[cvs] / xvidcore / src / dct / idct.c Repository:
ViewVC logotype

Annotation of /xvidcore/src/dct/idct.c

Parent Directory Parent Directory | Revision Log Revision Log

Revision 1.6.2.1 - (view) (download)

1 : edgomez 1.6 /* idct.c, inverse fast discrete cosine transform */
2 :    
3 :     /* Copyright (C) 1996, MPEG Software Simulation Group. All Rights Reserved. */
4 :    
5 :     /*
6 :     * Disclaimer of Warranty
7 : chl 1.3 *
8 : edgomez 1.6 * These software programs are available to the user without any license fee or
9 :     * royalty on an "as is" basis. The MPEG Software Simulation Group disclaims
10 :     * any and all warranties, whether express, implied, or statuary, including any
11 :     * implied warranties or merchantability or of fitness for a particular
12 :     * purpose. In no event shall the copyright-holder be liable for any
13 :     * incidental, punitive, or consequential damages of any kind whatsoever
14 :     * arising from the use of these programs.
15 :     *
16 :     * This disclaimer of warranty extends to the user of these programs and user's
17 :     * customers, employees, agents, transferees, successors, and assigns.
18 :     *
19 :     * The MPEG Software Simulation Group does not represent or warrant that the
20 :     * programs furnished hereunder are free of infringement of any third-party
21 :     * patents.
22 :     *
23 :     * Commercial implementations of MPEG-1 and MPEG-2 video, including shareware,
24 :     * are subject to royalty fees to patent holders. Many of these patents are
25 :     * general enough such that they are unavoidable regardless of implementation
26 :     * design.
27 :     *
28 :     * MPEG2AVI
29 :     * --------
30 :     * v0.16B33 renamed the initialization function to init_idct_int32()
31 :     * v0.16B32 removed the unused idct_row() and idct_col() functions
32 :     * v0.16B3 changed var declarations to static, to enforce data align
33 :     * v0.16B22 idct_FAST() renamed to idct_int32()
34 :     * also merged idct_FAST() into a single function, to help VC++
35 :     * optimize it.
36 :     *
37 :     * v0.14 changed int to long, to avoid confusion when compiling on x86
38 :     * platform ( in VC++ "int" -> 32bits )
39 :     */
40 : Isibaar 1.1
41 :     /**********************************************************/
42 :     /* inverse two dimensional DCT, Chen-Wang algorithm */
43 :     /* (cf. IEEE ASSP-32, pp. 803-816, Aug. 1984) */
44 :     /* 32-bit integer arithmetic (8 bit coefficients) */
45 :     /* 11 mults, 29 adds per DCT */
46 :     /* sE, 18.8.91 */
47 :     /**********************************************************/
48 :     /* coefficients extended to 12 bit for IEEE1180-1990 */
49 :     /* compliance sE, 2.1.94 */
50 :     /**********************************************************/
51 :    
52 :     /* this code assumes >> to be a two's-complement arithmetic */
53 :     /* right shift: (-2)>>1 == -1 , (-3)>>1 == -2 */
54 :    
55 :     #include "idct.h"
56 :    
57 : edgomez 1.2 #define W1 2841 /* 2048*sqrt(2)*cos(1*pi/16) */
58 :     #define W2 2676 /* 2048*sqrt(2)*cos(2*pi/16) */
59 :     #define W3 2408 /* 2048*sqrt(2)*cos(3*pi/16) */
60 :     #define W5 1609 /* 2048*sqrt(2)*cos(5*pi/16) */
61 :     #define W6 1108 /* 2048*sqrt(2)*cos(6*pi/16) */
62 :     #define W7 565 /* 2048*sqrt(2)*cos(7*pi/16) */
63 : Isibaar 1.1
64 :     /* private data */
65 : edgomez 1.2 static short iclip[1024]; /* clipping table */
66 : Isibaar 1.1 static short *iclp;
67 :    
68 :     /* private prototypes */
69 :    
70 :     /* row (horizontal) IDCT
71 :     *
72 :     * 7 pi 1
73 :     * dst[k] = sum c[l] * src[l] * cos( -- * ( k + - ) * l )
74 :     * l=0 8 2
75 :     *
76 :     * where: c[0] = 128
77 :     * c[1..7] = 128*sqrt(2)
78 :     */
79 :    
80 : edgomez 1.6.2.1 #if 0
81 : Isibaar 1.1 static void idctrow(blk)
82 :     short *blk;
83 :     {
84 :     int X0, X1, X2, X3, X4, X5, X6, X7, X8;
85 :    
86 : edgomez 1.6.2.1 /* shortcut */
87 : Isibaar 1.1 if (!((X1 = blk[4]<<11) | (X2 = blk[6]) | (X3 = blk[2]) |
88 :     (X4 = blk[1]) | (X5 = blk[7]) | (X6 = blk[5]) | (X7 = blk[3])))
89 :     {
90 :     blk[0]=blk[1]=blk[2]=blk[3]=blk[4]=blk[5]=blk[6]=blk[7]=blk[0]<<3;
91 :     return;
92 :     }
93 :    
94 : edgomez 1.6.2.1 X0 = (blk[0]<<11) + 128; /* for proper rounding in the fourth stage */
95 : Isibaar 1.1
96 : edgomez 1.6.2.1 /* first stage */
97 : Isibaar 1.1 X8 = W7*(X4+X5);
98 :     X4 = X8 + (W1-W7)*X4;
99 :     X5 = X8 - (W1+W7)*X5;
100 :     X8 = W3*(X6+X7);
101 :     X6 = X8 - (W3-W5)*X6;
102 :     X7 = X8 - (W3+W5)*X7;
103 :    
104 : edgomez 1.6.2.1 /* second stage */
105 : Isibaar 1.1 X8 = X0 + X1;
106 :     X0 -= X1;
107 :     X1 = W6*(X3+X2);
108 :     X2 = X1 - (W2+W6)*X2;
109 :     X3 = X1 + (W2-W6)*X3;
110 :     X1 = X4 + X6;
111 :     X4 -= X6;
112 :     X6 = X5 + X7;
113 :     X5 -= X7;
114 :    
115 : edgomez 1.6.2.1 /* third stage */
116 : Isibaar 1.1 X7 = X8 + X3;
117 :     X8 -= X3;
118 :     X3 = X0 + X2;
119 :     X0 -= X2;
120 :     X2 = (181*(X4+X5)+128)>>8;
121 :     X4 = (181*(X4-X5)+128)>>8;
122 :    
123 : edgomez 1.6.2.1 /* fourth stage */
124 : Isibaar 1.1 blk[0] = (X7+X1)>>8;
125 :     blk[1] = (X3+X2)>>8;
126 :     blk[2] = (X0+X4)>>8;
127 :     blk[3] = (X8+X6)>>8;
128 :     blk[4] = (X8-X6)>>8;
129 :     blk[5] = (X0-X4)>>8;
130 :     blk[6] = (X3-X2)>>8;
131 :     blk[7] = (X7-X1)>>8;
132 : edgomez 1.6.2.1 }
133 :     #endif
134 : Isibaar 1.1
135 :     /* column (vertical) IDCT
136 :     *
137 :     * 7 pi 1
138 :     * dst[8*k] = sum c[l] * src[8*l] * cos( -- * ( k + - ) * l )
139 :     * l=0 8 2
140 :     *
141 :     * where: c[0] = 1/1024
142 :     * c[1..7] = (1/1024)*sqrt(2)
143 :     */
144 : edgomez 1.6.2.1
145 :     #if 0
146 : Isibaar 1.1 static void idctcol(blk)
147 :     short *blk;
148 :     {
149 :     int X0, X1, X2, X3, X4, X5, X6, X7, X8;
150 :    
151 : edgomez 1.6.2.1 /* shortcut */
152 : Isibaar 1.1 if (!((X1 = (blk[8*4]<<8)) | (X2 = blk[8*6]) | (X3 = blk[8*2]) |
153 :     (X4 = blk[8*1]) | (X5 = blk[8*7]) | (X6 = blk[8*5]) | (X7 = blk[8*3])))
154 :     {
155 :     blk[8*0]=blk[8*1]=blk[8*2]=blk[8*3]=blk[8*4]=blk[8*5]=blk[8*6]=blk[8*7]=
156 :     iclp[(blk[8*0]+32)>>6];
157 :     return;
158 :     }
159 :    
160 :     X0 = (blk[8*0]<<8) + 8192;
161 :    
162 : edgomez 1.6.2.1 /* first stage */
163 : Isibaar 1.1 X8 = W7*(X4+X5) + 4;
164 :     X4 = (X8+(W1-W7)*X4)>>3;
165 :     X5 = (X8-(W1+W7)*X5)>>3;
166 :     X8 = W3*(X6+X7) + 4;
167 :     X6 = (X8-(W3-W5)*X6)>>3;
168 :     X7 = (X8-(W3+W5)*X7)>>3;
169 :    
170 : edgomez 1.6.2.1 /* second stage */
171 : Isibaar 1.1 X8 = X0 + X1;
172 :     X0 -= X1;
173 :     X1 = W6*(X3+X2) + 4;
174 :     X2 = (X1-(W2+W6)*X2)>>3;
175 :     X3 = (X1+(W2-W6)*X3)>>3;
176 :     X1 = X4 + X6;
177 :     X4 -= X6;
178 :     X6 = X5 + X7;
179 :     X5 -= X7;
180 :    
181 : edgomez 1.6.2.1 /* third stage */
182 : Isibaar 1.1 X7 = X8 + X3;
183 :     X8 -= X3;
184 :     X3 = X0 + X2;
185 :     X0 -= X2;
186 :     X2 = (181*(X4+X5)+128)>>8;
187 :     X4 = (181*(X4-X5)+128)>>8;
188 :    
189 : edgomez 1.6.2.1 /* fourth stage */
190 : Isibaar 1.1 blk[8*0] = iclp[(X7+X1)>>14];
191 :     blk[8*1] = iclp[(X3+X2)>>14];
192 :     blk[8*2] = iclp[(X0+X4)>>14];
193 :     blk[8*3] = iclp[(X8+X6)>>14];
194 :     blk[8*4] = iclp[(X8-X6)>>14];
195 :     blk[8*5] = iclp[(X0-X4)>>14];
196 :     blk[8*6] = iclp[(X3-X2)>>14];
197 :     blk[8*7] = iclp[(X7-X1)>>14];
198 : edgomez 1.6.2.1 }
199 :     #endif
200 : Isibaar 1.1
201 : edgomez 1.6.2.1 /* function pointer */
202 : Isibaar 1.1 idctFuncPtr idct;
203 :    
204 :     /* two dimensional inverse discrete cosine transform */
205 : edgomez 1.2 void
206 :     idct_int32(short *const block)
207 : Isibaar 1.1 {
208 :    
209 : edgomez 1.6.2.1 /*
210 :     * idct_int32_init() must be called before the first call to this
211 :     * function!
212 :     */
213 :    
214 :    
215 :     #if 0
216 :     int i;
217 :     long i;
218 :    
219 :     for (i=0; i<8; i++)
220 :     idctrow(block+8*i);
221 :    
222 :     for (i=0; i<8; i++)
223 :     idctcol(block+i);
224 :     #endif
225 : Isibaar 1.1
226 : edgomez 1.2 static short *blk;
227 :     static long i;
228 :     static long X0, X1, X2, X3, X4, X5, X6, X7, X8;
229 : Isibaar 1.1
230 :    
231 : edgomez 1.6.2.1 for (i = 0; i < 8; i++) /* idct rows */
232 : Isibaar 1.1 {
233 : edgomez 1.2 blk = block + (i << 3);
234 :     if (!
235 :     ((X1 = blk[4] << 11) | (X2 = blk[6]) | (X3 = blk[2]) | (X4 =
236 :     blk[1]) |
237 :     (X5 = blk[7]) | (X6 = blk[5]) | (X7 = blk[3]))) {
238 :     blk[0] = blk[1] = blk[2] = blk[3] = blk[4] = blk[5] = blk[6] =
239 :     blk[7] = blk[0] << 3;
240 :     continue;
241 :     }
242 :    
243 : edgomez 1.6.2.1 X0 = (blk[0] << 11) + 128; /* for proper rounding in the fourth stage */
244 : edgomez 1.2
245 : edgomez 1.6.2.1 /* first stage */
246 : edgomez 1.2 X8 = W7 * (X4 + X5);
247 :     X4 = X8 + (W1 - W7) * X4;
248 :     X5 = X8 - (W1 + W7) * X5;
249 :     X8 = W3 * (X6 + X7);
250 :     X6 = X8 - (W3 - W5) * X6;
251 :     X7 = X8 - (W3 + W5) * X7;
252 :    
253 : edgomez 1.6.2.1 /* second stage */
254 : edgomez 1.2 X8 = X0 + X1;
255 :     X0 -= X1;
256 :     X1 = W6 * (X3 + X2);
257 :     X2 = X1 - (W2 + W6) * X2;
258 :     X3 = X1 + (W2 - W6) * X3;
259 :     X1 = X4 + X6;
260 :     X4 -= X6;
261 :     X6 = X5 + X7;
262 :     X5 -= X7;
263 :    
264 : edgomez 1.6.2.1 /* third stage */
265 : edgomez 1.2 X7 = X8 + X3;
266 :     X8 -= X3;
267 :     X3 = X0 + X2;
268 :     X0 -= X2;
269 :     X2 = (181 * (X4 + X5) + 128) >> 8;
270 :     X4 = (181 * (X4 - X5) + 128) >> 8;
271 :    
272 : edgomez 1.6.2.1 /* fourth stage */
273 : edgomez 1.2
274 :     blk[0] = (short) ((X7 + X1) >> 8);
275 :     blk[1] = (short) ((X3 + X2) >> 8);
276 :     blk[2] = (short) ((X0 + X4) >> 8);
277 :     blk[3] = (short) ((X8 + X6) >> 8);
278 :     blk[4] = (short) ((X8 - X6) >> 8);
279 :     blk[5] = (short) ((X0 - X4) >> 8);
280 :     blk[6] = (short) ((X3 - X2) >> 8);
281 :     blk[7] = (short) ((X7 - X1) >> 8);
282 : Isibaar 1.1
283 : edgomez 1.6.2.1 } /* end for ( i = 0; i < 8; ++i ) IDCT-rows */
284 : Isibaar 1.1
285 :    
286 :    
287 : edgomez 1.6.2.1 for (i = 0; i < 8; i++) /* idct columns */
288 : Isibaar 1.1 {
289 : edgomez 1.2 blk = block + i;
290 : edgomez 1.6.2.1 /* shortcut */
291 : edgomez 1.2 if (!
292 :     ((X1 = (blk[8 * 4] << 8)) | (X2 = blk[8 * 6]) | (X3 =
293 :     blk[8 *
294 :     2]) | (X4 =
295 :     blk[8 *
296 :     1])
297 :     | (X5 = blk[8 * 7]) | (X6 = blk[8 * 5]) | (X7 = blk[8 * 3]))) {
298 :     blk[8 * 0] = blk[8 * 1] = blk[8 * 2] = blk[8 * 3] = blk[8 * 4] =
299 :     blk[8 * 5] = blk[8 * 6] = blk[8 * 7] =
300 :     iclp[(blk[8 * 0] + 32) >> 6];
301 :     continue;
302 :     }
303 :    
304 :     X0 = (blk[8 * 0] << 8) + 8192;
305 :    
306 : edgomez 1.6.2.1 /* first stage */
307 : edgomez 1.2 X8 = W7 * (X4 + X5) + 4;
308 :     X4 = (X8 + (W1 - W7) * X4) >> 3;
309 :     X5 = (X8 - (W1 + W7) * X5) >> 3;
310 :     X8 = W3 * (X6 + X7) + 4;
311 :     X6 = (X8 - (W3 - W5) * X6) >> 3;
312 :     X7 = (X8 - (W3 + W5) * X7) >> 3;
313 :    
314 : edgomez 1.6.2.1 /* second stage */
315 : edgomez 1.2 X8 = X0 + X1;
316 :     X0 -= X1;
317 :     X1 = W6 * (X3 + X2) + 4;
318 :     X2 = (X1 - (W2 + W6) * X2) >> 3;
319 :     X3 = (X1 + (W2 - W6) * X3) >> 3;
320 :     X1 = X4 + X6;
321 :     X4 -= X6;
322 :     X6 = X5 + X7;
323 :     X5 -= X7;
324 :    
325 : edgomez 1.6.2.1 /* third stage */
326 : edgomez 1.2 X7 = X8 + X3;
327 :     X8 -= X3;
328 :     X3 = X0 + X2;
329 :     X0 -= X2;
330 :     X2 = (181 * (X4 + X5) + 128) >> 8;
331 :     X4 = (181 * (X4 - X5) + 128) >> 8;
332 :    
333 : edgomez 1.6.2.1 /* fourth stage */
334 : edgomez 1.2 blk[8 * 0] = iclp[(X7 + X1) >> 14];
335 :     blk[8 * 1] = iclp[(X3 + X2) >> 14];
336 :     blk[8 * 2] = iclp[(X0 + X4) >> 14];
337 :     blk[8 * 3] = iclp[(X8 + X6) >> 14];
338 :     blk[8 * 4] = iclp[(X8 - X6) >> 14];
339 :     blk[8 * 5] = iclp[(X0 - X4) >> 14];
340 :     blk[8 * 6] = iclp[(X3 - X2) >> 14];
341 :     blk[8 * 7] = iclp[(X7 - X1) >> 14];
342 : Isibaar 1.1 }
343 :    
344 : edgomez 1.6.2.1 } /* end function idct_int32(block) */
345 : Isibaar 1.1
346 :    
347 : edgomez 1.2 void
348 : edgomez 1.6 idct_int32_init()
349 : Isibaar 1.1 {
350 : edgomez 1.2 int i;
351 : Isibaar 1.1
352 : edgomez 1.2 iclp = iclip + 512;
353 :     for (i = -512; i < 512; i++)
354 :     iclp[i] = (i < -256) ? -256 : ((i > 255) ? 255 : i);
355 : Isibaar 1.1 }
No admin address has been configured
 ViewVC Help
Powered by ViewVC 1.0.4