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

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

Parent Directory Parent Directory | Revision Log Revision Log


Revision 851 - (view) (download)

1 : edgomez 851 /* 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 : Isibaar 3 *
8 : edgomez 851 * 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 : Isibaar 3 *
16 : edgomez 851 * This disclaimer of warranty extends to the user of these programs and user's
17 :     * customers, employees, agents, transferees, successors, and assigns.
18 : Isibaar 3 *
19 : edgomez 851 * 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 : Isibaar 3 *
23 : edgomez 851 * 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 : Isibaar 3 *
28 : edgomez 851 * 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 3
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 : edgomez 851 //#include <windows.h>
56 : Isibaar 3 #include "idct.h"
57 :    
58 : edgomez 195 #define W1 2841 /* 2048*sqrt(2)*cos(1*pi/16) */
59 :     #define W2 2676 /* 2048*sqrt(2)*cos(2*pi/16) */
60 :     #define W3 2408 /* 2048*sqrt(2)*cos(3*pi/16) */
61 :     #define W5 1609 /* 2048*sqrt(2)*cos(5*pi/16) */
62 :     #define W6 1108 /* 2048*sqrt(2)*cos(6*pi/16) */
63 :     #define W7 565 /* 2048*sqrt(2)*cos(7*pi/16) */
64 : Isibaar 3
65 :    
66 :     /* global declarations */
67 : edgomez 851 //void init_idct_int32 (void);
68 :     //void idct_int32 (short *block);
69 : Isibaar 3
70 :     /* private data */
71 : edgomez 195 static short iclip[1024]; /* clipping table */
72 : Isibaar 3 static short *iclp;
73 :    
74 :     /* private prototypes */
75 : edgomez 851 //static void idctrow _ANSI_ARGS_((short *blk));
76 :     //static void idctcol _ANSI_ARGS_((short *blk));
77 : Isibaar 3
78 :     /* row (horizontal) IDCT
79 :     *
80 :     * 7 pi 1
81 :     * dst[k] = sum c[l] * src[l] * cos( -- * ( k + - ) * l )
82 :     * l=0 8 2
83 :     *
84 :     * where: c[0] = 128
85 :     * c[1..7] = 128*sqrt(2)
86 :     */
87 :    
88 : edgomez 851 /*
89 : Isibaar 3 static void idctrow(blk)
90 :     short *blk;
91 :     {
92 :     int X0, X1, X2, X3, X4, X5, X6, X7, X8;
93 :    
94 : edgomez 851 // shortcut
95 : Isibaar 3 if (!((X1 = blk[4]<<11) | (X2 = blk[6]) | (X3 = blk[2]) |
96 :     (X4 = blk[1]) | (X5 = blk[7]) | (X6 = blk[5]) | (X7 = blk[3])))
97 :     {
98 :     blk[0]=blk[1]=blk[2]=blk[3]=blk[4]=blk[5]=blk[6]=blk[7]=blk[0]<<3;
99 :     return;
100 :     }
101 :    
102 : edgomez 851 X0 = (blk[0]<<11) + 128; // for proper rounding in the fourth stage
103 : Isibaar 3
104 : edgomez 851 // first stage
105 : Isibaar 3 X8 = W7*(X4+X5);
106 :     X4 = X8 + (W1-W7)*X4;
107 :     X5 = X8 - (W1+W7)*X5;
108 :     X8 = W3*(X6+X7);
109 :     X6 = X8 - (W3-W5)*X6;
110 :     X7 = X8 - (W3+W5)*X7;
111 :    
112 : edgomez 851 // second stage
113 : Isibaar 3 X8 = X0 + X1;
114 :     X0 -= X1;
115 :     X1 = W6*(X3+X2);
116 :     X2 = X1 - (W2+W6)*X2;
117 :     X3 = X1 + (W2-W6)*X3;
118 :     X1 = X4 + X6;
119 :     X4 -= X6;
120 :     X6 = X5 + X7;
121 :     X5 -= X7;
122 :    
123 : edgomez 851 // third stage
124 : Isibaar 3 X7 = X8 + X3;
125 :     X8 -= X3;
126 :     X3 = X0 + X2;
127 :     X0 -= X2;
128 :     X2 = (181*(X4+X5)+128)>>8;
129 :     X4 = (181*(X4-X5)+128)>>8;
130 :    
131 : edgomez 851 // fourth stage
132 : Isibaar 3 blk[0] = (X7+X1)>>8;
133 :     blk[1] = (X3+X2)>>8;
134 :     blk[2] = (X0+X4)>>8;
135 :     blk[3] = (X8+X6)>>8;
136 :     blk[4] = (X8-X6)>>8;
137 :     blk[5] = (X0-X4)>>8;
138 :     blk[6] = (X3-X2)>>8;
139 :     blk[7] = (X7-X1)>>8;
140 : edgomez 851 }*/
141 : Isibaar 3
142 :     /* column (vertical) IDCT
143 :     *
144 :     * 7 pi 1
145 :     * dst[8*k] = sum c[l] * src[8*l] * cos( -- * ( k + - ) * l )
146 :     * l=0 8 2
147 :     *
148 :     * where: c[0] = 1/1024
149 :     * c[1..7] = (1/1024)*sqrt(2)
150 :     */
151 : edgomez 851 /*
152 : Isibaar 3 static void idctcol(blk)
153 :     short *blk;
154 :     {
155 :     int X0, X1, X2, X3, X4, X5, X6, X7, X8;
156 :    
157 : edgomez 851 // shortcut
158 : Isibaar 3 if (!((X1 = (blk[8*4]<<8)) | (X2 = blk[8*6]) | (X3 = blk[8*2]) |
159 :     (X4 = blk[8*1]) | (X5 = blk[8*7]) | (X6 = blk[8*5]) | (X7 = blk[8*3])))
160 :     {
161 :     blk[8*0]=blk[8*1]=blk[8*2]=blk[8*3]=blk[8*4]=blk[8*5]=blk[8*6]=blk[8*7]=
162 :     iclp[(blk[8*0]+32)>>6];
163 :     return;
164 :     }
165 :    
166 :     X0 = (blk[8*0]<<8) + 8192;
167 :    
168 : edgomez 851 // first stage
169 : Isibaar 3 X8 = W7*(X4+X5) + 4;
170 :     X4 = (X8+(W1-W7)*X4)>>3;
171 :     X5 = (X8-(W1+W7)*X5)>>3;
172 :     X8 = W3*(X6+X7) + 4;
173 :     X6 = (X8-(W3-W5)*X6)>>3;
174 :     X7 = (X8-(W3+W5)*X7)>>3;
175 :    
176 : edgomez 851 // second stage
177 : Isibaar 3 X8 = X0 + X1;
178 :     X0 -= X1;
179 :     X1 = W6*(X3+X2) + 4;
180 :     X2 = (X1-(W2+W6)*X2)>>3;
181 :     X3 = (X1+(W2-W6)*X3)>>3;
182 :     X1 = X4 + X6;
183 :     X4 -= X6;
184 :     X6 = X5 + X7;
185 :     X5 -= X7;
186 :    
187 : edgomez 851 // third stage
188 : Isibaar 3 X7 = X8 + X3;
189 :     X8 -= X3;
190 :     X3 = X0 + X2;
191 :     X0 -= X2;
192 :     X2 = (181*(X4+X5)+128)>>8;
193 :     X4 = (181*(X4-X5)+128)>>8;
194 :    
195 : edgomez 851 // fourth stage
196 : Isibaar 3 blk[8*0] = iclp[(X7+X1)>>14];
197 :     blk[8*1] = iclp[(X3+X2)>>14];
198 :     blk[8*2] = iclp[(X0+X4)>>14];
199 :     blk[8*3] = iclp[(X8+X6)>>14];
200 :     blk[8*4] = iclp[(X8-X6)>>14];
201 :     blk[8*5] = iclp[(X0-X4)>>14];
202 :     blk[8*6] = iclp[(X3-X2)>>14];
203 :     blk[8*7] = iclp[(X7-X1)>>14];
204 : edgomez 851 }*/
205 : Isibaar 3
206 : edgomez 851 // function pointer
207 : Isibaar 3 idctFuncPtr idct;
208 :    
209 :     /* two dimensional inverse discrete cosine transform */
210 : edgomez 851 //void j_rev_dct(block)
211 :     //short *block;
212 : edgomez 195 void
213 :     idct_int32(short *const block)
214 : Isibaar 3 {
215 :    
216 : edgomez 851 // idct_int32_init() must be called before the first call to this function!
217 : Isibaar 3
218 :    
219 : edgomez 195 /*int i;
220 :     long i;
221 : Isibaar 3
222 : edgomez 195 for (i=0; i<8; i++)
223 :     idctrow(block+8*i);
224 : Isibaar 3
225 : edgomez 195 for (i=0; i<8; i++)
226 :     idctcol(block+i); */
227 :     static short *blk;
228 :     static long i;
229 :     static long X0, X1, X2, X3, X4, X5, X6, X7, X8;
230 : Isibaar 3
231 :    
232 : edgomez 851 for (i = 0; i < 8; i++) // idct rows
233 : Isibaar 3 {
234 : edgomez 195 blk = block + (i << 3);
235 :     if (!
236 :     ((X1 = blk[4] << 11) | (X2 = blk[6]) | (X3 = blk[2]) | (X4 =
237 :     blk[1]) |
238 :     (X5 = blk[7]) | (X6 = blk[5]) | (X7 = blk[3]))) {
239 :     blk[0] = blk[1] = blk[2] = blk[3] = blk[4] = blk[5] = blk[6] =
240 :     blk[7] = blk[0] << 3;
241 :     continue;
242 :     }
243 : Isibaar 3
244 : edgomez 851 X0 = (blk[0] << 11) + 128; // for proper rounding in the fourth stage
245 : Isibaar 3
246 : edgomez 851 // first stage
247 : edgomez 195 X8 = W7 * (X4 + X5);
248 :     X4 = X8 + (W1 - W7) * X4;
249 :     X5 = X8 - (W1 + W7) * X5;
250 :     X8 = W3 * (X6 + X7);
251 :     X6 = X8 - (W3 - W5) * X6;
252 :     X7 = X8 - (W3 + W5) * X7;
253 : Isibaar 3
254 : edgomez 851 // second stage
255 : edgomez 195 X8 = X0 + X1;
256 :     X0 -= X1;
257 :     X1 = W6 * (X3 + X2);
258 :     X2 = X1 - (W2 + W6) * X2;
259 :     X3 = X1 + (W2 - W6) * X3;
260 :     X1 = X4 + X6;
261 :     X4 -= X6;
262 :     X6 = X5 + X7;
263 :     X5 -= X7;
264 : Isibaar 3
265 : edgomez 851 // third stage
266 : edgomez 195 X7 = X8 + X3;
267 :     X8 -= X3;
268 :     X3 = X0 + X2;
269 :     X0 -= X2;
270 :     X2 = (181 * (X4 + X5) + 128) >> 8;
271 :     X4 = (181 * (X4 - X5) + 128) >> 8;
272 : Isibaar 3
273 : edgomez 851 // fourth stage
274 : Isibaar 3
275 : edgomez 195 blk[0] = (short) ((X7 + X1) >> 8);
276 :     blk[1] = (short) ((X3 + X2) >> 8);
277 :     blk[2] = (short) ((X0 + X4) >> 8);
278 :     blk[3] = (short) ((X8 + X6) >> 8);
279 :     blk[4] = (short) ((X8 - X6) >> 8);
280 :     blk[5] = (short) ((X0 - X4) >> 8);
281 :     blk[6] = (short) ((X3 - X2) >> 8);
282 :     blk[7] = (short) ((X7 - X1) >> 8);
283 :    
284 : edgomez 851 } // end for ( i = 0; i < 8; ++i ) IDCT-rows
285 : edgomez 195
286 :    
287 :    
288 : edgomez 851 for (i = 0; i < 8; i++) // idct columns
289 : Isibaar 3 {
290 : edgomez 195 blk = block + i;
291 : edgomez 851 // shortcut
292 : edgomez 195 if (!
293 :     ((X1 = (blk[8 * 4] << 8)) | (X2 = blk[8 * 6]) | (X3 =
294 :     blk[8 *
295 :     2]) | (X4 =
296 :     blk[8 *
297 :     1])
298 :     | (X5 = blk[8 * 7]) | (X6 = blk[8 * 5]) | (X7 = blk[8 * 3]))) {
299 :     blk[8 * 0] = blk[8 * 1] = blk[8 * 2] = blk[8 * 3] = blk[8 * 4] =
300 :     blk[8 * 5] = blk[8 * 6] = blk[8 * 7] =
301 :     iclp[(blk[8 * 0] + 32) >> 6];
302 :     continue;
303 :     }
304 :    
305 :     X0 = (blk[8 * 0] << 8) + 8192;
306 :    
307 : edgomez 851 // first stage
308 : edgomez 195 X8 = W7 * (X4 + X5) + 4;
309 :     X4 = (X8 + (W1 - W7) * X4) >> 3;
310 :     X5 = (X8 - (W1 + W7) * X5) >> 3;
311 :     X8 = W3 * (X6 + X7) + 4;
312 :     X6 = (X8 - (W3 - W5) * X6) >> 3;
313 :     X7 = (X8 - (W3 + W5) * X7) >> 3;
314 :    
315 : edgomez 851 // second stage
316 : edgomez 195 X8 = X0 + X1;
317 :     X0 -= X1;
318 :     X1 = W6 * (X3 + X2) + 4;
319 :     X2 = (X1 - (W2 + W6) * X2) >> 3;
320 :     X3 = (X1 + (W2 - W6) * X3) >> 3;
321 :     X1 = X4 + X6;
322 :     X4 -= X6;
323 :     X6 = X5 + X7;
324 :     X5 -= X7;
325 :    
326 : edgomez 851 // third stage
327 : edgomez 195 X7 = X8 + X3;
328 :     X8 -= X3;
329 :     X3 = X0 + X2;
330 :     X0 -= X2;
331 :     X2 = (181 * (X4 + X5) + 128) >> 8;
332 :     X4 = (181 * (X4 - X5) + 128) >> 8;
333 :    
334 : edgomez 851 // fourth stage
335 : edgomez 195 blk[8 * 0] = iclp[(X7 + X1) >> 14];
336 :     blk[8 * 1] = iclp[(X3 + X2) >> 14];
337 :     blk[8 * 2] = iclp[(X0 + X4) >> 14];
338 :     blk[8 * 3] = iclp[(X8 + X6) >> 14];
339 :     blk[8 * 4] = iclp[(X8 - X6) >> 14];
340 :     blk[8 * 5] = iclp[(X0 - X4) >> 14];
341 :     blk[8 * 6] = iclp[(X3 - X2) >> 14];
342 :     blk[8 * 7] = iclp[(X7 - X1) >> 14];
343 : Isibaar 3 }
344 :    
345 : edgomez 851 } // end function idct_int32(block)
346 : Isibaar 3
347 :    
348 : edgomez 851 //void
349 :     //idct_int32_init()
350 : edgomez 195 void
351 : edgomez 851 idct_int32_init()
352 : Isibaar 3 {
353 : edgomez 195 int i;
354 : Isibaar 3
355 : edgomez 195 iclp = iclip + 512;
356 :     for (i = -512; i < 512; i++)
357 :     iclp[i] = (i < -256) ? -256 : ((i > 255) ? 255 : i);
358 : Isibaar 3 }

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