Parent Directory | Revision Log
Revision 1424 - (view) (download)
1 : | edgomez | 1382 | /***************************************************************************** |
2 : | edgomez | 195 | * |
3 : | edgomez | 1382 | * XVID MPEG-4 VIDEO CODEC |
4 : | * - Sum Of Absolute Difference related code - | ||
5 : | edgomez | 195 | * |
6 : | edgomez | 1382 | * Copyright(C) 2001-2003 Peter Ross <pross@xvid.org> |
7 : | edgomez | 195 | * |
8 : | edgomez | 1382 | * This program is free software ; you can redistribute it and/or modify |
9 : | * it under the terms of the GNU General Public License as published by | ||
10 : | * the Free Software Foundation ; either version 2 of the License, or | ||
11 : | * (at your option) any later version. | ||
12 : | edgomez | 195 | * |
13 : | edgomez | 1382 | * 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 : | edgomez | 195 | * |
18 : | edgomez | 1382 | * You should have received a copy of the GNU General Public License |
19 : | * along with this program ; if not, write to the Free Software | ||
20 : | * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA | ||
21 : | edgomez | 195 | * |
22 : | edgomez | 1424 | * $Id: sad.c,v 1.16 2004-04-12 15:49:56 edgomez Exp $ |
23 : | edgomez | 195 | * |
24 : | edgomez | 1382 | ****************************************************************************/ |
25 : | edgomez | 195 | |
26 : | #include "../portab.h" | ||
27 : | edgomez | 851 | #include "../global.h" |
28 : | edgomez | 195 | #include "sad.h" |
29 : | |||
30 : | edgomez | 1382 | #include <stdlib.h> |
31 : | |||
32 : | edgomez | 195 | sad16FuncPtr sad16; |
33 : | sad8FuncPtr sad8; | ||
34 : | sad16biFuncPtr sad16bi; | ||
35 : | edgomez | 1382 | sad8biFuncPtr sad8bi; /* not really sad16, but no difference in prototype */ |
36 : | edgomez | 195 | dev16FuncPtr dev16; |
37 : | edgomez | 851 | sad16vFuncPtr sad16v; |
38 : | edgomez | 1382 | sse8Func_16bitPtr sse8_16bit; |
39 : | edgomez | 1424 | sse8Func_8bitPtr sse8_8bit; |
40 : | edgomez | 195 | |
41 : | sadInitFuncPtr sadInit; | ||
42 : | |||
43 : | |||
44 : | uint32_t | ||
45 : | sad16_c(const uint8_t * const cur, | ||
46 : | const uint8_t * const ref, | ||
47 : | const uint32_t stride, | ||
48 : | const uint32_t best_sad) | ||
49 : | { | ||
50 : | |||
51 : | uint32_t sad = 0; | ||
52 : | edgomez | 851 | uint32_t j; |
53 : | edgomez | 195 | uint8_t const *ptr_cur = cur; |
54 : | uint8_t const *ptr_ref = ref; | ||
55 : | |||
56 : | for (j = 0; j < 16; j++) { | ||
57 : | edgomez | 1382 | sad += abs(ptr_cur[0] - ptr_ref[0]); |
58 : | sad += abs(ptr_cur[1] - ptr_ref[1]); | ||
59 : | sad += abs(ptr_cur[2] - ptr_ref[2]); | ||
60 : | sad += abs(ptr_cur[3] - ptr_ref[3]); | ||
61 : | sad += abs(ptr_cur[4] - ptr_ref[4]); | ||
62 : | sad += abs(ptr_cur[5] - ptr_ref[5]); | ||
63 : | sad += abs(ptr_cur[6] - ptr_ref[6]); | ||
64 : | sad += abs(ptr_cur[7] - ptr_ref[7]); | ||
65 : | sad += abs(ptr_cur[8] - ptr_ref[8]); | ||
66 : | sad += abs(ptr_cur[9] - ptr_ref[9]); | ||
67 : | sad += abs(ptr_cur[10] - ptr_ref[10]); | ||
68 : | sad += abs(ptr_cur[11] - ptr_ref[11]); | ||
69 : | sad += abs(ptr_cur[12] - ptr_ref[12]); | ||
70 : | sad += abs(ptr_cur[13] - ptr_ref[13]); | ||
71 : | sad += abs(ptr_cur[14] - ptr_ref[14]); | ||
72 : | sad += abs(ptr_cur[15] - ptr_ref[15]); | ||
73 : | edgomez | 195 | |
74 : | edgomez | 1382 | if (sad >= best_sad) |
75 : | edgomez | 195 | return sad; |
76 : | |||
77 : | edgomez | 851 | ptr_cur += stride; |
78 : | ptr_ref += stride; | ||
79 : | edgomez | 195 | |
80 : | } | ||
81 : | |||
82 : | return sad; | ||
83 : | |||
84 : | } | ||
85 : | |||
86 : | uint32_t | ||
87 : | sad16bi_c(const uint8_t * const cur, | ||
88 : | const uint8_t * const ref1, | ||
89 : | const uint8_t * const ref2, | ||
90 : | const uint32_t stride) | ||
91 : | { | ||
92 : | |||
93 : | uint32_t sad = 0; | ||
94 : | uint32_t i, j; | ||
95 : | uint8_t const *ptr_cur = cur; | ||
96 : | uint8_t const *ptr_ref1 = ref1; | ||
97 : | uint8_t const *ptr_ref2 = ref2; | ||
98 : | |||
99 : | for (j = 0; j < 16; j++) { | ||
100 : | |||
101 : | for (i = 0; i < 16; i++) { | ||
102 : | int pixel = (ptr_ref1[i] + ptr_ref2[i] + 1) / 2; | ||
103 : | edgomez | 1382 | sad += abs(ptr_cur[i] - pixel); |
104 : | edgomez | 195 | } |
105 : | |||
106 : | ptr_cur += stride; | ||
107 : | ptr_ref1 += stride; | ||
108 : | ptr_ref2 += stride; | ||
109 : | |||
110 : | } | ||
111 : | |||
112 : | return sad; | ||
113 : | |||
114 : | } | ||
115 : | |||
116 : | uint32_t | ||
117 : | sad8bi_c(const uint8_t * const cur, | ||
118 : | const uint8_t * const ref1, | ||
119 : | const uint8_t * const ref2, | ||
120 : | const uint32_t stride) | ||
121 : | { | ||
122 : | |||
123 : | uint32_t sad = 0; | ||
124 : | chl | 326 | uint32_t i, j; |
125 : | edgomez | 195 | uint8_t const *ptr_cur = cur; |
126 : | chl | 326 | uint8_t const *ptr_ref1 = ref1; |
127 : | uint8_t const *ptr_ref2 = ref2; | ||
128 : | |||
129 : | for (j = 0; j < 8; j++) { | ||
130 : | |||
131 : | for (i = 0; i < 8; i++) { | ||
132 : | int pixel = (ptr_ref1[i] + ptr_ref2[i] + 1) / 2; | ||
133 : | edgomez | 1382 | sad += abs(ptr_cur[i] - pixel); |
134 : | chl | 326 | } |
135 : | |||
136 : | ptr_cur += stride; | ||
137 : | ptr_ref1 += stride; | ||
138 : | ptr_ref2 += stride; | ||
139 : | |||
140 : | } | ||
141 : | |||
142 : | return sad; | ||
143 : | |||
144 : | } | ||
145 : | |||
146 : | |||
147 : | |||
148 : | uint32_t | ||
149 : | sad8_c(const uint8_t * const cur, | ||
150 : | const uint8_t * const ref, | ||
151 : | const uint32_t stride) | ||
152 : | { | ||
153 : | uint32_t sad = 0; | ||
154 : | edgomez | 851 | uint32_t j; |
155 : | chl | 326 | uint8_t const *ptr_cur = cur; |
156 : | edgomez | 195 | uint8_t const *ptr_ref = ref; |
157 : | |||
158 : | for (j = 0; j < 8; j++) { | ||
159 : | |||
160 : | edgomez | 1382 | sad += abs(ptr_cur[0] - ptr_ref[0]); |
161 : | sad += abs(ptr_cur[1] - ptr_ref[1]); | ||
162 : | sad += abs(ptr_cur[2] - ptr_ref[2]); | ||
163 : | sad += abs(ptr_cur[3] - ptr_ref[3]); | ||
164 : | sad += abs(ptr_cur[4] - ptr_ref[4]); | ||
165 : | sad += abs(ptr_cur[5] - ptr_ref[5]); | ||
166 : | sad += abs(ptr_cur[6] - ptr_ref[6]); | ||
167 : | sad += abs(ptr_cur[7] - ptr_ref[7]); | ||
168 : | |||
169 : | edgomez | 195 | ptr_cur += stride; |
170 : | ptr_ref += stride; | ||
171 : | |||
172 : | } | ||
173 : | |||
174 : | return sad; | ||
175 : | } | ||
176 : | |||
177 : | |||
178 : | /* average deviation from mean */ | ||
179 : | |||
180 : | uint32_t | ||
181 : | dev16_c(const uint8_t * const cur, | ||
182 : | const uint32_t stride) | ||
183 : | { | ||
184 : | |||
185 : | uint32_t mean = 0; | ||
186 : | uint32_t dev = 0; | ||
187 : | uint32_t i, j; | ||
188 : | uint8_t const *ptr_cur = cur; | ||
189 : | |||
190 : | for (j = 0; j < 16; j++) { | ||
191 : | |||
192 : | for (i = 0; i < 16; i++) | ||
193 : | mean += *(ptr_cur + i); | ||
194 : | |||
195 : | ptr_cur += stride; | ||
196 : | |||
197 : | } | ||
198 : | |||
199 : | mean /= (16 * 16); | ||
200 : | ptr_cur = cur; | ||
201 : | |||
202 : | for (j = 0; j < 16; j++) { | ||
203 : | |||
204 : | for (i = 0; i < 16; i++) | ||
205 : | edgomez | 1382 | dev += abs(*(ptr_cur + i) - (int32_t) mean); |
206 : | edgomez | 195 | |
207 : | ptr_cur += stride; | ||
208 : | |||
209 : | } | ||
210 : | |||
211 : | return dev; | ||
212 : | } | ||
213 : | edgomez | 851 | |
214 : | edgomez | 1382 | uint32_t sad16v_c(const uint8_t * const cur, |
215 : | const uint8_t * const ref, | ||
216 : | const uint32_t stride, | ||
217 : | edgomez | 851 | int32_t *sad) |
218 : | { | ||
219 : | sad[0] = sad8(cur, ref, stride); | ||
220 : | sad[1] = sad8(cur + 8, ref + 8, stride); | ||
221 : | sad[2] = sad8(cur + 8*stride, ref + 8*stride, stride); | ||
222 : | sad[3] = sad8(cur + 8*stride + 8, ref + 8*stride + 8, stride); | ||
223 : | edgomez | 1382 | |
224 : | edgomez | 851 | return sad[0]+sad[1]+sad[2]+sad[3]; |
225 : | } | ||
226 : | |||
227 : | edgomez | 1382 | uint32_t sad32v_c(const uint8_t * const cur, |
228 : | const uint8_t * const ref, | ||
229 : | const uint32_t stride, | ||
230 : | edgomez | 851 | int32_t *sad) |
231 : | { | ||
232 : | sad[0] = sad16(cur, ref, stride, 256*4096); | ||
233 : | edgomez | 1382 | sad[1] = sad16(cur + 16, ref + 16, stride, 256*4096); |
234 : | sad[2] = sad16(cur + 16*stride, ref + 16*stride, stride, 256*4096); | ||
235 : | sad[3] = sad16(cur + 16*stride + 16, ref + 16*stride + 16, stride, 256*4096); | ||
236 : | |||
237 : | edgomez | 851 | return sad[0]+sad[1]+sad[2]+sad[3]; |
238 : | } | ||
239 : | |||
240 : | |||
241 : | |||
242 : | #define MRSAD16_CORRFACTOR 8 | ||
243 : | uint32_t | ||
244 : | mrsad16_c(const uint8_t * const cur, | ||
245 : | const uint8_t * const ref, | ||
246 : | const uint32_t stride, | ||
247 : | const uint32_t best_sad) | ||
248 : | { | ||
249 : | |||
250 : | uint32_t sad = 0; | ||
251 : | int32_t mean = 0; | ||
252 : | uint32_t i, j; | ||
253 : | uint8_t const *ptr_cur = cur; | ||
254 : | uint8_t const *ptr_ref = ref; | ||
255 : | |||
256 : | for (j = 0; j < 16; j++) { | ||
257 : | for (i = 0; i < 16; i++) { | ||
258 : | mean += ((int) *(ptr_cur + i) - (int) *(ptr_ref + i)); | ||
259 : | } | ||
260 : | ptr_cur += stride; | ||
261 : | ptr_ref += stride; | ||
262 : | |||
263 : | } | ||
264 : | mean /= 256; | ||
265 : | |||
266 : | for (j = 0; j < 16; j++) { | ||
267 : | |||
268 : | ptr_cur -= stride; | ||
269 : | ptr_ref -= stride; | ||
270 : | |||
271 : | for (i = 0; i < 16; i++) { | ||
272 : | |||
273 : | edgomez | 1382 | sad += abs(*(ptr_cur + i) - *(ptr_ref + i) - mean); |
274 : | edgomez | 851 | if (sad >= best_sad) { |
275 : | return MRSAD16_CORRFACTOR * sad; | ||
276 : | } | ||
277 : | } | ||
278 : | } | ||
279 : | |||
280 : | return MRSAD16_CORRFACTOR * sad; | ||
281 : | } | ||
282 : | |||
283 : | edgomez | 1382 | uint32_t |
284 : | sse8_16bit_c(const int16_t * b1, | ||
285 : | const int16_t * b2, | ||
286 : | const uint32_t stride) | ||
287 : | { | ||
288 : | int i; | ||
289 : | int sse = 0; | ||
290 : | edgomez | 851 | |
291 : | edgomez | 1382 | for (i=0; i<8; i++) { |
292 : | sse += (b1[0] - b2[0])*(b1[0] - b2[0]); | ||
293 : | sse += (b1[1] - b2[1])*(b1[1] - b2[1]); | ||
294 : | sse += (b1[2] - b2[2])*(b1[2] - b2[2]); | ||
295 : | sse += (b1[3] - b2[3])*(b1[3] - b2[3]); | ||
296 : | sse += (b1[4] - b2[4])*(b1[4] - b2[4]); | ||
297 : | sse += (b1[5] - b2[5])*(b1[5] - b2[5]); | ||
298 : | sse += (b1[6] - b2[6])*(b1[6] - b2[6]); | ||
299 : | sse += (b1[7] - b2[7])*(b1[7] - b2[7]); | ||
300 : | |||
301 : | b1 = (const int16_t*)((int8_t*)b1+stride); | ||
302 : | b2 = (const int16_t*)((int8_t*)b2+stride); | ||
303 : | } | ||
304 : | |||
305 : | return(sse); | ||
306 : | } | ||
307 : | edgomez | 1424 | |
308 : | uint32_t | ||
309 : | sse8_8bit_c(const uint8_t * b1, | ||
310 : | const uint8_t * b2, | ||
311 : | const uint32_t stride) | ||
312 : | { | ||
313 : | int i; | ||
314 : | int sse = 0; | ||
315 : | |||
316 : | for (i=0; i<8; i++) { | ||
317 : | sse += (b1[0] - b2[0])*(b1[0] - b2[0]); | ||
318 : | sse += (b1[1] - b2[1])*(b1[1] - b2[1]); | ||
319 : | sse += (b1[2] - b2[2])*(b1[2] - b2[2]); | ||
320 : | sse += (b1[3] - b2[3])*(b1[3] - b2[3]); | ||
321 : | sse += (b1[4] - b2[4])*(b1[4] - b2[4]); | ||
322 : | sse += (b1[5] - b2[5])*(b1[5] - b2[5]); | ||
323 : | sse += (b1[6] - b2[6])*(b1[6] - b2[6]); | ||
324 : | sse += (b1[7] - b2[7])*(b1[7] - b2[7]); | ||
325 : | |||
326 : | b1 = b1+stride; | ||
327 : | b2 = b2+stride; | ||
328 : | } | ||
329 : | |||
330 : | return(sse); | ||
331 : | } |
No admin address has been configured | ViewVC Help |
Powered by ViewVC 1.0.4 |