1 /*****************************************************************************
2 * parse.c: Philips OGT (SVCD subtitle) packet parser
3 *****************************************************************************
4 * Copyright (C) 2003, 2004 VideoLAN
7 * Authors: Rocky Bernstein
9 * Julio Sanchez Fernandez (http://subhandler.sourceforge.net)
10 * Sam Hocevar <sam@zoy.org>
11 * Laurent Aimar <fenrir@via.ecp.fr>
13 * This program is free software; you can redistribute it and/or modify
14 * it under the terms of the GNU General Public License as published by
15 * the Free Software Foundation; either version 2 of the License, or
16 * (at your option) any later version.
18 * This program is distributed in the hope that it will be useful,
19 * but WITHOUT ANY WARRANTY; without even the implied warranty of
20 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
21 * GNU General Public License for more details.
23 * You should have received a copy of the GNU General Public License
24 * along with this program; if not, write to the Free Software
25 * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111, USA.
26 *****************************************************************************/
28 /*****************************************************************************
30 *****************************************************************************/
33 #include <vlc/decoder.h>
44 /* An image color is a two-bit palette entry: 0..3 */
45 typedef uint8_t ogt_color_t;
47 /*****************************************************************************
49 *****************************************************************************/
50 static int ParseImage ( decoder_t *, subpicture_t * );
53 We do not have information on the subtitle format used on CVD's
54 except the submux sample code and a couple of samples of dubious
55 origin. Thus, this is the result of reading some code whose
56 correctness is not known and some experimentation.
58 CVD subtitles are different in several ways from SVCD OGT subtitles.
59 Image comes first and metadata is at the end. So that the metadata
60 can be found easily, the subtitle packet starts with two bytes
61 (everything is big-endian again) that give the total size of the
62 subtitle data and the offset to the metadata - i.e. size of the
63 image data plus the four bytes at the beginning.
65 Image data comes interlaced is run-length encoded. Each field is a
66 four-bit nibble. Each nibble contains a two-bit repeat count and a
67 two-bit color number so that up to three pixels can be described in
68 four bits. The function of a 0 repeat count is unknown; it might be
69 used for RLE extension. However when the full nibble is zero, the
70 rest of the line is filled with the color value in the next nibble.
71 It is unknown what happens if the color value is greater than three.
72 The rest seems to use a 4-entries palette. It is not impossible
73 that the fill-line complete case above is not as described and the
74 zero repeat count means fill line. The sample code never produces
75 this, so it may be untested.
78 void E_(ParseHeader)( decoder_t *p_dec, uint8_t *p_buffer, block_t *p_block )
80 decoder_sys_t *p_sys = p_dec->p_sys;
81 uint8_t *p = p_buffer+1;
83 dbg_print( (DECODE_DBG_CALL|DECODE_DBG_PACKET),
84 "header: 0x%02x 0x%02x 0x%02x 0x%02x, 0x%02x, 0x%02x, size: %i",
85 p_buffer[0], p_buffer[1], p_buffer[2], p_buffer[3],
86 p_buffer[4], p_buffer[5],
89 dbg_print( (DECODE_DBG_CALL|DECODE_DBG_EXT) , "");
91 p_sys->i_pts = p_block->i_pts;
92 p_sys->i_spu_size = (p[0] << 8) + p[1] + 4; p += 2;
94 /* FIXME: check data sanity */
95 p_sys->metadata_offset = GETINT16(p);
96 p_sys->metadata_length = p_sys->i_spu_size - p_sys->metadata_offset;
98 p_sys->i_image_offset = 4;
99 p_sys->i_image_length = p_sys->metadata_offset - p_sys->i_image_offset;
101 dbg_print(DECODE_DBG_PACKET, "total size: %d image size: %d\n",
102 p_sys->i_spu_size, p_sys->i_image_length);
106 #define ExtractXY(x, y) \
107 x = ((p[1]&0x0f)<<6) + (p[2]>>2); \
108 y = ((p[2]&0x03)<<8) + p[3];
112 We parse the metadata information here.
114 Although metadata information does not have to come in a fixed field
115 order, every metadata field consists of a tag byte followed by
116 parameters. In all cases known, the size including tag byte is
117 exactly four bytes in length.
120 void E_(ParseMetaInfo)( decoder_t *p_dec )
122 /* last packet in subtitle block. */
124 decoder_sys_t *p_sys = p_dec->p_sys;
125 uint8_t *p = p_sys->subtitle_data + p_sys->metadata_offset;
126 uint8_t *p_end = p + p_sys->metadata_length;
128 dbg_print( (DECODE_DBG_PACKET),
129 "subtitle packet complete, size=%d", p_sys->i_spu );
131 p_sys->state = SUBTITLE_BLOCK_COMPLETE;
134 for ( ; p < p_end; p += 4 ) {
138 case 0x04: /* subtitle duration in 1/90000ths of a second */
140 mtime_t i_duration = (p[1]<<16) + (p[2]<<8) + p[3];
141 mtime_t i_duration_scale = config_GetInt( p_dec, MODULE_STRING
142 "-duration-scaling" );
144 dbg_print( DECODE_DBG_PACKET,
145 "subtitle display duration %lu secs (scaled %lu secs)",
146 (long unsigned int) (i_duration / 90000),
147 (long unsigned int) (i_duration * i_duration_scale / 90000)
149 p_sys->i_duration = i_duration * i_duration_scale ;
154 case 0x0c: /* unknown */
155 dbg_print( DECODE_DBG_PACKET,
156 "subtitle command unknown 0x%0x 0x%0x 0x%0x 0x%0x\n",
157 p[0], p[1], p[2], p[3]);
160 case 0x17: /* coordinates of subtitle upper left x, y position */
161 ExtractXY(p_sys->i_x_start, p_sys->i_y_start);
164 case 0x1f: /* coordinates of subtitle bottom right x, y position */
168 ExtractXY(lastx, lasty);
169 p_sys->i_width = lastx - p_sys->i_x_start + 1;
170 p_sys->i_height = lasty - p_sys->i_y_start + 1;
171 dbg_print( DECODE_DBG_PACKET,
172 "end position: (%d,%d): %.2x %.2x %.2x, w x h: %dx%d",
173 lastx, lasty, p[1], p[2], p[3],
174 p_sys->i_width, p_sys->i_height );
184 uint8_t v = p[0]-0x24;
186 /* Primary Palette */
187 dbg_print( DECODE_DBG_PACKET,
188 "primary palette %d (y,u,v): (0x%0x,0x%0x,0x%0x)",
189 v, p[1], p[2], p[3]);
191 p_sys->p_palette[v].s.y = p[1];
192 p_sys->p_palette[v].s.u = p[2];
193 p_sys->p_palette[v].s.v = p[3];
203 uint8_t v = p[0]-0x2c;
205 dbg_print( DECODE_DBG_PACKET,
206 "highlight palette %d (y,u,v): (0x%0x,0x%0x,0x%0x)",
207 v, p[1], p[2], p[3]);
209 /* Highlight Palette */
210 p_sys->p_palette_highlight[v].s.y = p[1];
211 p_sys->p_palette_highlight[v].s.u = p[2];
212 p_sys->p_palette_highlight[v].s.v = p[3];
217 /* transparency for primary palette */
218 p_sys->p_palette[0].s.t = p[3] & 0x0f;
219 p_sys->p_palette[1].s.t = p[3] >> 4;
220 p_sys->p_palette[2].s.t = p[2] & 0x0f;
221 p_sys->p_palette[3].s.t = p[2] >> 4;
223 dbg_print( DECODE_DBG_PACKET,
224 "transparency for primary palette 0..3: "
225 "0x%0x 0x%0x 0x%0x 0x%0x",
226 p_sys->p_palette[0].s.t,
227 p_sys->p_palette[1].s.t,
228 p_sys->p_palette[2].s.t,
229 p_sys->p_palette[3].s.t );
234 /* transparency for highlight palette */
235 p_sys->p_palette_highlight[0].s.t = p[2] & 0x0f;
236 p_sys->p_palette_highlight[1].s.t = p[2] >> 4;
237 p_sys->p_palette_highlight[2].s.t = p[1] & 0x0f;
238 p_sys->p_palette_highlight[3].s.t = p[1] >> 4;
240 dbg_print( DECODE_DBG_PACKET,
241 "transparency for primary palette 0..3: "
242 "0x%0x 0x%0x 0x%0x 0x%0x",
243 p_sys->p_palette_highlight[0].s.t,
244 p_sys->p_palette_highlight[1].s.t,
245 p_sys->p_palette_highlight[2].s.t,
246 p_sys->p_palette_highlight[3].s.t );
251 /* offset to start of even rows of interlaced image, we correct
252 to make it relative to i_image_offset (usually 4) */
253 p_sys->first_field_offset =
254 (p[2] << 8) + p[3] - p_sys->i_image_offset;
255 dbg_print( DECODE_DBG_PACKET,
256 "first_field_offset %d", p_sys->first_field_offset);
260 /* offset to start of odd rows of interlaced image, we correct
261 to make it relative to i_image_offset (usually 4) */
262 p_sys->second_field_offset =
263 (p[2] << 8) + p[3] - p_sys->i_image_offset;
264 dbg_print( DECODE_DBG_PACKET,
265 "second_field_offset %d", p_sys->second_field_offset);
270 "unknown sequence in control header "
271 "0x%0x 0x%0x 0x%0x 0x%0x",
272 p[0], p[1], p[2], p[3]);
274 p_sys->subtitle_data_pos = 0;
279 /*****************************************************************************
280 * ParsePacket: parse an SPU packet and send it to the video output
281 *****************************************************************************
282 * This function parses the SPU packet and, if valid, sends it to the
284 *****************************************************************************/
286 E_(ParsePacket)( decoder_t *p_dec)
288 decoder_sys_t *p_sys = p_dec->p_sys;
292 dbg_print( (DECODE_DBG_CALL|DECODE_DBG_EXT) , "");
294 /* Allocate the subpicture internal data. */
295 p_spu = vout_CreateSubPicture( p_sys->p_vout, SUBT1_CHAN, TEXT_CONTENT,
302 /* In ParseImage we expand the run-length encoded color 0's; also
303 we expand pixels and remove the color palette. This should
304 facilitate scaling and antialiasing and speed up rendering.
306 p_spu->p_sys = malloc( sizeof( subpicture_sys_t )
307 + PIXEL_SIZE * (p_sys->i_width * p_sys->i_height) );
309 /* Fill the p_spu structure */
310 vlc_mutex_init( p_dec, &p_spu->p_sys->lock );
312 p_spu->pf_render = VCDSubBlend;
313 p_spu->pf_destroy = VCDSubDestroySPU;
314 p_spu->p_sys->p_data = (uint8_t*)p_spu->p_sys + sizeof( subpicture_sys_t );
316 p_spu->p_sys->i_x_end = p_sys->i_x_start + p_sys->i_width - 1;
317 p_spu->p_sys->i_y_end = p_sys->i_y_start + p_sys->i_height - 1;
319 p_spu->i_x = p_sys->i_x_start
320 + config_GetInt( p_dec, MODULE_STRING "-horizontal-correct" );
322 p_spu->p_sys->p_palette[0] = p_sys->p_palette[0];
323 p_spu->p_sys->p_palette[1] = p_sys->p_palette[1];
324 p_spu->p_sys->p_palette[2] = p_sys->p_palette[2];
325 p_spu->p_sys->p_palette[3] = p_sys->p_palette[3];
327 /* FIXME: use aspect ratio for x? */
328 p_spu->i_x = (p_spu->i_x * 3) / 4;
329 p_spu->i_y = p_sys->i_y_start
330 + config_GetInt( p_dec, MODULE_STRING "-vertical-correct" );
332 p_spu->i_width = p_sys->i_width;
333 p_spu->i_height = p_sys->i_height;
335 p_spu->i_start = p_sys->i_pts;
336 p_spu->i_stop = p_sys->i_pts + (p_sys->i_duration);
338 p_spu->p_sys->b_crop = VLC_FALSE;
339 p_spu->p_sys->i_debug = p_sys->i_debug;
341 /* Get display time now. If we do it later, we may miss the PTS. */
342 p_spu->p_sys->i_pts = p_sys->i_pts;
344 /* Attach to our input thread */
345 p_spu->p_sys->p_input = vlc_object_find( p_dec,
346 VLC_OBJECT_INPUT, FIND_PARENT );
348 /* We try to display it */
349 if( ParseImage( p_dec, p_spu ) )
351 /* There was a parse error, delete the subpicture */
352 vout_DestroySubPicture( p_sys->p_vout, p_spu );
356 /* SPU is finished - we can ask the video output to display it */
357 vout_DisplaySubPicture( p_sys->p_vout, p_spu );
361 #define advance_color_byte_pointer \
363 i_nibble_field = 2; \
365 * This is wrong, it may exceed maxp if it is the last, check \
366 * should be moved to use location or the algorithm changed to \
371 "broken subtitle - overflow while decoding " \
372 " padding (%d,%d,%d)\n", \
373 i_field, i_row, i_column ); \
374 return VLC_EGENERIC; \
377 #define CVD_FIELD_BITS (4)
378 #define CVD_FIELD_MASK ((1<<CVD_FIELD_BITS) - 1)
380 /* Get the next field - a 2-bit palette index and a run count. To do
381 this we use byte image pointer p, and i_nibble_field which
382 indicates where we are in the byte.
384 static inline uint8_t
385 ExtractField(uint8_t *p, uint8_t i_nibble_field)
387 return ( ( *p >> (CVD_FIELD_BITS*(i_nibble_field-1)) ) & CVD_FIELD_MASK );
390 /*****************************************************************************
391 * ParseImage: parse the image part of the subtitle
392 *****************************************************************************
393 This part parses the subtitle graphical data and stores it in a more
394 convenient structure for later rendering.
396 Image data comes interlaced and is run-length encoded (RLE). Each
397 field is a four-bit nibbles that is further subdivided in a two-bit
398 repeat count and a two-bit color number - up to three pixels can be
399 described in four bits. What a 0 repeat count means is unknown. It
400 might be used for RLE extension. There is a special case of a 0
401 repeat count though. When the full nibble is zero, the rest of the
402 line is filled with the color value in the next nibble. It is
403 unknown what happens if the color value is greater than three. The
404 rest seems to use a 4-entries palette. It is not impossible that the
405 fill-line complete case above is not as described and the zero repeat
406 count means fill line. The sample code never produces this, so it
409 However we'll transform this so that that the RLE is expanded and
410 interlacing will also be removed. On output each pixel entry will by
411 a 4-bit alpha (filling 8 bits), and 8-bit y, u, and v entry.
413 *****************************************************************************/
415 ParseImage( decoder_t *p_dec, subpicture_t * p_spu )
417 decoder_sys_t *p_sys = p_dec->p_sys;
419 uint8_t i_field; /* The subtitles are interlaced, are we on an
420 even or odd scanline? */
421 unsigned int i_row; /* scanline row number */
422 unsigned int i_column; /* scanline column number */
424 unsigned int i_width = p_sys->i_width;
425 unsigned int i_height = p_sys->i_height;
427 uint8_t *p_dest = (uint8_t *)p_spu->p_sys->p_data;
429 uint8_t i_nibble_field; /* The 2-bit pixels remaining in byte of *p.
431 vlc_bool_t b_filling; /* Filling i_color to the of the line. */
432 uint8_t i_pending = 0; /* number of pixels to fill with
434 ogt_color_t i_color=0; /* current pixel color: 0..3 */
435 uint8_t *p = p_sys->subtitle_data + p_sys->i_image_offset;
436 uint8_t *maxp = p + p_sys->i_image_length;
438 dbg_print( (DECODE_DBG_CALL) , "width x height: %dx%d",
441 if (p_sys && p_sys->i_debug & DECODE_DBG_IMAGE)
446 for ( i_field=0; i_field < 2; i_field++ ) {
447 i_nibble_field = 2; /* 4-bit pieces available in *p */
452 for (i=0; i< i_width * i_height; i++)
453 printf ("%02x", b[i]);
457 for ( i_row=i_field; i_row < i_height; i_row += 2 ) {
458 b_filling = VLC_FALSE;
459 for ( i_column=0; i_column<i_width; i_column++ ) {
461 /* We are in the middle of a RLE expansion, just decrement and
462 fall through with current color value */
464 } else if ( b_filling ) {
465 /* We are just filling to the end of line with one color, just
466 reuse current color value */
468 uint8_t i_val = ExtractField(p, i_nibble_field--);
469 if ( i_nibble_field == 0 ) {
470 advance_color_byte_pointer;
473 /* fill the rest of the line with next color */
474 i_color = ExtractField( p, i_nibble_field-- );
475 if ( i_nibble_field == 0 ) {
479 This is wrong, it may exceed maxp if it is the
480 last, check should be moved to use location or the
481 algorithm changed to that in vob2sub
485 "broken subtitle - overflow while decoding "
486 " filling (%d,%d,%d)",
487 i_field, i_row, i_column);
488 /* return VLC_EGENERIC; */
491 b_filling = VLC_TRUE;
493 /* Normal case: get color and repeat count,
494 this iteration will output the first (or only)
496 i_pending = (i_val >> 2);
497 i_color = i_val & 0x3;
498 /* This time counts against the total */
503 p_dest[i_row*i_width+i_column] = i_color;
505 if (p_sys && p_sys->i_debug & DECODE_DBG_IMAGE)
506 printf("%1d", i_color);
510 if ( i_nibble_field == 1 ) {
511 advance_color_byte_pointer;
514 if (p_sys && p_sys->i_debug & DECODE_DBG_IMAGE)
519 if (p_sys && (p_sys->i_debug & DECODE_DBG_IMAGE)) {
520 /* Dump out image not interlaced... */
521 VCDSubDumpImage( p_dest, i_height, i_width );
525 if (p_sys && (p_sys->i_debug & DECODE_DBG_PNG)) {
527 /* Dump image to a file in PNG format. */
529 png_text text_ptr[TEXT_COUNT];
531 text_ptr[0].key = "Preparer";
532 text_ptr[0].text = "VLC";
533 text_ptr[0].compression = PNG_TEXT_COMPRESSION_NONE;
534 text_ptr[1].key = "Description";
535 text_ptr[1].text = "CVD Subtitle";
536 text_ptr[1].compression = PNG_TEXT_COMPRESSION_NONE;
538 snprintf(filename, 300, "%s%d.png", "/tmp/vlc-cvd-sub", p_sys->i_image);
539 VCDSubDumpPNG( p_dest, p_dec, i_height, i_width, filename,
540 text_ptr, TEXT_COUNT );
542 #endif /*HAVE_LIBPNG*/
544 VCDSubHandleScaling( p_spu, p_dec );