[svn] / branches / release-1_0-branch / xvidcore / src / dct / idct.c Repository:
ViewVC logotype

Annotation of /branches/release-1_0-branch/xvidcore/src/dct/idct.c

Parent Directory Parent Directory | Revision Log Revision Log


Revision 677 - (view) (download)
Original Path: trunk/xvidcore/src/dct/idct.c

1 : chl 461 /*****************************************************************************
2 : Isibaar 3 *
3 : chl 461 * XVID MPEG-4 VIDEO CODEC
4 :     * - inverse fast disrete cosine transformation - integer C version
5 : Isibaar 3 *
6 : chl 461 * These routines are from Independent JPEG Group's free JPEG software
7 :     * Copyright (C) 1991-1998, Thomas G. Lane (see the file README.IJG)
8 : Isibaar 3 *
9 : edgomez 649 * This file is part of XviD, a free MPEG-4 video encoder/decoder
10 : Isibaar 3 *
11 : edgomez 649 * XviD is free software; you can redistribute it and/or modify it
12 :     * under the terms of the GNU General Public License as published by
13 : chl 461 * the Free Software Foundation; either version 2 of the License, or
14 :     * (at your option) any later version.
15 : Isibaar 3 *
16 : chl 461 * This program is distributed in the hope that it will be useful,
17 :     * but WITHOUT ANY WARRANTY; without even the implied warranty of
18 :     * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
19 :     * GNU General Public License for more details.
20 :     *
21 :     * You should have received a copy of the GNU General Public License
22 :     * along with this program; if not, write to the Free Software
23 :     * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
24 :     *
25 : edgomez 649 * Under section 8 of the GNU General Public License, the copyright
26 :     * holders of XVID explicitly forbid distribution in the following
27 :     * countries:
28 :     *
29 :     * - Japan
30 :     * - United States of America
31 :     *
32 :     * Linking XviD statically or dynamically with other modules is making a
33 :     * combined work based on XviD. Thus, the terms and conditions of the
34 :     * GNU General Public License cover the whole combination.
35 :     *
36 :     * As a special exception, the copyright holders of XviD give you
37 :     * permission to link XviD with independent modules that communicate with
38 :     * XviD solely through the VFW1.1 and DShow interfaces, regardless of the
39 :     * license terms of these independent modules, and to copy and distribute
40 :     * the resulting combined work under terms of your choice, provided that
41 :     * every copy of the combined work is accompanied by a complete copy of
42 :     * the source code of XviD (the version of XviD used to produce the
43 :     * combined work), being distributed under the terms of the GNU General
44 :     * Public License plus this exception. An independent module is a module
45 :     * which is not derived from or based on XviD.
46 :     *
47 :     * Note that people who make modified versions of XviD are not obligated
48 :     * to grant this special exception for their modified versions; it is
49 :     * their choice whether to do so. The GNU General Public License gives
50 :     * permission to release a modified version without this exception; this
51 :     * exception also makes it possible to release a modified version which
52 :     * carries forward this exception.
53 :     *
54 : edgomez 677 * $Id: idct.c,v 1.5 2002-11-26 23:44:10 edgomez Exp $
55 : edgomez 649 *
56 : chl 461 *************************************************************************/
57 : Isibaar 3
58 :     /**********************************************************/
59 :     /* inverse two dimensional DCT, Chen-Wang algorithm */
60 :     /* (cf. IEEE ASSP-32, pp. 803-816, Aug. 1984) */
61 :     /* 32-bit integer arithmetic (8 bit coefficients) */
62 :     /* 11 mults, 29 adds per DCT */
63 :     /* sE, 18.8.91 */
64 :     /**********************************************************/
65 :     /* coefficients extended to 12 bit for IEEE1180-1990 */
66 :     /* compliance sE, 2.1.94 */
67 :     /**********************************************************/
68 :    
69 :     /* this code assumes >> to be a two's-complement arithmetic */
70 :     /* right shift: (-2)>>1 == -1 , (-3)>>1 == -2 */
71 :    
72 :     #include "idct.h"
73 :    
74 : edgomez 195 #define W1 2841 /* 2048*sqrt(2)*cos(1*pi/16) */
75 :     #define W2 2676 /* 2048*sqrt(2)*cos(2*pi/16) */
76 :     #define W3 2408 /* 2048*sqrt(2)*cos(3*pi/16) */
77 :     #define W5 1609 /* 2048*sqrt(2)*cos(5*pi/16) */
78 :     #define W6 1108 /* 2048*sqrt(2)*cos(6*pi/16) */
79 :     #define W7 565 /* 2048*sqrt(2)*cos(7*pi/16) */
80 : Isibaar 3
81 :    
82 :     /* global declarations */
83 : edgomez 677 /*void init_idct_int32 (void); */
84 :     /*void idct_int32 (short *block); */
85 : Isibaar 3
86 :     /* private data */
87 : edgomez 195 static short iclip[1024]; /* clipping table */
88 : Isibaar 3 static short *iclp;
89 :    
90 :     /* private prototypes */
91 : edgomez 677 /*static void idctrow _ANSI_ARGS_((short *blk)); */
92 :     /*static void idctcol _ANSI_ARGS_((short *blk)); */
93 : Isibaar 3
94 :     /* row (horizontal) IDCT
95 :     *
96 :     * 7 pi 1
97 :     * dst[k] = sum c[l] * src[l] * cos( -- * ( k + - ) * l )
98 :     * l=0 8 2
99 :     *
100 :     * where: c[0] = 128
101 :     * c[1..7] = 128*sqrt(2)
102 :     */
103 :    
104 : edgomez 677 #if 0
105 : Isibaar 3 static void idctrow(blk)
106 :     short *blk;
107 :     {
108 :     int X0, X1, X2, X3, X4, X5, X6, X7, X8;
109 :    
110 : edgomez 677 /* shortcut */
111 : Isibaar 3 if (!((X1 = blk[4]<<11) | (X2 = blk[6]) | (X3 = blk[2]) |
112 :     (X4 = blk[1]) | (X5 = blk[7]) | (X6 = blk[5]) | (X7 = blk[3])))
113 :     {
114 :     blk[0]=blk[1]=blk[2]=blk[3]=blk[4]=blk[5]=blk[6]=blk[7]=blk[0]<<3;
115 :     return;
116 :     }
117 :    
118 : edgomez 677 X0 = (blk[0]<<11) + 128; /* for proper rounding in the fourth stage */
119 : Isibaar 3
120 : edgomez 677 /* first stage */
121 : Isibaar 3 X8 = W7*(X4+X5);
122 :     X4 = X8 + (W1-W7)*X4;
123 :     X5 = X8 - (W1+W7)*X5;
124 :     X8 = W3*(X6+X7);
125 :     X6 = X8 - (W3-W5)*X6;
126 :     X7 = X8 - (W3+W5)*X7;
127 :    
128 : edgomez 677 /* second stage */
129 : Isibaar 3 X8 = X0 + X1;
130 :     X0 -= X1;
131 :     X1 = W6*(X3+X2);
132 :     X2 = X1 - (W2+W6)*X2;
133 :     X3 = X1 + (W2-W6)*X3;
134 :     X1 = X4 + X6;
135 :     X4 -= X6;
136 :     X6 = X5 + X7;
137 :     X5 -= X7;
138 :    
139 : edgomez 677 /* third stage */
140 : Isibaar 3 X7 = X8 + X3;
141 :     X8 -= X3;
142 :     X3 = X0 + X2;
143 :     X0 -= X2;
144 :     X2 = (181*(X4+X5)+128)>>8;
145 :     X4 = (181*(X4-X5)+128)>>8;
146 :    
147 : edgomez 677 /* fourth stage */
148 : Isibaar 3 blk[0] = (X7+X1)>>8;
149 :     blk[1] = (X3+X2)>>8;
150 :     blk[2] = (X0+X4)>>8;
151 :     blk[3] = (X8+X6)>>8;
152 :     blk[4] = (X8-X6)>>8;
153 :     blk[5] = (X0-X4)>>8;
154 :     blk[6] = (X3-X2)>>8;
155 :     blk[7] = (X7-X1)>>8;
156 : edgomez 677 }
157 :     #endif
158 : Isibaar 3
159 :     /* column (vertical) IDCT
160 :     *
161 :     * 7 pi 1
162 :     * dst[8*k] = sum c[l] * src[8*l] * cos( -- * ( k + - ) * l )
163 :     * l=0 8 2
164 :     *
165 :     * where: c[0] = 1/1024
166 :     * c[1..7] = (1/1024)*sqrt(2)
167 :     */
168 : edgomez 677 #if 0
169 : Isibaar 3 static void idctcol(blk)
170 :     short *blk;
171 :     {
172 :     int X0, X1, X2, X3, X4, X5, X6, X7, X8;
173 :    
174 : edgomez 677 /* shortcut */
175 : Isibaar 3 if (!((X1 = (blk[8*4]<<8)) | (X2 = blk[8*6]) | (X3 = blk[8*2]) |
176 :     (X4 = blk[8*1]) | (X5 = blk[8*7]) | (X6 = blk[8*5]) | (X7 = blk[8*3])))
177 :     {
178 :     blk[8*0]=blk[8*1]=blk[8*2]=blk[8*3]=blk[8*4]=blk[8*5]=blk[8*6]=blk[8*7]=
179 :     iclp[(blk[8*0]+32)>>6];
180 :     return;
181 :     }
182 :    
183 :     X0 = (blk[8*0]<<8) + 8192;
184 :    
185 : edgomez 677 /* first stage */
186 : Isibaar 3 X8 = W7*(X4+X5) + 4;
187 :     X4 = (X8+(W1-W7)*X4)>>3;
188 :     X5 = (X8-(W1+W7)*X5)>>3;
189 :     X8 = W3*(X6+X7) + 4;
190 :     X6 = (X8-(W3-W5)*X6)>>3;
191 :     X7 = (X8-(W3+W5)*X7)>>3;
192 :    
193 : edgomez 677 /* second stage */
194 : Isibaar 3 X8 = X0 + X1;
195 :     X0 -= X1;
196 :     X1 = W6*(X3+X2) + 4;
197 :     X2 = (X1-(W2+W6)*X2)>>3;
198 :     X3 = (X1+(W2-W6)*X3)>>3;
199 :     X1 = X4 + X6;
200 :     X4 -= X6;
201 :     X6 = X5 + X7;
202 :     X5 -= X7;
203 :    
204 : edgomez 677 /* third stage */
205 : Isibaar 3 X7 = X8 + X3;
206 :     X8 -= X3;
207 :     X3 = X0 + X2;
208 :     X0 -= X2;
209 :     X2 = (181*(X4+X5)+128)>>8;
210 :     X4 = (181*(X4-X5)+128)>>8;
211 :    
212 : edgomez 677 /* fourth stage */
213 : Isibaar 3 blk[8*0] = iclp[(X7+X1)>>14];
214 :     blk[8*1] = iclp[(X3+X2)>>14];
215 :     blk[8*2] = iclp[(X0+X4)>>14];
216 :     blk[8*3] = iclp[(X8+X6)>>14];
217 :     blk[8*4] = iclp[(X8-X6)>>14];
218 :     blk[8*5] = iclp[(X0-X4)>>14];
219 :     blk[8*6] = iclp[(X3-X2)>>14];
220 :     blk[8*7] = iclp[(X7-X1)>>14];
221 : edgomez 677 }
222 :     #endif
223 : Isibaar 3
224 : edgomez 677 /* function pointer */
225 : Isibaar 3 idctFuncPtr idct;
226 :    
227 :     /* two dimensional inverse discrete cosine transform */
228 : edgomez 677 /*void j_rev_dct(block) */
229 :     /*short *block; */
230 : edgomez 195 void
231 :     idct_int32(short *const block)
232 : Isibaar 3 {
233 :    
234 : edgomez 677 /* idct_int32_init() must be called before the first call to this function! */
235 : Isibaar 3
236 :    
237 : edgomez 195 /*int i;
238 :     long i;
239 : Isibaar 3
240 : edgomez 195 for (i=0; i<8; i++)
241 :     idctrow(block+8*i);
242 : Isibaar 3
243 : edgomez 195 for (i=0; i<8; i++)
244 :     idctcol(block+i); */
245 :     static short *blk;
246 :     static long i;
247 :     static long X0, X1, X2, X3, X4, X5, X6, X7, X8;
248 : Isibaar 3
249 :    
250 : edgomez 677 for (i = 0; i < 8; i++) /* idct rows */
251 : Isibaar 3 {
252 : edgomez 195 blk = block + (i << 3);
253 :     if (!
254 :     ((X1 = blk[4] << 11) | (X2 = blk[6]) | (X3 = blk[2]) | (X4 =
255 :     blk[1]) |
256 :     (X5 = blk[7]) | (X6 = blk[5]) | (X7 = blk[3]))) {
257 :     blk[0] = blk[1] = blk[2] = blk[3] = blk[4] = blk[5] = blk[6] =
258 :     blk[7] = blk[0] << 3;
259 :     continue;
260 :     }
261 : Isibaar 3
262 : edgomez 677 X0 = (blk[0] << 11) + 128; /* for proper rounding in the fourth stage */
263 : Isibaar 3
264 : edgomez 677 /* first stage */
265 : edgomez 195 X8 = W7 * (X4 + X5);
266 :     X4 = X8 + (W1 - W7) * X4;
267 :     X5 = X8 - (W1 + W7) * X5;
268 :     X8 = W3 * (X6 + X7);
269 :     X6 = X8 - (W3 - W5) * X6;
270 :     X7 = X8 - (W3 + W5) * X7;
271 : Isibaar 3
272 : edgomez 677 /* second stage */
273 : edgomez 195 X8 = X0 + X1;
274 :     X0 -= X1;
275 :     X1 = W6 * (X3 + X2);
276 :     X2 = X1 - (W2 + W6) * X2;
277 :     X3 = X1 + (W2 - W6) * X3;
278 :     X1 = X4 + X6;
279 :     X4 -= X6;
280 :     X6 = X5 + X7;
281 :     X5 -= X7;
282 : Isibaar 3
283 : edgomez 677 /* third stage */
284 : edgomez 195 X7 = X8 + X3;
285 :     X8 -= X3;
286 :     X3 = X0 + X2;
287 :     X0 -= X2;
288 :     X2 = (181 * (X4 + X5) + 128) >> 8;
289 :     X4 = (181 * (X4 - X5) + 128) >> 8;
290 : Isibaar 3
291 : edgomez 677 /* fourth stage */
292 : Isibaar 3
293 : edgomez 195 blk[0] = (short) ((X7 + X1) >> 8);
294 :     blk[1] = (short) ((X3 + X2) >> 8);
295 :     blk[2] = (short) ((X0 + X4) >> 8);
296 :     blk[3] = (short) ((X8 + X6) >> 8);
297 :     blk[4] = (short) ((X8 - X6) >> 8);
298 :     blk[5] = (short) ((X0 - X4) >> 8);
299 :     blk[6] = (short) ((X3 - X2) >> 8);
300 :     blk[7] = (short) ((X7 - X1) >> 8);
301 :    
302 : edgomez 677 } /* end for ( i = 0; i < 8; ++i ) IDCT-rows */
303 : edgomez 195
304 :    
305 :    
306 : edgomez 677 for (i = 0; i < 8; i++) /* idct columns */
307 : Isibaar 3 {
308 : edgomez 195 blk = block + i;
309 : edgomez 677 /* shortcut */
310 : edgomez 195 if (!
311 :     ((X1 = (blk[8 * 4] << 8)) | (X2 = blk[8 * 6]) | (X3 =
312 :     blk[8 *
313 :     2]) | (X4 =
314 :     blk[8 *
315 :     1])
316 :     | (X5 = blk[8 * 7]) | (X6 = blk[8 * 5]) | (X7 = blk[8 * 3]))) {
317 :     blk[8 * 0] = blk[8 * 1] = blk[8 * 2] = blk[8 * 3] = blk[8 * 4] =
318 :     blk[8 * 5] = blk[8 * 6] = blk[8 * 7] =
319 :     iclp[(blk[8 * 0] + 32) >> 6];
320 :     continue;
321 :     }
322 :    
323 :     X0 = (blk[8 * 0] << 8) + 8192;
324 :    
325 : edgomez 677 /* first stage */
326 : edgomez 195 X8 = W7 * (X4 + X5) + 4;
327 :     X4 = (X8 + (W1 - W7) * X4) >> 3;
328 :     X5 = (X8 - (W1 + W7) * X5) >> 3;
329 :     X8 = W3 * (X6 + X7) + 4;
330 :     X6 = (X8 - (W3 - W5) * X6) >> 3;
331 :     X7 = (X8 - (W3 + W5) * X7) >> 3;
332 :    
333 : edgomez 677 /* second stage */
334 : edgomez 195 X8 = X0 + X1;
335 :     X0 -= X1;
336 :     X1 = W6 * (X3 + X2) + 4;
337 :     X2 = (X1 - (W2 + W6) * X2) >> 3;
338 :     X3 = (X1 + (W2 - W6) * X3) >> 3;
339 :     X1 = X4 + X6;
340 :     X4 -= X6;
341 :     X6 = X5 + X7;
342 :     X5 -= X7;
343 :    
344 : edgomez 677 /* third stage */
345 : edgomez 195 X7 = X8 + X3;
346 :     X8 -= X3;
347 :     X3 = X0 + X2;
348 :     X0 -= X2;
349 :     X2 = (181 * (X4 + X5) + 128) >> 8;
350 :     X4 = (181 * (X4 - X5) + 128) >> 8;
351 :    
352 : edgomez 677 /* fourth stage */
353 : edgomez 195 blk[8 * 0] = iclp[(X7 + X1) >> 14];
354 :     blk[8 * 1] = iclp[(X3 + X2) >> 14];
355 :     blk[8 * 2] = iclp[(X0 + X4) >> 14];
356 :     blk[8 * 3] = iclp[(X8 + X6) >> 14];
357 :     blk[8 * 4] = iclp[(X8 - X6) >> 14];
358 :     blk[8 * 5] = iclp[(X0 - X4) >> 14];
359 :     blk[8 * 6] = iclp[(X3 - X2) >> 14];
360 :     blk[8 * 7] = iclp[(X7 - X1) >> 14];
361 : Isibaar 3 }
362 :    
363 : edgomez 677 } /* end function idct_int32(block) */
364 : Isibaar 3
365 :    
366 : edgomez 677 /*void */
367 :     /*idct_int32_init() */
368 : edgomez 195 void
369 : edgomez 677 idct_int32_init(void)
370 : Isibaar 3 {
371 : edgomez 195 int i;
372 : Isibaar 3
373 : edgomez 195 iclp = iclip + 512;
374 :     for (i = -512; i < 512; i++)
375 :     iclp[i] = (i < -256) ? -256 : ((i > 255) ? 255 : i);
376 : Isibaar 3 }

No admin address has been configured
ViewVC Help
Powered by ViewVC 1.0.4