scan)180 static void find_black_and_white(ReadEIA608Context *s, ScanItem *scan)
181 {
182 const int max = s->max;
183 int start = 0, end = 0, middle;
184 int black = 0, white = 0;
185 int cnt;
186
187 for (int i = 0; i <= max; i++) {
188 if (scan->histogram[i]) {
189 start = i;
190 break;
191 }
192 }
193
194 for (int i = max; i >= 0; i--) {
195 if (scan->histogram[i]) {
196 end = i;
197 break;
198 }
199 }
200
201 middle = start + (end - start) / 2;
202
203 cnt = 0;
204 for (int i = start; i <= middle; i++) {
205 if (scan->histogram[i] > cnt) {
206 cnt = scan->histogram[i];
207 black = i;
208 }
209 }
210
211 cnt = 0;
212 for (int i = end; i >= middle; i--) {
213 if (scan->histogram[i] > cnt) {
214 cnt = scan->histogram[i];
215 white = i;
216 }
217 }
218
219 scan->black = black;
220 scan->white = white;
221 }
222
meanf(const LineItem * line,int len)223 static float meanf(const LineItem *line, int len)
224 {
225 float sum = 0.0, mean = 0.0;
226
227 for (int i = 0; i < len; i++)
228 sum += line[i].filtered;
229
230 mean = sum / len;
231
232 return mean;
233 }
234
stddevf(const LineItem * line,int len)235 static float stddevf(const LineItem *line, int len)
236 {
237 float m = meanf(line, len);
238 float standard_deviation = 0.f;
239
240 for (int i = 0; i < len; i++)
241 standard_deviation += (line[i].filtered - m) * (line[i].filtered - m);
242
243 return sqrtf(standard_deviation / (len - 1));
244 }
245
thresholding(ReadEIA608Context * s,ScanItem * scan,LineItem * line,int lag,float threshold,float influence,int len)246 static void thresholding(ReadEIA608Context *s, ScanItem *scan, LineItem *line,
247 int lag, float threshold, float influence, int len)
248 {
249 for (int i = lag; i < len + lag; i++) {
250 line[i].unfiltered = line[i].input / 255.f;
251 line[i].filtered = line[i].unfiltered;
252 }
253
254 for (int i = 0; i < lag; i++) {
255 line[i].unfiltered = meanf(line, len * s->spw);
256 line[i].filtered = line[i].unfiltered;
257 }
258
259 line[lag - 1].average = meanf(line, lag);
260 line[lag - 1].deviation = stddevf(line, lag);
261
262 for (int i = lag; i < len + lag; i++) {
263 if (fabsf(line[i].unfiltered - line[i-1].average) > threshold * line[i-1].deviation) {
264 if (line[i].unfiltered > line[i-1].average) {
265 line[i].output = 255;
266 } else {
267 line[i].output = 0;
268 }
269
270 line[i].filtered = influence * line[i].unfiltered + (1.f - influence) * line[i-1].filtered;
271 } else {
272 int distance_from_black, distance_from_white;
273
274 distance_from_black = FFABS(line[i].input - scan->black);
275 distance_from_white = FFABS(line[i].input - scan->white);
276
277 line[i].output = distance_from_black <= distance_from_white ? 0 : 255;
278 }
279
280 line[i].average = meanf(line + i - lag, lag);
281 line[i].deviation = stddevf(line + i - lag, lag);
282 }
283 }
284
periods(const LineItem * line,CodeItem * code,int len)285 static int periods(const LineItem *line, CodeItem *code, int len)
286 {
287 int hold = line[LAG].output, cnt = 0;
288 int last = LAG;
289
290 memset(code, 0, len * sizeof(*code));
291
292 for (int i = LAG + 1; i < len + LAG; i++) {
293 if (line[i].output != hold) {
294 code[cnt].size = i - last;
295 code[cnt].bit = hold;
296 hold = line[i].output;
297 last = i;
298 cnt++;
299 }
300 }
301
302 code[cnt].size = LAG + len - last;
303 code[cnt].bit = hold;
304
305 return cnt + 1;
306 }
307
dump_code(AVFilterContext * ctx,ScanItem * scan,int len,int item)308 static void dump_code(AVFilterContext *ctx, ScanItem *scan, int len, int item)
309 {
310 av_log(ctx, AV_LOG_DEBUG, "%d:", item);
311 for (int i = 0; i < len; i++) {
312 av_log(ctx, AV_LOG_DEBUG, " %03d", scan->code[i].size);
313 }
314 av_log(ctx, AV_LOG_DEBUG, "\n");
315 }
316
317 #define READ_LINE(type, name) \
318 static void read_##name(AVFrame *in, int nb_line, LineItem *line, int lp, int w) \
319 { \
320 const type *src = (const type *)(&in->data[0][nb_line * in->linesize[0]]);\
321 \
322 if (lp) { \
323 for (int i = 0; i < w; i++) { \
324 int a = FFMAX(i - 3, 0); \
325 int b = FFMAX(i - 2, 0); \
326 int c = FFMAX(i - 1, 0); \
327 int d = FFMIN(i + 3, w-1); \
328 int e = FFMIN(i + 2, w-1); \
329 int f = FFMIN(i + 1, w-1); \
330 \
331 line[LAG + i].input = (src[a] + src[b] + src[c] + src[i] + \
332 src[d] + src[e] + src[f] + 6) / 7; \
333 } \
334 } else { \
335 for (int i = 0; i < w; i++) { \
336 line[LAG + i].input = src[i]; \
337 } \
338 } \
339 }
340
READ_LINE(uint8_t,byte)341 READ_LINE(uint8_t, byte)
342 READ_LINE(uint16_t, word)
343
344 static int config_input(AVFilterLink *inlink)
345 {
346 AVFilterContext *ctx = inlink->dst;
347 ReadEIA608Context *s = ctx->priv;
348 const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(inlink->format);
349
350 if (!desc)
351 return AVERROR_BUG;
352 s->depth = desc->comp[0].depth;
353 s->max = (1 << desc->comp[0].depth) - 1;
354 s->read_line[0] = read_byte;
355 s->read_line[1] = read_word;
356
357 return config_filter(ctx, s->start, s->end);
358 }
359
extract_line(AVFilterContext * ctx,AVFrame * in,ScanItem * scan,int w,int nb_line)360 static void extract_line(AVFilterContext *ctx, AVFrame *in, ScanItem *scan, int w, int nb_line)
361 {
362 ReadEIA608Context *s = ctx->priv;
363 LineItem *line = scan->line;
364 int i, j, ch, len;
365 uint8_t codes[19] = { 0 };
366 float bit_size = 0.f;
367 int parity;
368
369 memset(line, 0, (w + LAG) * sizeof(*line));
370 scan->byte[0] = scan->byte[1] = 0;
371 scan->found = 0;
372
373 s->read_line[s->depth > 8](in, nb_line, line, s->lp, w);
374
375 build_histogram(s, scan, line, w);
376 find_black_and_white(s, scan);
377 if (scan->white - scan->black < 5)
378 return;
379
380 thresholding(s, scan, line, LAG, 1, 0, w);
381 len = periods(line, scan->code, w);
382 dump_code(ctx, scan, len, nb_line);
383 if (len < 15 ||
384 scan->code[14].bit != 0 ||
385 w / (float)scan->code[14].size < SYNC_BITSIZE_MIN ||
386 w / (float)scan->code[14].size > SYNC_BITSIZE_MAX) {
387 return;
388 }
389
390 for (i = 14; i < len; i++) {
391 bit_size += scan->code[i].size;
392 }
393
394 bit_size /= 19.f;
395 for (i = 1; i < 14; i++) {
396 if (scan->code[i].size / bit_size > CLOCK_BITSIZE_MAX ||
397 scan->code[i].size / bit_size < CLOCK_BITSIZE_MIN) {
398 return;
399 }
400 }
401
402 if (scan->code[15].size / bit_size < 0.45f) {
403 return;
404 }
405
406 for (j = 0, i = 14; i < len; i++) {
407 int run, bit;
408
409 run = lrintf(scan->code[i].size / bit_size);
410 bit = scan->code[i].bit;
411
412 for (int k = 0; j < 19 && k < run; k++) {
413 codes[j++] = bit;
414 }
415
416 if (j >= 19)
417 break;
418 }
419
420 for (ch = 0; ch < 2; ch++) {
421 for (parity = 0, i = 0; i < 8; i++) {
422 int b = codes[3 + ch * 8 + i];
423
424 if (b == 255) {
425 parity++;
426 b = 1;
427 } else {
428 b = 0;
429 }
430 scan->byte[ch] |= b << i;
431 }
432
433 if (s->chp) {
434 if (!(parity & 1)) {
435 scan->byte[ch] = 0x7F;
436 }
437 }
438 }
439
440 scan->nb_line = nb_line;
441 scan->found = 1;
442 }
443
extract_lines(AVFilterContext * ctx,void * arg,int job,int nb_jobs)444 static int extract_lines(AVFilterContext *ctx, void *arg,
445 int job, int nb_jobs)
446 {
447 ReadEIA608Context *s = ctx->priv;
448 AVFilterLink *inlink = ctx->inputs[0];
449 const int h = s->end - s->start + 1;
450 const int start = (h * job) / nb_jobs;
451 const int end = (h * (job+1)) / nb_jobs;
452 AVFrame *in = arg;
453
454 for (int i = start; i < end; i++) {
455 ScanItem *scan = &s->scan[i];
456
457 extract_line(ctx, in, scan, inlink->w, s->start + i);
458 }
459
460 return 0;
461 }
462
filter_frame(AVFilterLink * inlink,AVFrame * in)463 static int filter_frame(AVFilterLink *inlink, AVFrame *in)
464 {
465 AVFilterContext *ctx = inlink->dst;
466 AVFilterLink *outlink = ctx->outputs[0];
467 ReadEIA608Context *s = ctx->priv;
468 int nb_found;
469
470 ff_filter_execute(ctx, extract_lines, in, NULL,
471 FFMIN(FFMAX(s->end - s->start + 1, 1), ff_filter_get_nb_threads(ctx)));
472
473 nb_found = 0;
474 for (int i = 0; i < s->end - s->start + 1; i++) {
475 ScanItem *scan = &s->scan[i];
476 uint8_t key[128], value[128];
477
478 if (!scan->found)
479 continue;
480
481 //snprintf(key, sizeof(key), "lavfi.readeia608.%d.bits", nb_found);
482 //snprintf(value, sizeof(value), "0b%d%d%d%d%d%d%d%d 0b%d%d%d%d%d%d%d%d", codes[3]==255,codes[4]==255,codes[5]==255,codes[6]==255,codes[7]==255,codes[8]==255,codes[9]==255,codes[10]==255,codes[11]==255,codes[12]==255,codes[13]==255,codes[14]==255,codes[15]==255,codes[16]==255,codes[17]==255,codes[18]==255);
483 //av_dict_set(&in->metadata, key, value, 0);
484
485 snprintf(key, sizeof(key), "lavfi.readeia608.%d.cc", nb_found);
486 snprintf(value, sizeof(value), "0x%02X%02X", scan->byte[0], scan->byte[1]);
487 av_dict_set(&in->metadata, key, value, 0);
488
489 snprintf(key, sizeof(key), "lavfi.readeia608.%d.line", nb_found);
490 av_dict_set_int(&in->metadata, key, scan->nb_line, 0);
491
492 nb_found++;
493 }
494
495 return ff_filter_frame(outlink, in);
496 }
497
uninit(AVFilterContext * ctx)498 static av_cold void uninit(AVFilterContext *ctx)
499 {
500 ReadEIA608Context *s = ctx->priv;
501
502 for (int i = 0; i < s->nb_allocated; i++) {
503 ScanItem *scan = &s->scan[i];
504
505 av_freep(&scan->histogram);
506 av_freep(&scan->code);
507 av_freep(&scan->line);
508 }
509
510 s->nb_allocated = 0;
511 av_freep(&s->scan);
512 }
513
process_command(AVFilterContext * ctx,const char * cmd,const char * args,char * res,int res_len,int flags)514 static int process_command(AVFilterContext *ctx, const char *cmd, const char *args,
515 char *res, int res_len, int flags)
516 {
517 ReadEIA608Context *s = ctx->priv;
518 int ret, start = s->start, end = s->end;
519
520 ret = ff_filter_process_command(ctx, cmd, args, res, res_len, flags);
521 if (ret < 0)
522 return ret;
523
524 ret = config_filter(ctx, s->start, s->end);
525 if (ret < 0) {
526 s->start = start;
527 s->end = end;
528 }
529
530 return 0;
531 }
532
533 static const AVFilterPad readeia608_inputs[] = {
534 {
535 .name = "default",
536 .type = AVMEDIA_TYPE_VIDEO,
537 .filter_frame = filter_frame,
538 .config_props = config_input,
539 },
540 };
541
542 static const AVFilterPad readeia608_outputs[] = {
543 {
544 .name = "default",
545 .type = AVMEDIA_TYPE_VIDEO,
546 },
547 };
548
549 const AVFilter ff_vf_readeia608 = {
550 .name = "readeia608",
551 .description = NULL_IF_CONFIG_SMALL("Read EIA-608 Closed Caption codes from input video and write them to frame metadata."),
552 .priv_size = sizeof(ReadEIA608Context),
553 .priv_class = &readeia608_class,
554 FILTER_INPUTS(readeia608_inputs),
555 FILTER_OUTPUTS(readeia608_outputs),
556 FILTER_PIXFMTS_ARRAY(pixel_fmts),
557 .uninit = uninit,
558 .flags = AVFILTER_FLAG_SUPPORT_TIMELINE_GENERIC |
559 AVFILTER_FLAG_SLICE_THREADS |
560 AVFILTER_FLAG_METADATA_ONLY,
561 .process_command = process_command,
562 };
563