• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 /*
2  * Implementation of JPEG Lite decoding algorithm
3  *
4  * Author & Copyright (c) 2003 : Sylvain Munaut <nw8xx ]at[ 246tNt.com>
5  *
6  * v4l library adaptation: Jean-François Moine <moinejf@free.fr>
7  *
8  * This program is free software; you can redistribute it and/or modify
9  * it under the terms of the GNU Lesser General Public License as published by
10  * the Free Software Foundation; either version 2.1 of the License, or
11  * (at your option) any later version.
12  *
13  * This program is distributed in the hope that it will be useful,
14  * but WITHOUT ANY WARRANTY; without even the implied warranty of
15  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
16  * GNU General Public License for more details.
17  *
18  * You should have received a copy of the GNU Lesser General Public License
19  * along with this program; if not, write to the Free Software
20  * Foundation, Inc., 51 Franklin Street, Suite 500, Boston, MA  02110-1335  USA
21 
22  * Note this code was originally licensed under the GNU GPL instead of the
23  * GNU LGPL, its license has been changed with permission, see the permission
24  * mail at the end of this file.
25  */
26 
27 /* Original WebSite: nw802.sourceforge.net */
28 
29 #include <stdlib.h>
30 #include "libv4lconvert-priv.h"
31 
32 #define RING_QUEUE_ADVANCE_INDEX(rq,ind,n) (rq)->ind = ((rq)->ind + (n))
33 #define RING_QUEUE_DEQUEUE_BYTES(rq,n) RING_QUEUE_ADVANCE_INDEX(rq,ri,n)
34 #define RING_QUEUE_PEEK(rq,ofs) ((rq)->queue[((ofs) + (rq)->ri)])
35 
36 struct RingQueue {
37 	const unsigned char *queue;
38 	int length;
39 	int ri;
40 };
41 
42 /* ============================================================================
43  * RingQueue bit reader
44  * ============================================================================
45  * All what is needed to read bit by nit from the RingQueue pump
46  * provided by usbvideo
47  * Critical part are macro and not functions to speed things up
48  * Rem: Data are read from the RingQueue as if they were 16bits Little Endian
49  *      words. Most Significants Bits are outputed first.
50  */
51 
52 /* Structure used to store what we need. */
53 /* (We may need multiple simultaneous instance from several cam) */
54 struct rqBitReader {
55 	int cur_bit;
56 	unsigned int cur_data;
57 	struct RingQueue *rq;
58 };
59 
rqBR_init(struct rqBitReader * br,struct RingQueue * rq)60 static inline void rqBR_init( struct rqBitReader *br, struct RingQueue *rq )
61 {
62 	br->cur_bit = 16;
63 	br->cur_data =
64 		RING_QUEUE_PEEK( rq, 2 )        |
65 		RING_QUEUE_PEEK( rq, 3 ) << 8   |
66 		RING_QUEUE_PEEK( rq, 0 ) << 16  |
67 		RING_QUEUE_PEEK( rq, 1 ) << 24  ;
68 	RING_QUEUE_DEQUEUE_BYTES( rq, 2 );
69 	br->rq = rq;
70 }
71 
72 #define rqBR_peekBits(br,n) ( br->cur_data >> (32-n) )
73 
74 #define rqBR_flushBits(br,n) do {                                   \
75         br->cur_data <<= n;                                         \
76         if ( (br->cur_bit -= n) <= 0 ) {                            \
77             br->cur_data |=                                         \
78                 RING_QUEUE_PEEK( br->rq, 2 ) << -br->cur_bit  |     \
79                 RING_QUEUE_PEEK( br->rq, 3 ) << (8 - br->cur_bit);  \
80             RING_QUEUE_DEQUEUE_BYTES( br->rq, 2 );                  \
81             br->cur_bit += 16;                                      \
82         }                                                           \
83 	} while (0)
84 
85 /* ============================================================================
86  * Real JPEG Lite stuff
87  * ============================================================================
88  *
89  * Precomputed tables
90  * Theses are computed at init time to make real-time operations faster.
91  * It takes some space ( about 9k ). But believe me it worth it !
92  */
93 
94 /* Variable Lenght Coding related tables, used for AC coefficient decoding
95  * TODO Check that 7 bits is enough ! */
96 static signed char vlcTbl_len[1<<10];	/* Meaningful bit count */
97 static signed char vlcTbl_run[1<<10];	/* Run */
98 static signed char vlcTbl_amp[1<<10];	/* Amplitude (without the sign) */
99 
100 /* YUV->RGB conversion table */
101 static int yuvTbl_y[256];
102 static int yuvTbl_u1[256];
103 static int yuvTbl_u2[256];
104 static int yuvTbl_v1[256];
105 static int yuvTbl_v2[256];
106 
107 /* Clamping table */
108 #define SAFE_CLAMP
109 #ifdef SAFE_CLAMP
clamp(int x)110 static inline unsigned char clamp(int x) {
111 	if (x > 255)
112 		return 255;
113 	if (x < 0)
114 		return 0;
115 	return x;
116 }
117 #define clamp_adjust(x) clamp(x+128)
118 #else
119 #define clamp(x) clampTbl[(x)+512]
120 #define clamp_adjust(x) clampTbl[(x)+640]
121 static char clampTbl[1280];
122 #endif
123 
124 /* Code to initialize those tables */
vlcTbl_init(void)125 static void vlcTbl_init(void)
126 {
127 	/* Bases tables used to compute the bigger one
128 	 * To understands theses, look at the VLC doc in the
129 	 * US patent document. */
130 
131 	static const int vlc_num = 28;
132 	static const int vlc_len[] =
133 		{ 2, 2, 3, 3, 4, 5, 5, 6, 6, 6, 6, 7, 7, 7, 7, 7,
134 		  8 ,8 ,8 ,9, 9, 9, 10, 10, 10, 10, 10, 10 };
135 	static const int vlc_run[] =
136 		{ 0, 0, 0, 1, 0, 2, 3, 1, 0, 4, 0, 5, 1, 0, -1, -2,
137 		  2, 6, 0, 3, 1, 0, 1, 0, 7, 2, 0, 8 };
138 	static const int vlc_amp[] =
139 		{ 0, 1, 2, 1, 3, 1, 1, 2, 4, 1 ,5 ,1 ,3 ,6, -1, -2,
140 		  2, 1, 7, 2, 4, 8, 5, 9, 1 ,3, 10, 1 };
141 	static const int vlc_cod[] =
142 		{ 0x000, 0x002, 0x003, 0x006, 0x00E, 0x008, 0x00B, 0x012,
143 		  0x014, 0x03D, 0x03E, 0x078, 0x079, 0x07E, 0x026, 0x027,
144 		  0x054, 0x057, 0x0FF, 0x0AA, 0x0AC, 0x1FC, 0x156, 0x157,
145 		  0x15A, 0x15B, 0x3FA, 0x3FB };
146 
147 	/* Vars */
148 	int i,j;
149 
150 	/* Main filling loop */
151 	for ( i=0 ; i<(1<<10) ; i++ ) {
152 
153 		/* Find the matching one */
154 		for ( j=0 ; j<vlc_num ; j++ ) {
155 			if ( (i >> (10-vlc_len[j])) == vlc_cod[j] ) {
156 				if ( vlc_run[j] >= 0 )
157 					if ( vlc_amp[j] != 0 )
158 						vlcTbl_len[i] = vlc_len[j] + 1;
159 					else
160 						vlcTbl_len[i] = vlc_len[j]; /* EOB */
161 				else
162 					vlcTbl_len[i] = 16;
163 				vlcTbl_run[i] = vlc_run[j];
164 				vlcTbl_amp[i] = vlc_amp[j];
165 				break;
166 			}
167 		}
168 	}
169 }
170 
yuvTbl_init(void)171 static void yuvTbl_init(void)
172 {
173 	/* These tables are just pre-multiplied and pre-offseted
174 	 * YUV by the book
175 	 * R = 1.164 * (Y-16) + 1.596 * (U-128)
176 	 * G = 1.164 * (Y-16) - 0.813 * (U-128) - 0.391 * (V-128)
177 	 * B = 1.164 * (Y-16)                   + 2.018 * (V-128) */
178 
179 	int i;
180 
181 	/* We use fixed point << 16 */
182 	for ( i=0 ; i < 256 ; i++ ) {
183 		yuvTbl_y[i]  =  76284 * (i- 16);
184 		yuvTbl_u1[i] = 104595 * (i-128);
185 		yuvTbl_u2[i] =  53281 * (i-128);
186 		yuvTbl_v1[i] =  25625 * (i-128);
187 		yuvTbl_v2[i] = 132252 * (i-128);
188 	}
189 }
190 
191 #ifndef SAFE_CLAMP
clampTbl_init(void)192 static void clampTbl_init(void)
193 {
194 	/* Instead of doing if(...) to test for overrange, we use
195 	 * a clamping table */
196 
197 	int i;
198 
199 	for (i=0 ; i < 512 ; i++)
200 		clampTbl[i] = 0;
201 	for (i=512 ; i < 768 ; i++ )
202 		clampTbl[i] = i - 512;
203 	for (i=768 ; i < 1280 ; i++ )
204 		clampTbl[i] = 255;
205 
206 }
207 #endif
208 
209 /*
210  * Internal helpers
211  */
212 
readAC(struct rqBitReader * br,int * run,int * amp)213 static inline int readAC( struct rqBitReader *br, int *run, int *amp )
214 {
215 	/* Vars */
216 	unsigned int cod;
217 
218 	/* Get 16 bits */
219 	cod = rqBR_peekBits(br,16);
220 
221 	/* Lookup in the table */
222 	*run = vlcTbl_run[cod>>6];
223 	*amp = vlcTbl_amp[cod>>6];
224 	rqBR_flushBits(br,vlcTbl_len[cod>>6]);
225 
226 	if (*amp > 0) {
227 
228 		/* Normal stuff, just correct the sign */
229 		if (cod & (0x10000 >> vlcTbl_len[cod>>6]))
230 			*amp = - *amp;
231 	} else {
232 
233 		/* Handle special cases */
234 		if (!*amp)
235 			return 0;
236 		if (*amp == -1) {
237 
238 			/* 0100110srrraaaaa */
239 			*run = ( cod >> 5 ) & 0x07;
240 			*amp = ( cod & 0x100) ?
241 				-(cod&0x1F) : (cod&0x1F);
242 		} else {
243 
244 			/* 0100111srrrraaaa */
245 			*run = ( cod >> 4 ) & 0x0F;
246 			*amp = ( cod & 0x100) ?
247 				-(cod&0x0F) : (cod&0x0F);
248 		}
249 	}
250 
251 	return 1;
252 }
253 
254 
255 #define iDCT_column(b0,b1,b2,b3) do {	\
256 	int t0,t1,t2,t3;                    \
257 					\
258 	t0 = ( b1 + b3 ) << 5;              \
259 	t2 = t0 - (b3 << 4);                \
260 	t3 = (b1 *  47) - t0;               \
261 	t0 = b0 + b2;                       \
262 	t1 = b0 - b2;                       \
263 					\
264 	b0 = ( t0 + t2 );                   \
265 	b1 = ( t1 + t3 );                   \
266 	b3 = ( t0 - t2 );                   \
267 	b2 = ( t1 - t3 );                   \
268 } while (0)
269 
270 #define iDCT_line(b0,b1,b2,b3) do {		\
271 	int t0,t1,t2,t3,bm0,bm2;            \
272 					\
273 	bm0 = b0 << 7;                      \
274 	bm2 = b2 << 7;                      \
275 					\
276 	t0 = bm0 + bm2;                     \
277 	t1 = bm0 - bm2;                     \
278 	t2 = b1 * 183 + b3 *  86;           \
279 	t3 = b1 *  86 - b3 * 183;           \
280 					\
281 	b0 = ( t0 + t2 ) >> 22;             \
282 	b1 = ( t1 + t3 ) >> 22;             \
283 	b3 = ( t0 - t2 ) >> 22;             \
284 	b2 = ( t1 - t3 ) >> 22;             \
285 } while (0)
286 
287 /* Decode a block
288  * Basic ops : get the DC - get the ACs - deZigZag - deWeighting -
289  *             deQuantization - iDCT
290  * Here they are a little mixed-up to speed all this up.
291  */
decodeBlock(struct rqBitReader * br,int * block,int * dc)292 static inline int decodeBlock( struct rqBitReader *br, int *block, int *dc )
293 {
294 	/* Tables used for block decoding */
295 
296 		/* deZigZag table
297 		 *
298 		 * ZigZag: each of the coefficient of the DCT transformed 4x4
299 		 *         matrix is taken in a certain order to make a linear
300 		 *         array with the high frequency AC at the end
301 		 *
302 		 * / 0  1  5  6 \    .
303 		 * | 2  4  7 12 |    This is the order taken. We must deZigZag
304 		 * | 3  8 11 13 |    to reconstitute the original matrix
305 		 * \ 9 10 14 15 /
306 		 */
307 	static const int iZigZagTbl[16] =
308 		{ 0, 1, 4, 8, 5, 2, 3, 6,  9,12, 13, 10, 7, 11, 14, 15 };
309 
310 		/* deQuantization, deWeighting & iDCT premultiply */
311 
312 		/*
313 		 * Weighting : Each DCT coefficient is weighted by a certain factor. We
314 		 *             must compensate for this to rebuilt the original DCT matrix.
315 		 *
316 		 * Quantization: According to the read Q factor, DCT coefficient are
317 		 *               quantized. We need to compensate for this.
318 		 *
319 		 * iDCT premultiply: Since for the first iDCT pass ( column ), we'll need
320 		 *                   to do some multiplication, the ones that we can
321 		 *                   integrate here, we do.
322 		 *
323 		 * Rem: - The factors are here presented in the ZigZaged order,
324 		 *      because we will need those BEFORE the deZigZag
325 		 *      - For more informations, consult jpgl_tbl.c, it's the little
326 		 *      prog that computes this table
327 		 */
328 	static const int iQWTbl[4][16] = {
329 		{  32768,  17808,    794,  18618,    850,  18618,  43115,   1828,
330 		   40960,   1924,   2089,  45511,   2089,  49648,   2216,   2521 },
331 		{  32768,  35617,   1589,  37236,   1700,  37236,  86231,   3656,
332 		   81920,   3849,   4179,  91022,   4179,  99296,   4432,   5043 },
333 		{  32768,  71234,   3179,  74472,   3401,  74472, 172463,   7313,
334 		  163840,   7698,   8358, 182044,   8358, 198593,   8865,  10087 },
335 		{  32768, 142469,   6359, 148945,   6803, 148945, 344926,  14627,
336 		  327680,  15397,  16716, 364088,  16716, 397187,  17730,  20175 }
337 	};
338 
339 	/* Vars */
340 	int hdr;
341 	int *eff_iQWTbl;
342 	int cc, run, amp;
343 
344 	/* Read & Decode the block header ( Q, T, DC ) */
345 	hdr = rqBR_peekBits(br,11);
346 
347 	if (hdr & 0x100) {
348 		/* Differential mode */
349 		if (hdr & 0x80)
350 			*dc += ( hdr >> 3 ) | ~0xF;
351 		else
352 			*dc += ( hdr >> 3 ) & 0xF;
353 
354 		/* Flush the header bits */
355 		rqBR_flushBits(br,8);
356 	} else {
357 		/* Direct mode */
358 		if ( hdr & 0x80 )
359 			*dc = hdr | ~0x7F;
360 		else
361 			*dc = hdr & 0x7F;
362 
363 		/* Flush the header bits */
364 		rqBR_flushBits(br,11);
365 	}
366 
367 	/* Clear the block & store DC ( with pre-multiply ) */
368 	block[0] = *dc << 15;
369 	block[1] = 0x00;
370 	block[2] = 0x00;
371 	block[3] = 0x00;
372 	block[4] = 0x00;
373 	block[5] = 0x00;
374 	block[6] = 0x00;
375 	block[7] = 0x00;
376 	block[8] = 0x00;
377 	block[9] = 0x00;
378 	block[10] = 0x00;
379 	block[11] = 0x00;
380 	block[12] = 0x00;
381 	block[13] = 0x00;
382 	block[14] = 0x00;
383 	block[15] = 0x00;
384 
385 	/* Read the AC coefficients
386 	 * at the same time, deZigZag, deQuantization, deWeighting & iDCT premultiply
387 	 */
388 	eff_iQWTbl = (int*) iQWTbl[hdr>>9];
389 	cc = 0;
390 
391 	while ( readAC(br,&run,&amp) ) {
392 		cc += run + 1;
393 		if ( cc > 15 )
394 			return -1;
395 		block[iZigZagTbl[cc]] = amp * eff_iQWTbl[cc];
396 	}
397 
398 	/* Do the column iDCT ( what's left to do ) */
399 	iDCT_column(block[0], block[4], block[8], block[12]);
400 	iDCT_column(block[1], block[5], block[9], block[13]);
401 	iDCT_column(block[2], block[6], block[10], block[14]);
402 	iDCT_column(block[3], block[7], block[11], block[15]);
403 
404 	/* Do the line iDCT ( complete one here ) */
405 	iDCT_line(block[0], block[1], block[2], block[3]);
406 	iDCT_line(block[4], block[5], block[6], block[7]);
407 	iDCT_line(block[8], block[9], block[10], block[11]);
408 	iDCT_line(block[12], block[13], block[14], block[15]);
409 
410 	return !(hdr & 0x700);
411 }
412 
v4lconvert_decode_jpgl(const unsigned char * inp,int src_size,unsigned int dest_pix_fmt,unsigned char * fb,int img_width,int img_height)413 int v4lconvert_decode_jpgl(const unsigned char *inp, int src_size,
414 		unsigned int dest_pix_fmt, unsigned char *fb,
415 		int img_width, int img_height)
416 {
417 	/* Vars */
418 	struct RingQueue rq;
419 	struct rqBitReader br;
420 
421 	int row, col;	/* Row & Column in the image */
422 
423 	int x,y;
424 	int block_idx;
425 
426 	unsigned char *Yline_baseptr, *Uline_baseptr, *Vline_baseptr;
427 	unsigned char *Yline, *Uline, *Vline;
428 	int Yline_baseofs, UVline_baseofs;
429 
430 	int dc_y, dc_u, dc_v;	/* DC Coefficients */
431 	int block_y[16*4];		/* Y blocks */
432 	int block_u[16];		/* U block */
433 	int block_v[16];		/* V block */
434 
435 	unsigned char *mainbuffer;
436 
437 	int yc,uc,vc;
438 
439 	/* init the decoder */
440 	if (yuvTbl_y[0] == 0) {
441 		vlcTbl_init();
442 		yuvTbl_init();
443 #ifndef SAFE_CLAMP
444 		clampTbl_init();
445 #endif
446 	}
447 
448 	img_height /= 4;
449 
450 	/* Prepare a bit-by-bit reader */
451 	rq.queue = inp;
452 	rq.length = src_size;
453 	rq.ri = 0;
454 	rqBR_init(&br, &rq);
455 
456 	/* Allocate a big buffer & setup pointers */
457 	switch (dest_pix_fmt) {
458 	default:
459 /*	case V4L2_PIX_FMT_RGB24: */
460 /*	case V4L2_PIX_FMT_BGR24: */
461 		mainbuffer = malloc(4 * (img_width + (img_width >> 1) + 2));
462 
463 		Yline_baseptr = mainbuffer;
464 		Uline_baseptr = mainbuffer + (4 * img_width);
465 		Vline_baseptr = Uline_baseptr + (img_width + 4);
466 		break;
467 	case V4L2_PIX_FMT_YUV420:
468 		mainbuffer = NULL;
469 		Yline_baseptr = fb;
470 		Uline_baseptr = fb + img_width * img_height * 16;
471 		Vline_baseptr = Uline_baseptr + img_width * img_height * 4;
472 		break;
473 	case V4L2_PIX_FMT_YVU420:
474 		mainbuffer = NULL;
475 		Yline_baseptr = fb;
476 		Vline_baseptr = fb + img_width * img_height * 16;
477 		Uline_baseptr = Vline_baseptr + img_width * img_height * 4;
478 		break;
479 	}
480 
481 	Yline_baseofs = img_width - 4;
482 	UVline_baseofs = (img_width >> 2) - 3;
483 
484 	/* Process 4 lines at a time ( one block height ) */
485 	for ( row=0 ; row<img_height ; row++ ) {
486 		/* Line start reset DC */
487 		dc_y = dc_u = dc_v = 0;
488 
489 		/* Process 16 columns at a time ( 4 block width ) */
490 		for ( col=0 ; col<img_width ; col+=16 ) {
491 			/* Decode blocks
492 			 * Block order : Y Y Y Y V U ( Why V before U ?
493 			 * that just depends what you call U&V ... I took the
494 			 * 'by-the-book' names and that makes V and then U,
495 			 * ... just ask the DivIO folks ;) )
496 			 */
497 			if ( decodeBlock(&br, block_y, &dc_y) && (!col) )
498 /*				return;		 * Bad block, so bad frame ... */
499 				;
500 
501 			decodeBlock(&br, block_y + 16, &dc_y);
502 			decodeBlock(&br, block_y + 32, &dc_y);
503 			decodeBlock(&br, block_y + 48, &dc_y);
504 			decodeBlock(&br, block_v, &dc_v);
505 			decodeBlock(&br, block_u, &dc_u);
506 
507 			/* Copy data to temporary buffers ( to make a complete line ) */
508 			block_idx = 0;
509 			Yline = Yline_baseptr + col;
510 			Uline = Uline_baseptr + (col >> 2);
511 			Vline = Vline_baseptr + (col >> 2);
512 
513 			for ( y=0 ; y<4 ; y++) {
514 				/* Scan line */
515 				for ( x=0 ; x<4 ; x++ ) {
516 					/* Y block */
517 					Yline[ 0] = clamp_adjust(block_y[block_idx   ]);
518 					Yline[ 4] = clamp_adjust(block_y[block_idx+16]);
519 					Yline[ 8] = clamp_adjust(block_y[block_idx+32]);
520 					Yline[12] = clamp_adjust(block_y[block_idx+48]);
521 
522 					/* U block */
523 					*Uline = clamp_adjust(block_u[block_idx]);
524 
525 					/* V block */
526 					*Vline = clamp_adjust(block_v[block_idx]);
527 
528 					/* Ajust pointers & index */
529 					block_idx++;
530 					Yline++;
531 					Uline++;
532 					Vline++;
533 				}
534 
535 				/* Adjust pointers */
536 				Yline += Yline_baseofs;
537 				Uline += UVline_baseofs;
538 				Vline += UVline_baseofs;
539 			}
540 		}
541 
542 		/* Handle interpolation special case ( at the end of the lines ) */
543 		Uline = Uline_baseptr + (UVline_baseofs+2);
544 		Vline = Vline_baseptr + (UVline_baseofs+2);
545 		for ( y=0 ; y<4 ; y++ ) {
546 			/* Copy the last pixel */
547 			Uline[1] = Uline[0];
548 			Vline[1] = Vline[0];
549 
550 			/* Adjust ptr */
551 			Uline += UVline_baseofs+4;
552 			Vline += UVline_baseofs+4;
553 		}
554 
555 		/* We have 4 complete lines, so tempbuffer<YUV> -> framebuffer<RGB>
556 		 * Go line by line */
557 
558 		switch (dest_pix_fmt) {
559 		case V4L2_PIX_FMT_RGB24:
560 		    Yline = Yline_baseptr;
561 		    Uline = Uline_baseptr;
562 		    Vline = Vline_baseptr;
563 		    for ( y=0 ; y<4 ; y++ ) {
564 			/* Process 4 pixel at a time to handle interpolation
565 			 * for U & V values */
566 			for ( x=0 ; x<img_width ; x+=4 ) {
567 				/* First pixel */
568 				yc = yuvTbl_y[*(Yline++)];
569 				uc = Uline[0];
570 				vc = Vline[0];
571 
572 				*(fb++) = clamp(( yc + yuvTbl_u1[uc] ) >> 16);
573 				*(fb++) = clamp(( yc - yuvTbl_u2[uc] - yuvTbl_v1[vc] ) >> 16);
574 				*(fb++) = clamp(( yc + yuvTbl_v2[vc] ) >> 16);
575 
576 				/* Second pixel */
577 				yc = yuvTbl_y[*(Yline++)];
578 				uc = ( 3*Uline[0] + Uline[1] ) >> 2;
579 				vc = ( 3*Vline[0] + Vline[1] ) >> 2;
580 
581 				*(fb++) = clamp(( yc + yuvTbl_u1[uc] ) >> 16);
582 				*(fb++) = clamp(( yc - yuvTbl_u2[uc] - yuvTbl_v1[vc] ) >> 16);
583 				*(fb++) = clamp(( yc + yuvTbl_v2[vc] ) >> 16);
584 
585 				/* Third pixel */
586 				yc = yuvTbl_y[*(Yline++)];
587 				uc = ( Uline[0] + Uline[1] ) >> 1;
588 				vc = ( Vline[0] + Vline[1] ) >> 1;
589 
590 				*(fb++) = clamp(( yc + yuvTbl_u1[uc] ) >> 16);
591 				*(fb++) = clamp(( yc - yuvTbl_u2[uc] - yuvTbl_v1[vc] ) >> 16);
592 				*(fb++) = clamp(( yc + yuvTbl_v2[vc] ) >> 16);
593 
594 				/* Fourth pixel */
595 				yc = yuvTbl_y[*(Yline++)];
596 				uc = ( Uline[0] + 3*Uline[1] ) >> 2;
597 				vc = ( Vline[0] + 3*Vline[1] ) >> 2;
598 
599 				*(fb++) = clamp(( yc + yuvTbl_u1[uc] ) >> 16);
600 				*(fb++) = clamp(( yc - yuvTbl_u2[uc] - yuvTbl_v1[vc] ) >> 16);
601 				*(fb++) = clamp(( yc + yuvTbl_v2[vc] ) >> 16);
602 
603 				/* Adjust pointers */
604 				Uline++;
605 				Vline++;
606 			}
607 
608 			/* Adjust pointers */
609 			Uline++;
610 			Vline++;
611 		    }
612 		    break;
613 		case V4L2_PIX_FMT_BGR24:
614 		    Yline = Yline_baseptr;
615 		    Uline = Uline_baseptr;
616 		    Vline = Vline_baseptr;
617 		    for ( y=0 ; y<4 ; y++ ) {
618 			/* Process 4 pixel at a time to handle interpolation
619 			 * for U & V values */
620 			for ( x=0 ; x<img_width ; x+=4 ) {
621 				/* First pixel */
622 				yc = yuvTbl_y[*(Yline++)];
623 				uc = Uline[0];
624 				vc = Vline[0];
625 
626 				*(fb++) = clamp(( yc + yuvTbl_v2[vc] ) >> 16);
627 				*(fb++) = clamp(( yc - yuvTbl_u2[uc] - yuvTbl_v1[vc] ) >> 16);
628 				*(fb++) = clamp(( yc + yuvTbl_u1[uc] ) >> 16);
629 
630 				/* Second pixel */
631 				yc = yuvTbl_y[*(Yline++)];
632 				uc = ( 3*Uline[0] + Uline[1] ) >> 2;
633 				vc = ( 3*Vline[0] + Vline[1] ) >> 2;
634 
635 				*(fb++) = clamp(( yc + yuvTbl_v2[vc] ) >> 16);
636 				*(fb++) = clamp(( yc - yuvTbl_u2[uc] - yuvTbl_v1[vc] ) >> 16);
637 				*(fb++) = clamp(( yc + yuvTbl_u1[uc] ) >> 16);
638 
639 				/* Third pixel */
640 				yc = yuvTbl_y[*(Yline++)];
641 				uc = ( Uline[0] + Uline[1] ) >> 1;
642 				vc = ( Vline[0] + Vline[1] ) >> 1;
643 
644 				*(fb++) = clamp(( yc + yuvTbl_v2[vc] ) >> 16);
645 				*(fb++) = clamp(( yc - yuvTbl_u2[uc] - yuvTbl_v1[vc] ) >> 16);
646 				*(fb++) = clamp(( yc + yuvTbl_u1[uc] ) >> 16);
647 
648 				/* Fourth pixel */
649 				yc = yuvTbl_y[*(Yline++)];
650 				uc = ( Uline[0] + 3*Uline[1] ) >> 2;
651 				vc = ( Vline[0] + 3*Vline[1] ) >> 2;
652 
653 				*(fb++) = clamp(( yc + yuvTbl_v2[vc] ) >> 16);
654 				*(fb++) = clamp(( yc - yuvTbl_u2[uc] - yuvTbl_v1[vc] ) >> 16);
655 				*(fb++) = clamp(( yc + yuvTbl_u1[uc] ) >> 16);
656 
657 				/* Adjust pointers */
658 				Uline++;
659 				Vline++;
660 			}
661 
662 			/* Adjust pointers */
663 			Uline++;
664 			Vline++;
665 		    }
666 		    break;
667 		case V4L2_PIX_FMT_YUV420:
668 		case V4L2_PIX_FMT_YVU420:
669 		    Yline_baseptr += img_width * 4;
670 		    Uline_baseptr += img_width;
671 		    Vline_baseptr += img_width;
672 		    break;
673 		}
674 	}
675 
676 	/* Free our buffer */
677 	if (mainbuffer != NULL)
678 		free(mainbuffer);
679 
680 	return 0;
681 }
682 
683 /*
684 Return-Path: tnt@246tNt.com
685 Received: from zimbra16-e3.priv.proxad.net (LHLO
686  zimbra16-e3.priv.proxad.net) (172.20.243.166) by
687  zimbra16-e3.priv.proxad.net with LMTP; Mon, 14 Feb 2011 21:10:38 +0100
688  (CET)
689 Received: from mailrelay011.isp.belgacom.be (mx26-g26.priv.proxad.net [172.20.243.96])
690 	by zimbra16-e3.priv.proxad.net (Postfix) with ESMTP id 1A661157C5B
691 	for <moinejf@free.fr>; Mon, 14 Feb 2011 21:10:38 +0100 (CET)
692 Received: from mailrelay011.isp.belgacom.be ([195.238.6.178])
693 	by mx1-g20.free.fr (MXproxy) for moinejf@free.fr ;
694 	Mon, 14 Feb 2011 21:10:36 +0100 (CET)
695 X-ProXaD-SC: state=HAM score=0
696 X-Belgacom-Dynamic: yes
697 X-IronPort-Anti-Spam-Filtered: true
698 X-IronPort-Anti-Spam-Result: ApIBAKsaWU1XQ5W2/2dsb2JhbAAMhBHOEpA5gSeBaYFYdgSLfw
699 Received: from 182.149-67-87.adsl-dyn.isp.belgacom.be (HELO [10.0.0.129]) ([87.67.149.182])
700   by relay.skynet.be with ESMTP; 14 Feb 2011 21:10:36 +0100
701 Message-ID: <4D598C7C.7080307@246tNt.com>
702 Date: Mon, 14 Feb 2011 21:11:40 +0100
703 From: Sylvain Munaut <tnt@246tNt.com>
704 User-Agent: Mozilla/5.0 (X11; U; Linux x86_64; en-US; rv:1.9.2.13) Gecko/20101219 Lightning/1.0b3pre Thunderbird/3.1.7
705 MIME-Version: 1.0
706 To: Jean-Francois Moine <moinejf@free.fr>
707 CC: Kjell Claesson <keyson@users.sourceforge.net>
708 Subject: Re: nw80x as a gspca subdriv
709 References: <20110209204208.4b19df88@tele>	<4D53B3BF.9050908@246tNt.com> <20110214205107.18c29303@tele>
710 In-Reply-To: <20110214205107.18c29303@tele>
711 Content-Type: text/plain; charset=UTF-8
712 Content-Transfer-Encoding: 7bit
713 
714 	[snip]
715 > May I have your permission to relicense your JPEG Lite decompression
716 > code under the LGPL (version 2 or later)?
717 
718 Yes, sure.
719 
720 """
721 I hereby allow the nw80x driver code, including the jpeg lite decoding
722 routines, to be used and distributed under the LGPL v2 or later.
723 """
724 	[snip]
725 Cheers,
726 
727     Sylvain
728  */
729