1 /*****************************************************************************
2 * parse.c: Philips OGT (SVCD subtitle) packet parser
3 *****************************************************************************
4 * Copyright (C) 2003, 2004 VideoLAN
5 * $Id: cvd_parse.c,v 1.9 2004/01/10 13:59:25 rocky Exp $
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->comp_image_offset = 4;
99 p_sys->comp_image_length = p_sys->metadata_offset - p_sys->comp_image_offset;
101 dbg_print(DECODE_DBG_PACKET, "total size: %d image size: %d\n",
102 p_sys->i_spu_size, p_sys->comp_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 p_sys->i_duration = (p[1]<<16) + (p[2]<<8) + p[3];
142 dbg_print( DECODE_DBG_PACKET,
143 "subtitle display duration %u", p_sys->i_duration);
146 case 0x0c: /* unknown */
147 dbg_print( DECODE_DBG_PACKET,
148 "subtitle command unknown 0x%0x 0x%0x 0x%0x 0x%0x\n",
149 p[0], p[1], p[2], p[3]);
152 case 0x17: /* coordinates of subtitle upper left x, y position */
153 ExtractXY(p_sys->i_x_start, p_sys->i_y_start);
156 case 0x1f: /* coordinates of subtitle bottom right x, y position */
160 ExtractXY(lastx, lasty);
161 p_sys->i_width = lastx - p_sys->i_x_start + 1;
162 p_sys->i_height = lasty - p_sys->i_y_start + 1;
163 dbg_print( DECODE_DBG_PACKET,
164 "end position: (%d,%d): %.2x %.2x %.2x, w x h: %dx%d",
165 lastx, lasty, p[1], p[2], p[3],
166 p_sys->i_width, p_sys->i_height );
176 uint8_t v = p[0]-0x24;
178 /* Primary Palette */
179 dbg_print( DECODE_DBG_PACKET,
180 "primary palette %d (y,u,v): (0x%0x,0x%0x,0x%0x)",
181 v, p[1], p[2], p[3]);
183 p_sys->p_palette[v].s.y = p[1];
184 p_sys->p_palette[v].s.u = p[2];
185 p_sys->p_palette[v].s.v = p[3];
195 uint8_t v = p[0]-0x2c;
197 dbg_print( DECODE_DBG_PACKET,
198 "highlight palette %d (y,u,v): (0x%0x,0x%0x,0x%0x)",
199 v, p[1], p[2], p[3]);
201 /* Highlight Palette */
202 p_sys->p_palette_highlight[v].s.y = p[1];
203 p_sys->p_palette_highlight[v].s.u = p[2];
204 p_sys->p_palette_highlight[v].s.v = p[3];
209 /* transparency for primary palette */
210 p_sys->p_palette[0].s.t = p[3] & 0x0f;
211 p_sys->p_palette[1].s.t = p[3] >> 4;
212 p_sys->p_palette[2].s.t = p[2] & 0x0f;
213 p_sys->p_palette[3].s.t = p[2] >> 4;
215 dbg_print( DECODE_DBG_PACKET,
216 "transparency for primary palette 0..3: "
217 "0x%0x 0x%0x 0x%0x 0x%0x",
218 p_sys->p_palette[0].s.t,
219 p_sys->p_palette[1].s.t,
220 p_sys->p_palette[2].s.t,
221 p_sys->p_palette[3].s.t );
226 /* transparency for highlight palette */
227 p_sys->p_palette_highlight[0].s.t = p[2] & 0x0f;
228 p_sys->p_palette_highlight[1].s.t = p[2] >> 4;
229 p_sys->p_palette_highlight[2].s.t = p[1] & 0x0f;
230 p_sys->p_palette_highlight[3].s.t = p[1] >> 4;
232 dbg_print( DECODE_DBG_PACKET,
233 "transparency for primary palette 0..3: "
234 "0x%0x 0x%0x 0x%0x 0x%0x",
235 p_sys->p_palette_highlight[0].s.t,
236 p_sys->p_palette_highlight[1].s.t,
237 p_sys->p_palette_highlight[2].s.t,
238 p_sys->p_palette_highlight[3].s.t );
243 /* offset to start of even rows of interlaced image, we correct
244 to make it relative to comp_image_offset (usually 4) */
245 p_sys->first_field_offset =
246 (p[2] << 8) + p[3] - p_sys->comp_image_offset;
247 dbg_print( DECODE_DBG_PACKET,
248 "first_field_offset %d", p_sys->first_field_offset);
252 /* offset to start of odd rows of interlaced image, we correct
253 to make it relative to comp_image_offset (usually 4) */
254 p_sys->second_field_offset =
255 (p[2] << 8) + p[3] - p_sys->comp_image_offset;
256 dbg_print( DECODE_DBG_PACKET,
257 "second_field_offset %d", p_sys->second_field_offset);
262 "unknown sequence in control header "
263 "0x%0x 0x%0x 0x%0x 0x%0x",
264 p[0], p[1], p[2], p[3]);
266 p_sys->subtitle_data_pos = 0;
271 /*****************************************************************************
272 * ParsePacket: parse an SPU packet and send it to the video output
273 *****************************************************************************
274 * This function parses the SPU packet and, if valid, sends it to the
276 *****************************************************************************/
278 E_(ParsePacket)( decoder_t *p_dec)
280 decoder_sys_t *p_sys = p_dec->p_sys;
284 dbg_print( (DECODE_DBG_CALL|DECODE_DBG_EXT) , "");
286 /* Allocate the subpicture internal data. */
287 p_spu = vout_CreateSubPicture( p_sys->p_vout, MEMORY_SUBPICTURE );
293 /* In ParseImage we expand the run-length encoded color 0's; also
294 we expand pixels and remove the color palette. This should
295 facilitate scaling and antialiasing and speed up rendering.
297 p_spu->p_sys = malloc( sizeof( subpicture_sys_t )
298 + PIXEL_SIZE * (p_sys->i_width * p_sys->i_height) );
300 /* Fill the p_spu structure */
301 vlc_mutex_init( p_dec, &p_spu->p_sys->lock );
303 p_spu->pf_render = VCDSubRender;
304 p_spu->pf_destroy = VCDSubDestroySPU;
305 p_spu->p_sys->p_data = (uint8_t*)p_spu->p_sys + sizeof( subpicture_sys_t );
307 p_spu->p_sys->i_x_end = p_sys->i_x_start + p_sys->i_width - 1;
308 p_spu->p_sys->i_y_end = p_sys->i_y_start + p_sys->i_height - 1;
310 /* FIXME: use aspect ratio for x? */
311 p_spu->i_x = p_sys->i_x_start * 3 / 4;
312 p_spu->i_y = p_sys->i_y_start;
313 p_spu->i_width = p_sys->i_width;
314 p_spu->i_height = p_sys->i_height;
316 p_spu->i_start = p_sys->i_pts;
317 p_spu->i_stop = p_sys->i_pts + (p_sys->i_duration);
319 p_spu->p_sys->b_crop = VLC_FALSE;
320 p_spu->p_sys->i_debug = p_sys->i_debug;
322 /* Get display time now. If we do it later, we may miss the PTS. */
323 p_spu->p_sys->i_pts = p_sys->i_pts;
325 /* Attach to our input thread */
326 p_spu->p_sys->p_input = vlc_object_find( p_dec,
327 VLC_OBJECT_INPUT, FIND_PARENT );
329 /* We try to display it */
330 if( ParseImage( p_dec, p_spu ) )
332 /* There was a parse error, delete the subpicture */
333 vout_DestroySubPicture( p_sys->p_vout, p_spu );
337 /* SPU is finished - we can ask the video output to display it */
338 vout_DisplaySubPicture( p_sys->p_vout, p_spu );
342 #define advance_color_byte_pointer \
344 i_nibble_field = 2; \
346 * This is wrong, it may exceed maxp if it is the last, check \
347 * should be moved to use location or the algorithm changed to \
352 "broken subtitle - overflow while decoding " \
353 " padding (%d,%d,%d)\n", \
354 i_field, i_row, i_column ); \
355 return VLC_EGENERIC; \
358 #define CVD_FIELD_BITS (4)
359 #define CVD_FIELD_MASK ((1<<CVD_FIELD_BITS) - 1)
361 /* Get the next field - a 2-bit palette index and a run count. To do
362 this we use byte image pointer p, and i_nibble_field which
363 indicates where we are in the byte.
365 static inline uint8_t
366 ExtractField(uint8_t *p, uint8_t i_nibble_field)
368 return ( ( *p >> (CVD_FIELD_BITS*(i_nibble_field-1)) ) & CVD_FIELD_MASK );
371 /*****************************************************************************
372 * ParseImage: parse the image part of the subtitle
373 *****************************************************************************
374 This part parses the subtitle graphical data and stores it in a more
375 convenient structure for later rendering.
377 Image data comes interlaced and is run-length encoded (RLE). Each
378 field is a four-bit nibbles that is further subdivided in a two-bit
379 repeat count and a two-bit color number - up to three pixels can be
380 described in four bits. What a 0 repeat count means is unknown. It
381 might be used for RLE extension. There is a special case of a 0
382 repeat count though. When the full nibble is zero, the rest of the
383 line is filled with the color value in the next nibble. It is
384 unknown what happens if the color value is greater than three. The
385 rest seems to use a 4-entries palette. It is not impossible that the
386 fill-line complete case above is not as described and the zero repeat
387 count means fill line. The sample code never produces this, so it
390 However we'll transform this so that that the RLE is expanded and
391 interlacing will also be removed. On output each pixel entry will by
392 a 4-bit alpha (filling 8 bits), and 8-bit y, u, and v entry.
394 *****************************************************************************/
396 ParseImage( decoder_t *p_dec, subpicture_t * p_spu )
398 decoder_sys_t *p_sys = p_dec->p_sys;
400 uint8_t i_field; /* The subtitles are interlaced, are we on an
401 even or odd scanline? */
402 unsigned int i_row; /* scanline row number */
403 unsigned int i_column; /* scanline column number */
405 unsigned int i_width = p_sys->i_width;
406 unsigned int i_height = p_sys->i_height;
408 uint8_t *p_dest = (uint8_t *)p_spu->p_sys->p_data;
410 uint8_t i_nibble_field; /* The 2-bit pixels remaining in byte of *p.
412 vlc_bool_t b_filling; /* Filling i_color to the of the line. */
413 uint8_t i_pending = 0; /* number of pixels to fill with
415 ogt_color_t i_color=0; /* current pixel color: 0..3 */
416 uint8_t *p = p_sys->subtitle_data + p_sys->comp_image_offset;
417 uint8_t *maxp = p + p_sys->comp_image_length;
419 dbg_print( (DECODE_DBG_CALL) , "width x height: %dx%d",
422 if (p_sys && p_sys->i_debug & DECODE_DBG_IMAGE)
427 for ( i_field=0; i_field < 2; i_field++ ) {
428 i_nibble_field = 2; /* 4-bit pieces available in *p */
433 for (i=0; i< i_width * i_height; i++)
434 printf ("%02x", b[i]);
438 for ( i_row=i_field; i_row < i_height; i_row += 2 ) {
439 b_filling = VLC_FALSE;
440 for ( i_column=0; i_column<i_width; i_column++ ) {
442 /* We are in the middle of a RLE expansion, just decrement and
443 fall through with current color value */
445 } else if ( b_filling ) {
446 /* We are just filling to the end of line with one color, just
447 reuse current color value */
449 uint8_t i_val = ExtractField(p, i_nibble_field--);
450 if ( i_nibble_field == 0 ) {
451 advance_color_byte_pointer;
454 /* fill the rest of the line with next color */
455 i_color = ExtractField( p, i_nibble_field-- );
456 if ( i_nibble_field == 0 ) {
460 This is wrong, it may exceed maxp if it is the
461 last, check should be moved to use location or the
462 algorithm changed to that in vob2sub
466 "broken subtitle - overflow while decoding "
467 " filling (%d,%d,%d)",
468 i_field, i_row, i_column);
469 /* return VLC_EGENERIC; */
472 b_filling = VLC_TRUE;
474 /* Normal case: get color and repeat count,
475 this iteration will output the first (or only)
477 i_pending = (i_val >> 2);
478 i_color = i_val & 0x3;
479 /* This time counts against the total */
484 p_dest[i_row*i_width+i_column] = i_color;
486 if (p_sys && p_sys->i_debug & DECODE_DBG_IMAGE)
487 printf("%1d", i_color);
491 if ( i_nibble_field == 1 ) {
492 advance_color_byte_pointer;
495 if (p_sys && p_sys->i_debug & DECODE_DBG_IMAGE)
500 if (p_sys && (p_sys->i_debug & DECODE_DBG_IMAGE)) {
501 /* Dump out image not interlaced... */
502 VCDSubDumpImage( p_dest, i_height, i_width );
506 if (p_sys && (p_sys->i_debug & DECODE_DBG_PNG)) {
508 /* Dump image to a file in PNG format. */
510 png_text text_ptr[TEXT_COUNT];
512 text_ptr[0].key = "Preparer";
513 text_ptr[0].text = "VLC";
514 text_ptr[0].compression = PNG_TEXT_COMPRESSION_NONE;
515 text_ptr[1].key = "Description";
516 text_ptr[1].text = "CVD Subtitle";
517 text_ptr[1].compression = PNG_TEXT_COMPRESSION_NONE;
519 snprintf(filename, 300, "%s%d.png", "/tmp/vlc-cvd-sub", p_sys->i_image);
520 VCDSubDumpPNG( p_dest, p_dec, i_height, i_width, filename,
521 text_ptr, TEXT_COUNT );
523 #endif /*HAVE_LIBPNG*/
525 VCDInlinePalette( p_dest, p_sys, i_height, i_width );
527 /* The video is automatically scaled. However subtitle bitmaps
528 assume a 1:1 aspect ratio. So we need to scale to compensate for
529 or undo the effects of video output scaling.
531 Perhaps this should go in the Render routine? The advantage would
532 be that it will deal with a dynamically changing aspect ratio.
533 The downside is having to scale many times for each render call.
537 vout_thread_t *p_vout = vlc_object_find( p_spu->p_sys->p_input,
540 unsigned int i_aspect_x, i_aspect_y;
542 vout_AspectRatio( p_vout->render.i_aspect, &i_aspect_x, &i_aspect_y );
543 VCDSubScaleX( p_dec, p_spu, i_aspect_y, i_aspect_x );
547 /* To be finished...*/