Parent Directory | Revision Log
Revision 1205 - (view) (download)
1 : | edgomez | 1054 | /***************************************************************************** |
2 : | edgomez | 195 | * |
3 : | edgomez | 1054 | * XVID MPEG-4 VIDEO CODEC |
4 : | * - Sum Of Absolute Difference related code - | ||
5 : | edgomez | 195 | * |
6 : | edgomez | 1054 | * Copyright(C) 2001-2003 Peter Ross <pross@xvid.org> |
7 : | edgomez | 195 | * |
8 : | edgomez | 1054 | * 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 | 1054 | * 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 | 1054 | * 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 | 1205 | * $Id: sad.c,v 1.13.2.8 2003-11-13 23:11:24 edgomez Exp $ |
23 : | edgomez | 195 | * |
24 : | edgomez | 1054 | ****************************************************************************/ |
25 : | edgomez | 195 | |
26 : | #include "../portab.h" | ||
27 : | edgomez | 851 | #include "../global.h" |
28 : | edgomez | 195 | #include "sad.h" |
29 : | |||
30 : | edgomez | 982 | #include <stdlib.h> |
31 : | |||
32 : | edgomez | 195 | sad16FuncPtr sad16; |
33 : | sad8FuncPtr sad8; | ||
34 : | sad16biFuncPtr sad16bi; | ||
35 : | edgomez | 1053 | sad8biFuncPtr sad8bi; /* not really sad16, but no difference in prototype */ |
36 : | edgomez | 195 | dev16FuncPtr dev16; |
37 : | edgomez | 851 | sad16vFuncPtr sad16v; |
38 : | edgomez | 1205 | sse8Func_16bitPtr sse8_16bit; |
39 : | edgomez | 195 | |
40 : | sadInitFuncPtr sadInit; | ||
41 : | |||
42 : | |||
43 : | uint32_t | ||
44 : | sad16_c(const uint8_t * const cur, | ||
45 : | const uint8_t * const ref, | ||
46 : | const uint32_t stride, | ||
47 : | const uint32_t best_sad) | ||
48 : | { | ||
49 : | |||
50 : | uint32_t sad = 0; | ||
51 : | edgomez | 851 | uint32_t j; |
52 : | edgomez | 195 | uint8_t const *ptr_cur = cur; |
53 : | uint8_t const *ptr_ref = ref; | ||
54 : | |||
55 : | for (j = 0; j < 16; j++) { | ||
56 : | edgomez | 982 | sad += abs(ptr_cur[0] - ptr_ref[0]); |
57 : | sad += abs(ptr_cur[1] - ptr_ref[1]); | ||
58 : | sad += abs(ptr_cur[2] - ptr_ref[2]); | ||
59 : | sad += abs(ptr_cur[3] - ptr_ref[3]); | ||
60 : | sad += abs(ptr_cur[4] - ptr_ref[4]); | ||
61 : | sad += abs(ptr_cur[5] - ptr_ref[5]); | ||
62 : | sad += abs(ptr_cur[6] - ptr_ref[6]); | ||
63 : | sad += abs(ptr_cur[7] - ptr_ref[7]); | ||
64 : | sad += abs(ptr_cur[8] - ptr_ref[8]); | ||
65 : | sad += abs(ptr_cur[9] - ptr_ref[9]); | ||
66 : | sad += abs(ptr_cur[10] - ptr_ref[10]); | ||
67 : | sad += abs(ptr_cur[11] - ptr_ref[11]); | ||
68 : | sad += abs(ptr_cur[12] - ptr_ref[12]); | ||
69 : | sad += abs(ptr_cur[13] - ptr_ref[13]); | ||
70 : | sad += abs(ptr_cur[14] - ptr_ref[14]); | ||
71 : | sad += abs(ptr_cur[15] - ptr_ref[15]); | ||
72 : | edgomez | 195 | |
73 : | edgomez | 1161 | if (sad >= best_sad) |
74 : | edgomez | 195 | return sad; |
75 : | |||
76 : | edgomez | 851 | ptr_cur += stride; |
77 : | ptr_ref += stride; | ||
78 : | edgomez | 195 | |
79 : | } | ||
80 : | |||
81 : | return sad; | ||
82 : | |||
83 : | } | ||
84 : | |||
85 : | uint32_t | ||
86 : | sad16bi_c(const uint8_t * const cur, | ||
87 : | const uint8_t * const ref1, | ||
88 : | const uint8_t * const ref2, | ||
89 : | const uint32_t stride) | ||
90 : | { | ||
91 : | |||
92 : | uint32_t sad = 0; | ||
93 : | uint32_t i, j; | ||
94 : | uint8_t const *ptr_cur = cur; | ||
95 : | uint8_t const *ptr_ref1 = ref1; | ||
96 : | uint8_t const *ptr_ref2 = ref2; | ||
97 : | |||
98 : | for (j = 0; j < 16; j++) { | ||
99 : | |||
100 : | for (i = 0; i < 16; i++) { | ||
101 : | int pixel = (ptr_ref1[i] + ptr_ref2[i] + 1) / 2; | ||
102 : | edgomez | 982 | sad += abs(ptr_cur[i] - pixel); |
103 : | edgomez | 195 | } |
104 : | |||
105 : | ptr_cur += stride; | ||
106 : | ptr_ref1 += stride; | ||
107 : | ptr_ref2 += stride; | ||
108 : | |||
109 : | } | ||
110 : | |||
111 : | return sad; | ||
112 : | |||
113 : | } | ||
114 : | |||
115 : | uint32_t | ||
116 : | sad8bi_c(const uint8_t * const cur, | ||
117 : | const uint8_t * const ref1, | ||
118 : | const uint8_t * const ref2, | ||
119 : | const uint32_t stride) | ||
120 : | { | ||
121 : | |||
122 : | uint32_t sad = 0; | ||
123 : | chl | 326 | uint32_t i, j; |
124 : | edgomez | 195 | uint8_t const *ptr_cur = cur; |
125 : | chl | 326 | uint8_t const *ptr_ref1 = ref1; |
126 : | uint8_t const *ptr_ref2 = ref2; | ||
127 : | |||
128 : | for (j = 0; j < 8; j++) { | ||
129 : | |||
130 : | for (i = 0; i < 8; i++) { | ||
131 : | int pixel = (ptr_ref1[i] + ptr_ref2[i] + 1) / 2; | ||
132 : | edgomez | 982 | sad += abs(ptr_cur[i] - pixel); |
133 : | chl | 326 | } |
134 : | |||
135 : | ptr_cur += stride; | ||
136 : | ptr_ref1 += stride; | ||
137 : | ptr_ref2 += stride; | ||
138 : | |||
139 : | } | ||
140 : | |||
141 : | return sad; | ||
142 : | |||
143 : | } | ||
144 : | |||
145 : | |||
146 : | |||
147 : | uint32_t | ||
148 : | sad8_c(const uint8_t * const cur, | ||
149 : | const uint8_t * const ref, | ||
150 : | const uint32_t stride) | ||
151 : | { | ||
152 : | uint32_t sad = 0; | ||
153 : | edgomez | 851 | uint32_t j; |
154 : | chl | 326 | uint8_t const *ptr_cur = cur; |
155 : | edgomez | 195 | uint8_t const *ptr_ref = ref; |
156 : | |||
157 : | for (j = 0; j < 8; j++) { | ||
158 : | |||
159 : | edgomez | 982 | sad += abs(ptr_cur[0] - ptr_ref[0]); |
160 : | sad += abs(ptr_cur[1] - ptr_ref[1]); | ||
161 : | sad += abs(ptr_cur[2] - ptr_ref[2]); | ||
162 : | sad += abs(ptr_cur[3] - ptr_ref[3]); | ||
163 : | sad += abs(ptr_cur[4] - ptr_ref[4]); | ||
164 : | sad += abs(ptr_cur[5] - ptr_ref[5]); | ||
165 : | sad += abs(ptr_cur[6] - ptr_ref[6]); | ||
166 : | sad += abs(ptr_cur[7] - ptr_ref[7]); | ||
167 : | edgomez | 1161 | |
168 : | edgomez | 195 | ptr_cur += stride; |
169 : | ptr_ref += stride; | ||
170 : | |||
171 : | } | ||
172 : | |||
173 : | return sad; | ||
174 : | } | ||
175 : | |||
176 : | |||
177 : | /* average deviation from mean */ | ||
178 : | |||
179 : | uint32_t | ||
180 : | dev16_c(const uint8_t * const cur, | ||
181 : | const uint32_t stride) | ||
182 : | { | ||
183 : | |||
184 : | uint32_t mean = 0; | ||
185 : | uint32_t dev = 0; | ||
186 : | uint32_t i, j; | ||
187 : | uint8_t const *ptr_cur = cur; | ||
188 : | |||
189 : | for (j = 0; j < 16; j++) { | ||
190 : | |||
191 : | for (i = 0; i < 16; i++) | ||
192 : | mean += *(ptr_cur + i); | ||
193 : | |||
194 : | ptr_cur += stride; | ||
195 : | |||
196 : | } | ||
197 : | |||
198 : | mean /= (16 * 16); | ||
199 : | ptr_cur = cur; | ||
200 : | |||
201 : | for (j = 0; j < 16; j++) { | ||
202 : | |||
203 : | for (i = 0; i < 16; i++) | ||
204 : | edgomez | 982 | dev += abs(*(ptr_cur + i) - (int32_t) mean); |
205 : | edgomez | 195 | |
206 : | ptr_cur += stride; | ||
207 : | |||
208 : | } | ||
209 : | |||
210 : | return dev; | ||
211 : | } | ||
212 : | edgomez | 851 | |
213 : | edgomez | 1161 | uint32_t sad16v_c(const uint8_t * const cur, |
214 : | const uint8_t * const ref, | ||
215 : | const uint32_t stride, | ||
216 : | edgomez | 851 | int32_t *sad) |
217 : | { | ||
218 : | sad[0] = sad8(cur, ref, stride); | ||
219 : | sad[1] = sad8(cur + 8, ref + 8, stride); | ||
220 : | sad[2] = sad8(cur + 8*stride, ref + 8*stride, stride); | ||
221 : | sad[3] = sad8(cur + 8*stride + 8, ref + 8*stride + 8, stride); | ||
222 : | edgomez | 1161 | |
223 : | edgomez | 851 | return sad[0]+sad[1]+sad[2]+sad[3]; |
224 : | } | ||
225 : | |||
226 : | edgomez | 1161 | uint32_t sad32v_c(const uint8_t * const cur, |
227 : | const uint8_t * const ref, | ||
228 : | const uint32_t stride, | ||
229 : | edgomez | 851 | int32_t *sad) |
230 : | { | ||
231 : | sad[0] = sad16(cur, ref, stride, 256*4096); | ||
232 : | edgomez | 1124 | sad[1] = sad16(cur + 16, ref + 16, stride, 256*4096); |
233 : | sad[2] = sad16(cur + 16*stride, ref + 16*stride, stride, 256*4096); | ||
234 : | sad[3] = sad16(cur + 16*stride + 16, ref + 16*stride + 16, stride, 256*4096); | ||
235 : | edgomez | 1161 | |
236 : | edgomez | 851 | return sad[0]+sad[1]+sad[2]+sad[3]; |
237 : | } | ||
238 : | |||
239 : | |||
240 : | |||
241 : | #define MRSAD16_CORRFACTOR 8 | ||
242 : | uint32_t | ||
243 : | mrsad16_c(const uint8_t * const cur, | ||
244 : | const uint8_t * const ref, | ||
245 : | const uint32_t stride, | ||
246 : | const uint32_t best_sad) | ||
247 : | { | ||
248 : | |||
249 : | uint32_t sad = 0; | ||
250 : | int32_t mean = 0; | ||
251 : | uint32_t i, j; | ||
252 : | uint8_t const *ptr_cur = cur; | ||
253 : | uint8_t const *ptr_ref = ref; | ||
254 : | |||
255 : | for (j = 0; j < 16; j++) { | ||
256 : | for (i = 0; i < 16; i++) { | ||
257 : | mean += ((int) *(ptr_cur + i) - (int) *(ptr_ref + i)); | ||
258 : | } | ||
259 : | ptr_cur += stride; | ||
260 : | ptr_ref += stride; | ||
261 : | |||
262 : | } | ||
263 : | mean /= 256; | ||
264 : | |||
265 : | for (j = 0; j < 16; j++) { | ||
266 : | |||
267 : | ptr_cur -= stride; | ||
268 : | ptr_ref -= stride; | ||
269 : | |||
270 : | for (i = 0; i < 16; i++) { | ||
271 : | |||
272 : | edgomez | 982 | sad += abs(*(ptr_cur + i) - *(ptr_ref + i) - mean); |
273 : | edgomez | 851 | if (sad >= best_sad) { |
274 : | return MRSAD16_CORRFACTOR * sad; | ||
275 : | } | ||
276 : | } | ||
277 : | } | ||
278 : | |||
279 : | return MRSAD16_CORRFACTOR * sad; | ||
280 : | } | ||
281 : | edgomez | 1205 | |
282 : | uint32_t | ||
283 : | sse8_16bit_c(const int16_t * b1, | ||
284 : | const int16_t * b2, | ||
285 : | const uint32_t stride) | ||
286 : | { | ||
287 : | int i; | ||
288 : | int sse = 0; | ||
289 : | |||
290 : | for (i=0; i<8; i++) { | ||
291 : | sse += (b1[0] - b2[0])*(b1[0] - b2[0]); | ||
292 : | sse += (b1[1] - b2[1])*(b1[1] - b2[1]); | ||
293 : | sse += (b1[2] - b2[2])*(b1[2] - b2[2]); | ||
294 : | sse += (b1[3] - b2[3])*(b1[3] - b2[3]); | ||
295 : | sse += (b1[4] - b2[4])*(b1[4] - b2[4]); | ||
296 : | sse += (b1[5] - b2[5])*(b1[5] - b2[5]); | ||
297 : | sse += (b1[6] - b2[6])*(b1[6] - b2[6]); | ||
298 : | sse += (b1[7] - b2[7])*(b1[7] - b2[7]); | ||
299 : | |||
300 : | b1 = (const int16_t*)((int8_t*)b1+stride); | ||
301 : | b2 = (const int16_t*)((int8_t*)b2+stride); | ||
302 : | } | ||
303 : | |||
304 : | return(sse); | ||
305 : | } |
No admin address has been configured | ViewVC Help |
Powered by ViewVC 1.0.4 |