2 * Closed Caption Decoding
3 * Copyright (c) 2015 Anshul Maheshwari
5 * This file is part of FFmpeg.
7 * FFmpeg is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU Lesser General Public
9 * License as published by the Free Software Foundation; either
10 * version 2.1 of the License, or (at your option) any later version.
12 * FFmpeg is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * Lesser General Public License for more details.
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with FFmpeg; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
24 #include "libavutil/opt.h"
26 #define SCREEN_ROWS 15
27 #define SCREEN_COLUMNS 32
29 #define SET_FLAG(var, val) ( (var) |= ( 1 << (val)) )
30 #define UNSET_FLAG(var, val) ( (var) &= ~( 1 << (val)) )
31 #define CHECK_FLAG(var, val) ( (var) & ( 1 << (val)) )
33 static const AVRational ms_tb = {1, 1000};
37 * 1) handle font and color completely
63 CCFONT_UNDERLINED_ITALICS,
68 CCSET_SPECIAL_AMERICAN,
69 CCSET_EXTENDED_SPANISH_FRENCH_MISC,
70 CCSET_EXTENDED_PORTUGUESE_GERMAN_DANISH,
73 static const char *charset_overrides[4][128] =
75 [CCSET_BASIC_AMERICAN] = {
88 [CCSET_SPECIAL_AMERICAN] = {
106 [CCSET_EXTENDED_SPANISH_FRENCH_MISC] = {
140 [CCSET_EXTENDED_PORTUGUESE_GERMAN_DANISH] = {
176 static const unsigned char pac2_attribs[32][3] = // Color, font, ident
178 { CCCOL_WHITE, CCFONT_REGULAR, 0 }, // 0x40 || 0x60
179 { CCCOL_WHITE, CCFONT_UNDERLINED, 0 }, // 0x41 || 0x61
180 { CCCOL_GREEN, CCFONT_REGULAR, 0 }, // 0x42 || 0x62
181 { CCCOL_GREEN, CCFONT_UNDERLINED, 0 }, // 0x43 || 0x63
182 { CCCOL_BLUE, CCFONT_REGULAR, 0 }, // 0x44 || 0x64
183 { CCCOL_BLUE, CCFONT_UNDERLINED, 0 }, // 0x45 || 0x65
184 { CCCOL_CYAN, CCFONT_REGULAR, 0 }, // 0x46 || 0x66
185 { CCCOL_CYAN, CCFONT_UNDERLINED, 0 }, // 0x47 || 0x67
186 { CCCOL_RED, CCFONT_REGULAR, 0 }, // 0x48 || 0x68
187 { CCCOL_RED, CCFONT_UNDERLINED, 0 }, // 0x49 || 0x69
188 { CCCOL_YELLOW, CCFONT_REGULAR, 0 }, // 0x4a || 0x6a
189 { CCCOL_YELLOW, CCFONT_UNDERLINED, 0 }, // 0x4b || 0x6b
190 { CCCOL_MAGENTA, CCFONT_REGULAR, 0 }, // 0x4c || 0x6c
191 { CCCOL_MAGENTA, CCFONT_UNDERLINED, 0 }, // 0x4d || 0x6d
192 { CCCOL_WHITE, CCFONT_ITALICS, 0 }, // 0x4e || 0x6e
193 { CCCOL_WHITE, CCFONT_UNDERLINED_ITALICS, 0 }, // 0x4f || 0x6f
194 { CCCOL_WHITE, CCFONT_REGULAR, 0 }, // 0x50 || 0x70
195 { CCCOL_WHITE, CCFONT_UNDERLINED, 0 }, // 0x51 || 0x71
196 { CCCOL_WHITE, CCFONT_REGULAR, 4 }, // 0x52 || 0x72
197 { CCCOL_WHITE, CCFONT_UNDERLINED, 4 }, // 0x53 || 0x73
198 { CCCOL_WHITE, CCFONT_REGULAR, 8 }, // 0x54 || 0x74
199 { CCCOL_WHITE, CCFONT_UNDERLINED, 8 }, // 0x55 || 0x75
200 { CCCOL_WHITE, CCFONT_REGULAR, 12 }, // 0x56 || 0x76
201 { CCCOL_WHITE, CCFONT_UNDERLINED, 12 }, // 0x57 || 0x77
202 { CCCOL_WHITE, CCFONT_REGULAR, 16 }, // 0x58 || 0x78
203 { CCCOL_WHITE, CCFONT_UNDERLINED, 16 }, // 0x59 || 0x79
204 { CCCOL_WHITE, CCFONT_REGULAR, 20 }, // 0x5a || 0x7a
205 { CCCOL_WHITE, CCFONT_UNDERLINED, 20 }, // 0x5b || 0x7b
206 { CCCOL_WHITE, CCFONT_REGULAR, 24 }, // 0x5c || 0x7c
207 { CCCOL_WHITE, CCFONT_UNDERLINED, 24 }, // 0x5d || 0x7d
208 { CCCOL_WHITE, CCFONT_REGULAR, 28 }, // 0x5e || 0x7e
209 { CCCOL_WHITE, CCFONT_UNDERLINED, 28 } // 0x5f || 0x7f
210 /* total 32 entries */
214 /* +1 is used to compensate null character of string */
215 uint8_t characters[SCREEN_ROWS][SCREEN_COLUMNS+1];
216 uint8_t charsets[SCREEN_ROWS][SCREEN_COLUMNS+1];
217 uint8_t colors[SCREEN_ROWS][SCREEN_COLUMNS+1];
218 uint8_t fonts[SCREEN_ROWS][SCREEN_COLUMNS+1];
220 * Bitmask of used rows; if a bit is not set, the
221 * corresponding row is not used.
222 * for setting row 1 use row | (1 << 0)
223 * for setting row 15 use row | (1 << 14)
228 typedef struct CCaptionSubContext {
231 struct Screen screen[2];
234 uint8_t cursor_column;
235 uint8_t cursor_color;
237 uint8_t cursor_charset;
243 /* visible screen time */
247 int64_t last_real_time;
249 /* buffer to store pkt data */
252 } CCaptionSubContext;
255 static av_cold int init_decoder(AVCodecContext *avctx)
258 CCaptionSubContext *ctx = avctx->priv_data;
260 av_bprint_init(&ctx->buffer, 0, AV_BPRINT_SIZE_UNLIMITED);
261 /* taking by default roll up to 2 */
262 ctx->mode = CCMODE_ROLLUP;
264 ctx->cursor_row = 10;
265 ret = ff_ass_subtitle_header(avctx, "Monospace",
266 ASS_DEFAULT_FONT_SIZE,
268 ASS_DEFAULT_BACK_COLOR,
271 ASS_DEFAULT_UNDERLINE,
273 ASS_DEFAULT_ALIGNMENT);
277 /* allocate pkt buffer */
278 ctx->pktbuf = av_buffer_alloc(128);
280 ret = AVERROR(ENOMEM);
285 static av_cold int close_decoder(AVCodecContext *avctx)
287 CCaptionSubContext *ctx = avctx->priv_data;
288 av_bprint_finalize(&ctx->buffer, NULL);
289 av_buffer_unref(&ctx->pktbuf);
293 static void flush_decoder(AVCodecContext *avctx)
295 CCaptionSubContext *ctx = avctx->priv_data;
296 ctx->screen[0].row_used = 0;
297 ctx->screen[1].row_used = 0;
298 ctx->prev_cmd[0] = 0;
299 ctx->prev_cmd[1] = 0;
300 ctx->mode = CCMODE_ROLLUP;
302 ctx->cursor_row = 10;
303 ctx->cursor_column = 0;
304 ctx->cursor_font = 0;
305 ctx->cursor_color = 0;
306 ctx->cursor_charset = 0;
307 ctx->active_screen = 0;
308 ctx->last_real_time = 0;
309 ctx->screen_touched = 0;
310 ctx->buffer_changed = 0;
311 if (!(avctx->flags2 & AV_CODEC_FLAG2_RO_FLUSH_NOOP))
313 av_bprint_clear(&ctx->buffer);
317 * @param ctx closed caption context just to print log
319 static void write_char(CCaptionSubContext *ctx, struct Screen *screen, char ch)
321 uint8_t col = ctx->cursor_column;
322 char *row = screen->characters[ctx->cursor_row];
323 char *font = screen->fonts[ctx->cursor_row];
324 char *charset = screen->charsets[ctx->cursor_row];
326 if (col < SCREEN_COLUMNS) {
328 font[col] = ctx->cursor_font;
329 charset[col] = ctx->cursor_charset;
330 ctx->cursor_charset = CCSET_BASIC_AMERICAN;
331 if (ch) ctx->cursor_column++;
334 /* We have extra space at end only for null character */
335 else if (col == SCREEN_COLUMNS && ch == 0) {
340 av_log(ctx, AV_LOG_WARNING, "Data Ignored since exceeding screen width\n");
346 * This function after validating parity bit, also remove it from data pair.
347 * The first byte doesn't pass parity, we replace it with a solid blank
348 * and process the pair.
349 * If the second byte doesn't pass parity, it returns INVALIDDATA
350 * user can ignore the whole pair and pass the other pair.
352 static int validate_cc_data_pair(uint8_t *cc_data_pair)
354 uint8_t cc_valid = (*cc_data_pair & 4) >>2;
355 uint8_t cc_type = *cc_data_pair & 3;
358 return AVERROR_INVALIDDATA;
360 // if EIA-608 data then verify parity.
361 if (cc_type==0 || cc_type==1) {
362 if (!av_parity(cc_data_pair[2])) {
363 return AVERROR_INVALIDDATA;
365 if (!av_parity(cc_data_pair[1])) {
366 cc_data_pair[1]=0x7F;
371 if ((cc_data_pair[0] == 0xFA || cc_data_pair[0] == 0xFC || cc_data_pair[0] == 0xFD)
372 && (cc_data_pair[1] & 0x7F) == 0 && (cc_data_pair[2] & 0x7F) == 0)
373 return AVERROR_PATCHWELCOME;
376 if (cc_type == 3 || cc_type == 2)
377 return AVERROR_PATCHWELCOME;
379 /* remove parity bit */
380 cc_data_pair[1] &= 0x7F;
381 cc_data_pair[2] &= 0x7F;
386 static struct Screen *get_writing_screen(CCaptionSubContext *ctx)
390 // use Inactive screen
391 return ctx->screen + !ctx->active_screen;
396 return ctx->screen + ctx->active_screen;
398 /* It was never an option */
402 static void roll_up(CCaptionSubContext *ctx)
404 struct Screen *screen;
407 if (ctx->mode == CCMODE_TEXT)
410 screen = get_writing_screen(ctx);
412 /* +1 signify cursor_row starts from 0
413 * Can't keep lines less then row cursor pos
415 keep_lines = FFMIN(ctx->cursor_row + 1, ctx->rollup);
417 for (i = 0; i < SCREEN_ROWS; i++) {
418 if (i > ctx->cursor_row - keep_lines && i <= ctx->cursor_row)
420 UNSET_FLAG(screen->row_used, i);
423 for (i = 0; i < keep_lines && screen->row_used; i++) {
424 const int i_row = ctx->cursor_row - keep_lines + i + 1;
426 memcpy(screen->characters[i_row], screen->characters[i_row+1], SCREEN_COLUMNS);
427 memcpy(screen->colors[i_row], screen->colors[i_row+1], SCREEN_COLUMNS);
428 memcpy(screen->fonts[i_row], screen->fonts[i_row+1], SCREEN_COLUMNS);
429 memcpy(screen->charsets[i_row], screen->charsets[i_row+1], SCREEN_COLUMNS);
430 if (CHECK_FLAG(screen->row_used, i_row + 1))
431 SET_FLAG(screen->row_used, i_row);
434 UNSET_FLAG(screen->row_used, ctx->cursor_row);
437 static int capture_screen(CCaptionSubContext *ctx)
440 struct Screen *screen = ctx->screen + ctx->active_screen;
441 enum cc_font prev_font = CCFONT_REGULAR;
442 av_bprint_clear(&ctx->buffer);
444 for (i = 0; screen->row_used && i < SCREEN_ROWS; i++)
446 if (CHECK_FLAG(screen->row_used, i)) {
447 const char *row = screen->characters[i];
448 const char *charset = screen->charsets[i];
450 while (row[j] == ' ' && charset[j] == CCSET_BASIC_AMERICAN)
457 for (i = 0; screen->row_used && i < SCREEN_ROWS; i++)
459 if (CHECK_FLAG(screen->row_used, i)) {
460 const char *row = screen->characters[i];
461 const char *font = screen->fonts[i];
462 const char *charset = screen->charsets[i];
463 const char *override;
464 int x, y, seen_char = 0;
467 /* skip leading space */
468 while (row[j] == ' ' && charset[j] == CCSET_BASIC_AMERICAN && j < tab)
471 x = ASS_DEFAULT_PLAYRESX * (0.1 + 0.0250 * j);
472 y = ASS_DEFAULT_PLAYRESY * (0.1 + 0.0533 * i);
473 av_bprintf(&ctx->buffer, "{\\an7}{\\pos(%d,%d)}", x, y);
475 for (; j < SCREEN_COLUMNS; j++) {
476 const char *e_tag = "", *s_tag = "";
481 if (prev_font != font[j]) {
486 case CCFONT_UNDERLINED:
489 case CCFONT_UNDERLINED_ITALICS:
490 e_tag = "{\\u0}{\\i0}";
497 case CCFONT_UNDERLINED:
500 case CCFONT_UNDERLINED_ITALICS:
501 s_tag = "{\\u1}{\\i1}";
506 override = charset_overrides[(int)charset[j]][(int)row[j]];
508 av_bprintf(&ctx->buffer, "%s%s%s", e_tag, s_tag, override);
510 } else if (row[j] == ' ' && !seen_char) {
511 av_bprintf(&ctx->buffer, "%s%s\\h", e_tag, s_tag);
513 av_bprintf(&ctx->buffer, "%s%s%c", e_tag, s_tag, row[j]);
518 av_bprintf(&ctx->buffer, "\\N");
521 if (!av_bprint_is_complete(&ctx->buffer))
522 return AVERROR(ENOMEM);
523 if (screen->row_used && ctx->buffer.len >= 2) {
524 ctx->buffer.len -= 2;
525 ctx->buffer.str[ctx->buffer.len] = 0;
527 ctx->buffer_changed = 1;
531 static int reap_screen(CCaptionSubContext *ctx, int64_t pts)
533 ctx->start_time = ctx->startv_time;
534 ctx->startv_time = pts;
536 return capture_screen(ctx);
539 static void handle_textattr(CCaptionSubContext *ctx, uint8_t hi, uint8_t lo)
542 struct Screen *screen = get_writing_screen(ctx);
547 ctx->cursor_color = pac2_attribs[i][0];
548 ctx->cursor_font = pac2_attribs[i][1];
550 SET_FLAG(screen->row_used, ctx->cursor_row);
551 write_char(ctx, screen, ' ');
554 static void handle_pac(CCaptionSubContext *ctx, uint8_t hi, uint8_t lo)
556 static const int8_t row_map[] = {
557 11, -1, 1, 2, 3, 4, 12, 13, 14, 15, 5, 6, 7, 8, 9, 10
559 const int index = ( (hi<<1) & 0x0e) | ( (lo>>5) & 0x01 );
560 struct Screen *screen = get_writing_screen(ctx);
563 if (row_map[index] <= 0) {
564 av_log(ctx, AV_LOG_DEBUG, "Invalid pac index encountered\n");
570 ctx->cursor_row = row_map[index] - 1;
571 ctx->cursor_color = pac2_attribs[lo][0];
572 ctx->cursor_font = pac2_attribs[lo][1];
573 ctx->cursor_charset = CCSET_BASIC_AMERICAN;
574 ctx->cursor_column = 0;
575 indent = pac2_attribs[lo][2];
576 for (i = 0; i < indent; i++) {
577 write_char(ctx, screen, ' ');
582 * @param pts it is required to set end time
584 static void handle_edm(CCaptionSubContext *ctx, int64_t pts)
586 struct Screen *screen = ctx->screen + ctx->active_screen;
588 // In buffered mode, keep writing to screen until it is wiped.
589 // Before wiping the display, capture contents to emit subtitle.
591 reap_screen(ctx, pts);
593 screen->row_used = 0;
595 // In realtime mode, emit an empty caption so the last one doesn't
596 // stay on the screen.
598 reap_screen(ctx, pts);
601 static void handle_eoc(CCaptionSubContext *ctx, int64_t pts)
603 // In buffered mode, we wait til the *next* EOC and
604 // reap what was already on the screen since the last EOC.
608 ctx->active_screen = !ctx->active_screen;
609 ctx->cursor_column = 0;
611 // In realtime mode, we display the buffered contents (after
612 // flipping the buffer to active above) as soon as EOC arrives.
614 reap_screen(ctx, pts);
617 static void handle_delete_end_of_row(CCaptionSubContext *ctx, char hi, char lo)
619 struct Screen *screen = get_writing_screen(ctx);
620 write_char(ctx, screen, 0);
623 static void handle_char(CCaptionSubContext *ctx, char hi, char lo, int64_t pts)
625 struct Screen *screen = get_writing_screen(ctx);
627 SET_FLAG(screen->row_used, ctx->cursor_row);
631 ctx->cursor_charset = CCSET_SPECIAL_AMERICAN;
634 if (ctx->cursor_column > 0)
635 ctx->cursor_column -= 1;
636 ctx->cursor_charset = CCSET_EXTENDED_SPANISH_FRENCH_MISC;
639 if (ctx->cursor_column > 0)
640 ctx->cursor_column -= 1;
641 ctx->cursor_charset = CCSET_EXTENDED_PORTUGUESE_GERMAN_DANISH;
644 ctx->cursor_charset = CCSET_BASIC_AMERICAN;
645 write_char(ctx, screen, hi);
650 write_char(ctx, screen, lo);
652 write_char(ctx, screen, 0);
654 if (ctx->mode != CCMODE_POPON)
655 ctx->screen_touched = 1;
658 ff_dlog(ctx, "(%c,%c)\n", hi, lo);
660 ff_dlog(ctx, "(%c)\n", hi);
663 static void process_cc608(CCaptionSubContext *ctx, int64_t pts, uint8_t hi, uint8_t lo)
665 if (hi == ctx->prev_cmd[0] && lo == ctx->prev_cmd[1]) {
666 /* ignore redundant command */
670 /* set prev command */
671 ctx->prev_cmd[0] = hi;
672 ctx->prev_cmd[1] = lo;
674 if ( (hi == 0x10 && (lo >= 0x40 && lo <= 0x5f)) ||
675 ( (hi >= 0x11 && hi <= 0x17) && (lo >= 0x40 && lo <= 0x7f) ) ) {
676 handle_pac(ctx, hi, lo);
677 } else if ( ( hi == 0x11 && lo >= 0x20 && lo <= 0x2f ) ||
678 ( hi == 0x17 && lo >= 0x2e && lo <= 0x2f) ) {
679 handle_textattr(ctx, hi, lo);
680 } else if (hi == 0x14 || hi == 0x15 || hi == 0x1c) {
683 /* resume caption loading */
684 ctx->mode = CCMODE_POPON;
687 handle_delete_end_of_row(ctx, hi, lo);
692 ctx->rollup = lo - 0x23;
693 ctx->mode = CCMODE_ROLLUP;
696 /* resume direct captioning */
697 ctx->mode = CCMODE_PAINTON;
700 /* resume text display */
701 ctx->mode = CCMODE_TEXT;
704 /* erase display memory */
705 handle_edm(ctx, pts);
708 /* carriage return */
709 ff_dlog(ctx, "carriage return\n");
711 reap_screen(ctx, pts);
713 ctx->cursor_column = 0;
716 /* erase buffered (non displayed) memory */
717 // Only in realtime mode. In buffered mode, we re-use the inactive screen
718 // for our own buffering.
719 if (ctx->real_time) {
720 struct Screen *screen = ctx->screen + !ctx->active_screen;
721 screen->row_used = 0;
726 ff_dlog(ctx, "handle_eoc\n");
727 handle_eoc(ctx, pts);
730 ff_dlog(ctx, "Unknown command 0x%hhx 0x%hhx\n", hi, lo);
733 } else if (hi >= 0x11 && hi <= 0x13) {
734 /* Special characters */
735 handle_char(ctx, hi, lo, pts);
736 } else if (hi >= 0x20) {
737 /* Standard characters (always in pairs) */
738 handle_char(ctx, hi, lo, pts);
739 ctx->prev_cmd[0] = ctx->prev_cmd[1] = 0;
740 } else if (hi == 0x17 && lo >= 0x21 && lo <= 0x23) {
742 /* Tab offsets (spacing) */
743 for (i = 0; i < lo - 0x20; i++) {
744 handle_char(ctx, ' ', 0, pts);
747 /* Ignoring all other non data code */
748 ff_dlog(ctx, "Unknown command 0x%hhx 0x%hhx\n", hi, lo);
752 static int decode(AVCodecContext *avctx, void *data, int *got_sub, AVPacket *avpkt)
754 CCaptionSubContext *ctx = avctx->priv_data;
755 AVSubtitle *sub = data;
756 const int64_t start_time = sub->pts;
757 uint8_t *bptr = NULL;
758 int len = avpkt->size;
762 if (ctx->pktbuf->size < len) {
763 ret = av_buffer_realloc(&ctx->pktbuf, len);
765 av_log(ctx, AV_LOG_WARNING, "Insufficient Memory of %d truncated to %d\n", len, ctx->pktbuf->size);
766 len = ctx->pktbuf->size;
770 memcpy(ctx->pktbuf->data, avpkt->data, len);
771 bptr = ctx->pktbuf->data;
773 for (i = 0; i < len; i += 3) {
774 uint8_t cc_type = *(bptr + i) & 3;
775 if (validate_cc_data_pair(bptr + i))
777 /* ignoring data field 1 */
781 process_cc608(ctx, start_time, *(bptr + i + 1) & 0x7f, *(bptr + i + 2) & 0x7f);
783 if (!ctx->buffer_changed)
785 ctx->buffer_changed = 0;
787 if (*ctx->buffer.str || ctx->real_time)
789 ff_dlog(ctx, "cdp writing data (%s)\n",ctx->buffer.str);
790 ret = ff_ass_add_rect(sub, ctx->buffer.str, ctx->readorder++, 0, NULL, NULL);
793 sub->pts = ctx->start_time;
795 sub->end_display_time = av_rescale_q(ctx->end_time - ctx->start_time,
796 AV_TIME_BASE_Q, ms_tb);
798 sub->end_display_time = -1;
799 ctx->buffer_changed = 0;
800 ctx->last_real_time = sub->pts;
801 ctx->screen_touched = 0;
805 if (ctx->real_time && ctx->screen_touched &&
806 sub->pts > ctx->last_real_time + av_rescale_q(200, ms_tb, AV_TIME_BASE_Q)) {
807 ctx->last_real_time = sub->pts;
808 ctx->screen_touched = 0;
811 ctx->buffer_changed = 0;
813 ret = ff_ass_add_rect(sub, ctx->buffer.str, ctx->readorder++, 0, NULL, NULL);
816 sub->end_display_time = -1;
819 *got_sub = sub->num_rects > 0;
823 #define OFFSET(x) offsetof(CCaptionSubContext, x)
824 #define SD AV_OPT_FLAG_SUBTITLE_PARAM | AV_OPT_FLAG_DECODING_PARAM
825 static const AVOption options[] = {
826 { "real_time", "emit subtitle events as they are decoded for real-time display", OFFSET(real_time), AV_OPT_TYPE_BOOL, { .i64 = 0 }, 0, 1, SD },
830 static const AVClass ccaption_dec_class = {
831 .class_name = "Closed caption Decoder",
832 .item_name = av_default_item_name,
834 .version = LIBAVUTIL_VERSION_INT,
837 AVCodec ff_ccaption_decoder = {
839 .long_name = NULL_IF_CONFIG_SMALL("Closed Caption (EIA-608 / CEA-708) Decoder"),
840 .type = AVMEDIA_TYPE_SUBTITLE,
841 .id = AV_CODEC_ID_EIA_608,
842 .priv_data_size = sizeof(CCaptionSubContext),
843 .init = init_decoder,
844 .close = close_decoder,
845 .flush = flush_decoder,
847 .priv_class = &ccaption_dec_class,