Parent Directory | Revision Log
Revision 1566 - (view) (download)
1 : | edgomez | 851 | /************************************************************************** |
2 : | albeu | 315 | * |
3 : | edgomez | 1382 | * XVID MPEG-4 VIDEO CODEC |
4 : | * - Image management functions - | ||
5 : | albeu | 315 | * |
6 : | suxen_drol | 1397 | * Copyright(C) 2001-2004 Peter Ross <pross@xvid.org> |
7 : | albeu | 315 | * |
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 : | albeu | 315 | * |
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 : | albeu | 315 | * |
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 : | albeu | 315 | * |
22 : | syskin | 1566 | * $Id: image.c,v 1.30 2004-12-05 13:56:13 syskin Exp $ |
23 : | albeu | 315 | * |
24 : | edgomez | 1382 | ****************************************************************************/ |
25 : | albeu | 315 | |
26 : | #include <stdlib.h> | ||
27 : | edgomez | 1382 | #include <string.h> /* memcpy, memset */ |
28 : | albeu | 315 | #include <math.h> |
29 : | #include "../portab.h" | ||
30 : | edgomez | 1382 | #include "../global.h" |
31 : | #include "../xvid.h" | ||
32 : | albeu | 315 | #include "image.h" |
33 : | #include "colorspace.h" | ||
34 : | #include "interpolate8x8.h" | ||
35 : | #include "../utils/mem_align.h" | ||
36 : | edgomez | 1424 | #include "../motion/sad.h" |
37 : | albeu | 315 | |
38 : | edgomez | 1382 | #include "font.h" |
39 : | edgomez | 851 | |
40 : | albeu | 315 | #define SAFETY 64 |
41 : | #define EDGE_SIZE2 (EDGE_SIZE/2) | ||
42 : | |||
43 : | |||
44 : | int32_t | ||
45 : | image_create(IMAGE * image, | ||
46 : | uint32_t edged_width, | ||
47 : | uint32_t edged_height) | ||
48 : | { | ||
49 : | const uint32_t edged_width2 = edged_width / 2; | ||
50 : | const uint32_t edged_height2 = edged_height / 2; | ||
51 : | |||
52 : | image->y = | ||
53 : | xvid_malloc(edged_width * (edged_height + 1) + SAFETY, CACHE_LINE); | ||
54 : | if (image->y == NULL) { | ||
55 : | return -1; | ||
56 : | } | ||
57 : | edgomez | 1382 | memset(image->y, 0, edged_width * (edged_height + 1) + SAFETY); |
58 : | albeu | 315 | |
59 : | image->u = xvid_malloc(edged_width2 * edged_height2 + SAFETY, CACHE_LINE); | ||
60 : | if (image->u == NULL) { | ||
61 : | xvid_free(image->y); | ||
62 : | edgomez | 1382 | image->y = NULL; |
63 : | albeu | 315 | return -1; |
64 : | } | ||
65 : | edgomez | 1382 | memset(image->u, 0, edged_width2 * edged_height2 + SAFETY); |
66 : | |||
67 : | albeu | 315 | image->v = xvid_malloc(edged_width2 * edged_height2 + SAFETY, CACHE_LINE); |
68 : | if (image->v == NULL) { | ||
69 : | xvid_free(image->u); | ||
70 : | edgomez | 1382 | image->u = NULL; |
71 : | albeu | 315 | xvid_free(image->y); |
72 : | edgomez | 1382 | image->y = NULL; |
73 : | albeu | 315 | return -1; |
74 : | } | ||
75 : | edgomez | 1382 | memset(image->v, 0, edged_width2 * edged_height2 + SAFETY); |
76 : | albeu | 315 | |
77 : | image->y += EDGE_SIZE * edged_width + EDGE_SIZE; | ||
78 : | image->u += EDGE_SIZE2 * edged_width2 + EDGE_SIZE2; | ||
79 : | image->v += EDGE_SIZE2 * edged_width2 + EDGE_SIZE2; | ||
80 : | |||
81 : | return 0; | ||
82 : | } | ||
83 : | |||
84 : | |||
85 : | |||
86 : | void | ||
87 : | image_destroy(IMAGE * image, | ||
88 : | uint32_t edged_width, | ||
89 : | uint32_t edged_height) | ||
90 : | { | ||
91 : | const uint32_t edged_width2 = edged_width / 2; | ||
92 : | |||
93 : | if (image->y) { | ||
94 : | xvid_free(image->y - (EDGE_SIZE * edged_width + EDGE_SIZE)); | ||
95 : | edgomez | 1382 | image->y = NULL; |
96 : | albeu | 315 | } |
97 : | if (image->u) { | ||
98 : | xvid_free(image->u - (EDGE_SIZE2 * edged_width2 + EDGE_SIZE2)); | ||
99 : | edgomez | 1382 | image->u = NULL; |
100 : | albeu | 315 | } |
101 : | if (image->v) { | ||
102 : | xvid_free(image->v - (EDGE_SIZE2 * edged_width2 + EDGE_SIZE2)); | ||
103 : | edgomez | 1382 | image->v = NULL; |
104 : | albeu | 315 | } |
105 : | } | ||
106 : | |||
107 : | |||
108 : | void | ||
109 : | image_swap(IMAGE * image1, | ||
110 : | IMAGE * image2) | ||
111 : | { | ||
112 : | edgomez | 1382 | SWAP(uint8_t*, image1->y, image2->y); |
113 : | SWAP(uint8_t*, image1->u, image2->u); | ||
114 : | SWAP(uint8_t*, image1->v, image2->v); | ||
115 : | albeu | 315 | } |
116 : | |||
117 : | |||
118 : | void | ||
119 : | image_copy(IMAGE * image1, | ||
120 : | IMAGE * image2, | ||
121 : | uint32_t edged_width, | ||
122 : | uint32_t height) | ||
123 : | { | ||
124 : | memcpy(image1->y, image2->y, edged_width * height); | ||
125 : | memcpy(image1->u, image2->u, edged_width * height / 4); | ||
126 : | memcpy(image1->v, image2->v, edged_width * height / 4); | ||
127 : | } | ||
128 : | |||
129 : | edgomez | 1382 | /* setedges bug was fixed in this BS version */ |
130 : | #define SETEDGES_BUG_BEFORE 18 | ||
131 : | albeu | 315 | |
132 : | void | ||
133 : | image_setedges(IMAGE * image, | ||
134 : | uint32_t edged_width, | ||
135 : | uint32_t edged_height, | ||
136 : | uint32_t width, | ||
137 : | edgomez | 1382 | uint32_t height, |
138 : | int bs_version) | ||
139 : | albeu | 315 | { |
140 : | const uint32_t edged_width2 = edged_width / 2; | ||
141 : | edgomez | 1382 | uint32_t width2; |
142 : | albeu | 315 | uint32_t i; |
143 : | uint8_t *dst; | ||
144 : | uint8_t *src; | ||
145 : | |||
146 : | dst = image->y - (EDGE_SIZE + EDGE_SIZE * edged_width); | ||
147 : | src = image->y; | ||
148 : | |||
149 : | edgomez | 1382 | /* According to the Standard Clause 7.6.4, padding is done starting at 16 |
150 : | * pixel width and height multiples. This was not respected in old xvids */ | ||
151 : | if (bs_version == 0 || bs_version >= SETEDGES_BUG_BEFORE) { | ||
152 : | width = (width+15)&~15; | ||
153 : | height = (height+15)&~15; | ||
154 : | } | ||
155 : | |||
156 : | width2 = width/2; | ||
157 : | |||
158 : | albeu | 315 | for (i = 0; i < EDGE_SIZE; i++) { |
159 : | edgomez | 851 | memset(dst, *src, EDGE_SIZE); |
160 : | memcpy(dst + EDGE_SIZE, src, width); | ||
161 : | memset(dst + edged_width - EDGE_SIZE, *(src + width - 1), | ||
162 : | EDGE_SIZE); | ||
163 : | albeu | 315 | dst += edged_width; |
164 : | } | ||
165 : | |||
166 : | for (i = 0; i < height; i++) { | ||
167 : | memset(dst, *src, EDGE_SIZE); | ||
168 : | memset(dst + edged_width - EDGE_SIZE, src[width - 1], EDGE_SIZE); | ||
169 : | dst += edged_width; | ||
170 : | src += edged_width; | ||
171 : | } | ||
172 : | |||
173 : | src -= edged_width; | ||
174 : | for (i = 0; i < EDGE_SIZE; i++) { | ||
175 : | edgomez | 851 | memset(dst, *src, EDGE_SIZE); |
176 : | memcpy(dst + EDGE_SIZE, src, width); | ||
177 : | memset(dst + edged_width - EDGE_SIZE, *(src + width - 1), | ||
178 : | albeu | 315 | EDGE_SIZE); |
179 : | dst += edged_width; | ||
180 : | } | ||
181 : | |||
182 : | |||
183 : | edgomez | 1382 | /* U */ |
184 : | albeu | 315 | dst = image->u - (EDGE_SIZE2 + EDGE_SIZE2 * edged_width2); |
185 : | src = image->u; | ||
186 : | |||
187 : | for (i = 0; i < EDGE_SIZE2; i++) { | ||
188 : | memset(dst, *src, EDGE_SIZE2); | ||
189 : | memcpy(dst + EDGE_SIZE2, src, width2); | ||
190 : | memset(dst + edged_width2 - EDGE_SIZE2, *(src + width2 - 1), | ||
191 : | EDGE_SIZE2); | ||
192 : | dst += edged_width2; | ||
193 : | } | ||
194 : | |||
195 : | for (i = 0; i < height / 2; i++) { | ||
196 : | memset(dst, *src, EDGE_SIZE2); | ||
197 : | memset(dst + edged_width2 - EDGE_SIZE2, src[width2 - 1], EDGE_SIZE2); | ||
198 : | dst += edged_width2; | ||
199 : | src += edged_width2; | ||
200 : | } | ||
201 : | src -= edged_width2; | ||
202 : | for (i = 0; i < EDGE_SIZE2; i++) { | ||
203 : | memset(dst, *src, EDGE_SIZE2); | ||
204 : | memcpy(dst + EDGE_SIZE2, src, width2); | ||
205 : | memset(dst + edged_width2 - EDGE_SIZE2, *(src + width2 - 1), | ||
206 : | EDGE_SIZE2); | ||
207 : | dst += edged_width2; | ||
208 : | } | ||
209 : | |||
210 : | |||
211 : | edgomez | 1382 | /* V */ |
212 : | albeu | 315 | dst = image->v - (EDGE_SIZE2 + EDGE_SIZE2 * edged_width2); |
213 : | src = image->v; | ||
214 : | |||
215 : | for (i = 0; i < EDGE_SIZE2; i++) { | ||
216 : | memset(dst, *src, EDGE_SIZE2); | ||
217 : | memcpy(dst + EDGE_SIZE2, src, width2); | ||
218 : | memset(dst + edged_width2 - EDGE_SIZE2, *(src + width2 - 1), | ||
219 : | EDGE_SIZE2); | ||
220 : | dst += edged_width2; | ||
221 : | } | ||
222 : | |||
223 : | for (i = 0; i < height / 2; i++) { | ||
224 : | memset(dst, *src, EDGE_SIZE2); | ||
225 : | memset(dst + edged_width2 - EDGE_SIZE2, src[width2 - 1], EDGE_SIZE2); | ||
226 : | dst += edged_width2; | ||
227 : | src += edged_width2; | ||
228 : | } | ||
229 : | src -= edged_width2; | ||
230 : | for (i = 0; i < EDGE_SIZE2; i++) { | ||
231 : | memset(dst, *src, EDGE_SIZE2); | ||
232 : | memcpy(dst + EDGE_SIZE2, src, width2); | ||
233 : | memset(dst + edged_width2 - EDGE_SIZE2, *(src + width2 - 1), | ||
234 : | EDGE_SIZE2); | ||
235 : | dst += edged_width2; | ||
236 : | } | ||
237 : | } | ||
238 : | |||
239 : | edgomez | 1382 | /* bframe encoding requires image-based u,v interpolation */ |
240 : | albeu | 315 | void |
241 : | image_interpolate(const IMAGE * refn, | ||
242 : | IMAGE * refh, | ||
243 : | IMAGE * refv, | ||
244 : | IMAGE * refhv, | ||
245 : | uint32_t edged_width, | ||
246 : | uint32_t edged_height, | ||
247 : | edgomez | 851 | uint32_t quarterpel, |
248 : | albeu | 315 | uint32_t rounding) |
249 : | { | ||
250 : | edgomez | 1382 | const uint32_t offset = EDGE_SIZE2 * (edged_width + 1); /* we only interpolate half of the edge area */ |
251 : | albeu | 315 | const uint32_t stride_add = 7 * edged_width; |
252 : | edgomez | 1382 | #if 0 |
253 : | edgomez | 851 | const uint32_t edged_width2 = edged_width / 2; |
254 : | const uint32_t edged_height2 = edged_height / 2; | ||
255 : | const uint32_t offset2 = EDGE_SIZE2 * (edged_width2 + 1); | ||
256 : | const uint32_t stride_add2 = 7 * edged_width2; | ||
257 : | #endif | ||
258 : | albeu | 315 | uint8_t *n_ptr, *h_ptr, *v_ptr, *hv_ptr; |
259 : | uint32_t x, y; | ||
260 : | |||
261 : | |||
262 : | n_ptr = refn->y; | ||
263 : | h_ptr = refh->y; | ||
264 : | v_ptr = refv->y; | ||
265 : | |||
266 : | n_ptr -= offset; | ||
267 : | h_ptr -= offset; | ||
268 : | v_ptr -= offset; | ||
269 : | |||
270 : | edgomez | 1382 | /* Note we initialize the hv pointer later, as we can optimize code a bit |
271 : | * doing it down to up in quarterpel and up to down in halfpel */ | ||
272 : | edgomez | 851 | if(quarterpel) { |
273 : | edgomez | 1382 | |
274 : | edgomez | 851 | for (y = 0; y < (edged_height - EDGE_SIZE); y += 8) { |
275 : | for (x = 0; x < (edged_width - EDGE_SIZE); x += 8) { | ||
276 : | interpolate8x8_6tap_lowpass_h(h_ptr, n_ptr, edged_width, rounding); | ||
277 : | interpolate8x8_6tap_lowpass_v(v_ptr, n_ptr, edged_width, rounding); | ||
278 : | albeu | 315 | |
279 : | edgomez | 851 | n_ptr += 8; |
280 : | h_ptr += 8; | ||
281 : | v_ptr += 8; | ||
282 : | } | ||
283 : | edgomez | 1382 | |
284 : | edgomez | 851 | n_ptr += EDGE_SIZE; |
285 : | h_ptr += EDGE_SIZE; | ||
286 : | v_ptr += EDGE_SIZE; | ||
287 : | |||
288 : | h_ptr += stride_add; | ||
289 : | v_ptr += stride_add; | ||
290 : | n_ptr += stride_add; | ||
291 : | } | ||
292 : | |||
293 : | edgomez | 1382 | h_ptr = refh->y + (edged_height - EDGE_SIZE - EDGE_SIZE2)*edged_width - EDGE_SIZE2; |
294 : | hv_ptr = refhv->y + (edged_height - EDGE_SIZE - EDGE_SIZE2)*edged_width - EDGE_SIZE2; | ||
295 : | edgomez | 851 | |
296 : | for (y = 0; y < (edged_height - EDGE_SIZE); y = y + 8) { | ||
297 : | edgomez | 1382 | hv_ptr -= stride_add; |
298 : | h_ptr -= stride_add; | ||
299 : | hv_ptr -= EDGE_SIZE; | ||
300 : | h_ptr -= EDGE_SIZE; | ||
301 : | |||
302 : | edgomez | 851 | for (x = 0; x < (edged_width - EDGE_SIZE); x = x + 8) { |
303 : | edgomez | 1382 | hv_ptr -= 8; |
304 : | h_ptr -= 8; | ||
305 : | edgomez | 851 | interpolate8x8_6tap_lowpass_v(hv_ptr, h_ptr, edged_width, rounding); |
306 : | } | ||
307 : | } | ||
308 : | edgomez | 1382 | } else { |
309 : | edgomez | 851 | |
310 : | edgomez | 1382 | hv_ptr = refhv->y; |
311 : | hv_ptr -= offset; | ||
312 : | |||
313 : | edgomez | 851 | for (y = 0; y < (edged_height - EDGE_SIZE); y += 8) { |
314 : | for (x = 0; x < (edged_width - EDGE_SIZE); x += 8) { | ||
315 : | interpolate8x8_halfpel_h(h_ptr, n_ptr, edged_width, rounding); | ||
316 : | interpolate8x8_halfpel_v(v_ptr, n_ptr, edged_width, rounding); | ||
317 : | interpolate8x8_halfpel_hv(hv_ptr, n_ptr, edged_width, rounding); | ||
318 : | |||
319 : | n_ptr += 8; | ||
320 : | h_ptr += 8; | ||
321 : | v_ptr += 8; | ||
322 : | hv_ptr += 8; | ||
323 : | } | ||
324 : | edgomez | 1382 | |
325 : | edgomez | 851 | h_ptr += EDGE_SIZE; |
326 : | v_ptr += EDGE_SIZE; | ||
327 : | hv_ptr += EDGE_SIZE; | ||
328 : | n_ptr += EDGE_SIZE; | ||
329 : | |||
330 : | h_ptr += stride_add; | ||
331 : | v_ptr += stride_add; | ||
332 : | hv_ptr += stride_add; | ||
333 : | n_ptr += stride_add; | ||
334 : | } | ||
335 : | } | ||
336 : | /* | ||
337 : | #ifdef BFRAMES | ||
338 : | n_ptr = refn->u; | ||
339 : | h_ptr = refh->u; | ||
340 : | v_ptr = refv->u; | ||
341 : | hv_ptr = refhv->u; | ||
342 : | |||
343 : | n_ptr -= offset2; | ||
344 : | h_ptr -= offset2; | ||
345 : | v_ptr -= offset2; | ||
346 : | hv_ptr -= offset2; | ||
347 : | |||
348 : | for (y = 0; y < edged_height2; y += 8) { | ||
349 : | for (x = 0; x < edged_width2; x += 8) { | ||
350 : | interpolate8x8_halfpel_h(h_ptr, n_ptr, edged_width2, rounding); | ||
351 : | interpolate8x8_halfpel_v(v_ptr, n_ptr, edged_width2, rounding); | ||
352 : | interpolate8x8_halfpel_hv(hv_ptr, n_ptr, edged_width2, rounding); | ||
353 : | |||
354 : | albeu | 315 | n_ptr += 8; |
355 : | h_ptr += 8; | ||
356 : | v_ptr += 8; | ||
357 : | hv_ptr += 8; | ||
358 : | } | ||
359 : | edgomez | 851 | h_ptr += stride_add2; |
360 : | v_ptr += stride_add2; | ||
361 : | hv_ptr += stride_add2; | ||
362 : | n_ptr += stride_add2; | ||
363 : | albeu | 315 | } |
364 : | |||
365 : | edgomez | 851 | n_ptr = refn->v; |
366 : | h_ptr = refh->v; | ||
367 : | v_ptr = refv->v; | ||
368 : | hv_ptr = refhv->v; | ||
369 : | |||
370 : | n_ptr -= offset2; | ||
371 : | h_ptr -= offset2; | ||
372 : | v_ptr -= offset2; | ||
373 : | hv_ptr -= offset2; | ||
374 : | |||
375 : | for (y = 0; y < edged_height2; y = y + 8) { | ||
376 : | for (x = 0; x < edged_width2; x = x + 8) { | ||
377 : | interpolate8x8_halfpel_h(h_ptr, n_ptr, edged_width2, rounding); | ||
378 : | interpolate8x8_halfpel_v(v_ptr, n_ptr, edged_width2, rounding); | ||
379 : | interpolate8x8_halfpel_hv(hv_ptr, n_ptr, edged_width2, rounding); | ||
380 : | |||
381 : | n_ptr += 8; | ||
382 : | h_ptr += 8; | ||
383 : | v_ptr += 8; | ||
384 : | hv_ptr += 8; | ||
385 : | } | ||
386 : | h_ptr += stride_add2; | ||
387 : | v_ptr += stride_add2; | ||
388 : | hv_ptr += stride_add2; | ||
389 : | n_ptr += stride_add2; | ||
390 : | } | ||
391 : | #endif | ||
392 : | */ | ||
393 : | albeu | 315 | /* |
394 : | interpolate_halfpel_h( | ||
395 : | refh->y - offset, | ||
396 : | edgomez | 1382 | refn->y - offset, |
397 : | albeu | 315 | edged_width, edged_height, |
398 : | rounding); | ||
399 : | |||
400 : | interpolate_halfpel_v( | ||
401 : | refv->y - offset, | ||
402 : | edgomez | 1382 | refn->y - offset, |
403 : | albeu | 315 | edged_width, edged_height, |
404 : | rounding); | ||
405 : | |||
406 : | interpolate_halfpel_hv( | ||
407 : | refhv->y - offset, | ||
408 : | refn->y - offset, | ||
409 : | edged_width, edged_height, | ||
410 : | rounding); | ||
411 : | */ | ||
412 : | |||
413 : | /* uv-image-based compensation | ||
414 : | offset = EDGE_SIZE2 * (edged_width / 2 + 1); | ||
415 : | |||
416 : | interpolate_halfpel_h( | ||
417 : | refh->u - offset, | ||
418 : | edgomez | 1382 | refn->u - offset, |
419 : | albeu | 315 | edged_width / 2, edged_height / 2, |
420 : | rounding); | ||
421 : | |||
422 : | interpolate_halfpel_v( | ||
423 : | refv->u - offset, | ||
424 : | edgomez | 1382 | refn->u - offset, |
425 : | albeu | 315 | edged_width / 2, edged_height / 2, |
426 : | rounding); | ||
427 : | |||
428 : | interpolate_halfpel_hv( | ||
429 : | refhv->u - offset, | ||
430 : | edgomez | 1382 | refn->u - offset, |
431 : | albeu | 315 | edged_width / 2, edged_height / 2, |
432 : | rounding); | ||
433 : | |||
434 : | |||
435 : | interpolate_halfpel_h( | ||
436 : | refh->v - offset, | ||
437 : | edgomez | 1382 | refn->v - offset, |
438 : | albeu | 315 | edged_width / 2, edged_height / 2, |
439 : | rounding); | ||
440 : | |||
441 : | interpolate_halfpel_v( | ||
442 : | refv->v - offset, | ||
443 : | edgomez | 1382 | refn->v - offset, |
444 : | albeu | 315 | edged_width / 2, edged_height / 2, |
445 : | rounding); | ||
446 : | |||
447 : | interpolate_halfpel_hv( | ||
448 : | refhv->v - offset, | ||
449 : | edgomez | 1382 | refn->v - offset, |
450 : | albeu | 315 | edged_width / 2, edged_height / 2, |
451 : | rounding); | ||
452 : | */ | ||
453 : | } | ||
454 : | |||
455 : | |||
456 : | edgomez | 851 | /* |
457 : | chroma optimize filter, invented by mf | ||
458 : | a chroma pixel is average from the surrounding pixels, when the | ||
459 : | correpsonding luma pixels are pure black or white. | ||
460 : | */ | ||
461 : | |||
462 : | void | ||
463 : | image_chroma_optimize(IMAGE * img, int width, int height, int edged_width) | ||
464 : | { | ||
465 : | int x,y; | ||
466 : | int pixels = 0; | ||
467 : | |||
468 : | for (y = 1; y < height/2 - 1; y++) | ||
469 : | for (x = 1; x < width/2 - 1; x++) | ||
470 : | { | ||
471 : | #define IS_PURE(a) ((a)<=16||(a)>=235) | ||
472 : | #define IMG_Y(Y,X) img->y[(Y)*edged_width + (X)] | ||
473 : | #define IMG_U(Y,X) img->u[(Y)*edged_width/2 + (X)] | ||
474 : | #define IMG_V(Y,X) img->v[(Y)*edged_width/2 + (X)] | ||
475 : | |||
476 : | edgomez | 1382 | if (IS_PURE(IMG_Y(y*2 ,x*2 )) && |
477 : | edgomez | 851 | IS_PURE(IMG_Y(y*2 ,x*2+1)) && |
478 : | edgomez | 1382 | IS_PURE(IMG_Y(y*2+1,x*2 )) && |
479 : | edgomez | 851 | IS_PURE(IMG_Y(y*2+1,x*2+1))) |
480 : | { | ||
481 : | IMG_U(y,x) = (IMG_U(y,x-1) + IMG_U(y-1, x) + IMG_U(y, x+1) + IMG_U(y+1, x)) / 4; | ||
482 : | IMG_V(y,x) = (IMG_V(y,x-1) + IMG_V(y-1, x) + IMG_V(y, x+1) + IMG_V(y+1, x)) / 4; | ||
483 : | pixels++; | ||
484 : | } | ||
485 : | |||
486 : | #undef IS_PURE | ||
487 : | #undef IMG_Y | ||
488 : | #undef IMG_U | ||
489 : | #undef IMG_V | ||
490 : | } | ||
491 : | edgomez | 1382 | |
492 : | DPRINTF(XVID_DEBUG_DEBUG,"chroma_optimized_pixels = %i/%i\n", pixels, width*height/4); | ||
493 : | edgomez | 851 | } |
494 : | |||
495 : | |||
496 : | |||
497 : | |||
498 : | |||
499 : | /* | ||
500 : | perform safe packed colorspace conversion, by splitting | ||
501 : | the image up into an optimized area (pixel width divisible by 16), | ||
502 : | and two unoptimized/plain-c areas (pixel width divisible by 2) | ||
503 : | */ | ||
504 : | |||
505 : | edgomez | 1382 | static void |
506 : | edgomez | 851 | safe_packed_conv(uint8_t * x_ptr, int x_stride, |
507 : | uint8_t * y_ptr, uint8_t * u_ptr, uint8_t * v_ptr, | ||
508 : | int y_stride, int uv_stride, | ||
509 : | int width, int height, int vflip, | ||
510 : | packedFunc * func_opt, packedFunc func_c, int size) | ||
511 : | { | ||
512 : | int width_opt, width_c; | ||
513 : | |||
514 : | if (func_opt != func_c && x_stride < size*((width+15)/16)*16) | ||
515 : | { | ||
516 : | width_opt = width & (~15); | ||
517 : | width_c = width - width_opt; | ||
518 : | } | ||
519 : | else | ||
520 : | { | ||
521 : | width_opt = width; | ||
522 : | width_c = 0; | ||
523 : | } | ||
524 : | |||
525 : | func_opt(x_ptr, x_stride, | ||
526 : | y_ptr, u_ptr, v_ptr, y_stride, uv_stride, | ||
527 : | width_opt, height, vflip); | ||
528 : | |||
529 : | if (width_c) | ||
530 : | { | ||
531 : | func_c(x_ptr + size*width_opt, x_stride, | ||
532 : | y_ptr + width_opt, u_ptr + width_opt/2, v_ptr + width_opt/2, | ||
533 : | y_stride, uv_stride, width_c, height, vflip); | ||
534 : | } | ||
535 : | } | ||
536 : | |||
537 : | |||
538 : | |||
539 : | albeu | 315 | int |
540 : | image_input(IMAGE * image, | ||
541 : | uint32_t width, | ||
542 : | int height, | ||
543 : | uint32_t edged_width, | ||
544 : | edgomez | 1382 | uint8_t * src[4], |
545 : | int src_stride[4], | ||
546 : | edgomez | 851 | int csp, |
547 : | int interlacing) | ||
548 : | albeu | 315 | { |
549 : | edgomez | 851 | const int edged_width2 = edged_width/2; |
550 : | const int width2 = width/2; | ||
551 : | const int height2 = height/2; | ||
552 : | edgomez | 1382 | #if 0 |
553 : | const int height_signed = (csp & XVID_CSP_VFLIP) ? -height : height; | ||
554 : | #endif | ||
555 : | albeu | 315 | |
556 : | switch (csp & ~XVID_CSP_VFLIP) { | ||
557 : | case XVID_CSP_RGB555: | ||
558 : | edgomez | 851 | safe_packed_conv( |
559 : | edgomez | 1382 | src[0], src_stride[0], image->y, image->u, image->v, |
560 : | edgomez | 851 | edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
561 : | interlacing?rgb555i_to_yv12 :rgb555_to_yv12, | ||
562 : | interlacing?rgb555i_to_yv12_c:rgb555_to_yv12_c, 2); | ||
563 : | break; | ||
564 : | albeu | 315 | |
565 : | case XVID_CSP_RGB565: | ||
566 : | edgomez | 851 | safe_packed_conv( |
567 : | edgomez | 1382 | src[0], src_stride[0], image->y, image->u, image->v, |
568 : | edgomez | 851 | edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
569 : | interlacing?rgb565i_to_yv12 :rgb565_to_yv12, | ||
570 : | interlacing?rgb565i_to_yv12_c:rgb565_to_yv12_c, 2); | ||
571 : | break; | ||
572 : | albeu | 315 | |
573 : | |||
574 : | edgomez | 1382 | case XVID_CSP_BGR: |
575 : | edgomez | 851 | safe_packed_conv( |
576 : | edgomez | 1382 | src[0], src_stride[0], image->y, image->u, image->v, |
577 : | edgomez | 851 | edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
578 : | interlacing?bgri_to_yv12 :bgr_to_yv12, | ||
579 : | interlacing?bgri_to_yv12_c:bgr_to_yv12_c, 3); | ||
580 : | break; | ||
581 : | albeu | 315 | |
582 : | edgomez | 1382 | case XVID_CSP_BGRA: |
583 : | edgomez | 851 | safe_packed_conv( |
584 : | edgomez | 1382 | src[0], src_stride[0], image->y, image->u, image->v, |
585 : | edgomez | 851 | edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
586 : | interlacing?bgrai_to_yv12 :bgra_to_yv12, | ||
587 : | interlacing?bgrai_to_yv12_c:bgra_to_yv12_c, 4); | ||
588 : | break; | ||
589 : | albeu | 315 | |
590 : | edgomez | 851 | case XVID_CSP_ABGR : |
591 : | safe_packed_conv( | ||
592 : | edgomez | 1382 | src[0], src_stride[0], image->y, image->u, image->v, |
593 : | edgomez | 851 | edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
594 : | interlacing?abgri_to_yv12 :abgr_to_yv12, | ||
595 : | interlacing?abgri_to_yv12_c:abgr_to_yv12_c, 4); | ||
596 : | break; | ||
597 : | albeu | 315 | |
598 : | edgomez | 851 | case XVID_CSP_RGBA : |
599 : | safe_packed_conv( | ||
600 : | edgomez | 1382 | src[0], src_stride[0], image->y, image->u, image->v, |
601 : | edgomez | 851 | edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
602 : | interlacing?rgbai_to_yv12 :rgba_to_yv12, | ||
603 : | interlacing?rgbai_to_yv12_c:rgba_to_yv12_c, 4); | ||
604 : | break; | ||
605 : | edgomez | 1382 | |
606 : | case XVID_CSP_ARGB: | ||
607 : | safe_packed_conv( | ||
608 : | src[0], src_stride[0], image->y, image->u, image->v, | ||
609 : | edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), | ||
610 : | interlacing?argbi_to_yv12 : argb_to_yv12, | ||
611 : | interlacing?argbi_to_yv12_c: argb_to_yv12_c, 4); | ||
612 : | break; | ||
613 : | albeu | 315 | |
614 : | case XVID_CSP_YUY2: | ||
615 : | edgomez | 851 | safe_packed_conv( |
616 : | edgomez | 1382 | src[0], src_stride[0], image->y, image->u, image->v, |
617 : | edgomez | 851 | edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
618 : | interlacing?yuyvi_to_yv12 :yuyv_to_yv12, | ||
619 : | interlacing?yuyvi_to_yv12_c:yuyv_to_yv12_c, 2); | ||
620 : | break; | ||
621 : | albeu | 315 | |
622 : | case XVID_CSP_YVYU: /* u/v swapped */ | ||
623 : | edgomez | 851 | safe_packed_conv( |
624 : | edgomez | 1382 | src[0], src_stride[0], image->y, image->v, image->u, |
625 : | edgomez | 851 | edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
626 : | interlacing?yuyvi_to_yv12 :yuyv_to_yv12, | ||
627 : | interlacing?yuyvi_to_yv12_c:yuyv_to_yv12_c, 2); | ||
628 : | break; | ||
629 : | albeu | 315 | |
630 : | case XVID_CSP_UYVY: | ||
631 : | edgomez | 851 | safe_packed_conv( |
632 : | edgomez | 1382 | src[0], src_stride[0], image->y, image->u, image->v, |
633 : | edgomez | 851 | edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
634 : | interlacing?uyvyi_to_yv12 :uyvy_to_yv12, | ||
635 : | interlacing?uyvyi_to_yv12_c:uyvy_to_yv12_c, 2); | ||
636 : | break; | ||
637 : | albeu | 315 | |
638 : | edgomez | 1382 | case XVID_CSP_I420: /* YCbCr == YUV == internal colorspace for MPEG */ |
639 : | edgomez | 851 | yv12_to_yv12(image->y, image->u, image->v, edged_width, edged_width2, |
640 : | edgomez | 1382 | src[0], src[0] + src_stride[0]*height, src[0] + src_stride[0]*height + (src_stride[0]/2)*height2, |
641 : | src_stride[0], src_stride[0]/2, width, height, (csp & XVID_CSP_VFLIP)); | ||
642 : | break; | ||
643 : | |||
644 : | case XVID_CSP_YV12: /* YCrCb == YVA == U and V plane swapped */ | ||
645 : | edgomez | 851 | yv12_to_yv12(image->y, image->v, image->u, edged_width, edged_width2, |
646 : | edgomez | 1382 | src[0], src[0] + src_stride[0]*height, src[0] + src_stride[0]*height + (src_stride[0]/2)*height2, |
647 : | src_stride[0], src_stride[0]/2, width, height, (csp & XVID_CSP_VFLIP)); | ||
648 : | edgomez | 851 | break; |
649 : | |||
650 : | edgomez | 1382 | case XVID_CSP_PLANAR: /* YCbCr with arbitrary pointers and different strides for Y and UV */ |
651 : | yv12_to_yv12(image->y, image->u, image->v, edged_width, edged_width2, | ||
652 : | src[0], src[1], src[2], src_stride[0], src_stride[1], /* v: dst_stride[2] not yet supported */ | ||
653 : | width, height, (csp & XVID_CSP_VFLIP)); | ||
654 : | edgomez | 851 | break; |
655 : | albeu | 315 | |
656 : | case XVID_CSP_NULL: | ||
657 : | break; | ||
658 : | |||
659 : | edgomez | 851 | default : |
660 : | return -1; | ||
661 : | albeu | 315 | } |
662 : | |||
663 : | edgomez | 851 | |
664 : | /* pad out image when the width and/or height is not a multiple of 16 */ | ||
665 : | |||
666 : | if (width & 15) | ||
667 : | { | ||
668 : | int i; | ||
669 : | int pad_width = 16 - (width&15); | ||
670 : | for (i = 0; i < height; i++) | ||
671 : | { | ||
672 : | edgomez | 1382 | memset(image->y + i*edged_width + width, |
673 : | edgomez | 851 | *(image->y + i*edged_width + width - 1), pad_width); |
674 : | } | ||
675 : | for (i = 0; i < height/2; i++) | ||
676 : | { | ||
677 : | edgomez | 1382 | memset(image->u + i*edged_width2 + width2, |
678 : | edgomez | 851 | *(image->u + i*edged_width2 + width2 - 1),pad_width/2); |
679 : | edgomez | 1382 | memset(image->v + i*edged_width2 + width2, |
680 : | edgomez | 851 | *(image->v + i*edged_width2 + width2 - 1),pad_width/2); |
681 : | } | ||
682 : | } | ||
683 : | |||
684 : | if (height & 15) | ||
685 : | { | ||
686 : | edgomez | 1382 | int pad_height = 16 - (height&15); |
687 : | edgomez | 851 | int length = ((width+15)/16)*16; |
688 : | int i; | ||
689 : | for (i = 0; i < pad_height; i++) | ||
690 : | { | ||
691 : | memcpy(image->y + (height+i)*edged_width, | ||
692 : | image->y + (height-1)*edged_width,length); | ||
693 : | } | ||
694 : | |||
695 : | for (i = 0; i < pad_height/2; i++) | ||
696 : | { | ||
697 : | memcpy(image->u + (height2+i)*edged_width2, | ||
698 : | image->u + (height2-1)*edged_width2,length/2); | ||
699 : | memcpy(image->v + (height2+i)*edged_width2, | ||
700 : | image->v + (height2-1)*edged_width2,length/2); | ||
701 : | } | ||
702 : | } | ||
703 : | |||
704 : | /* | ||
705 : | if (interlacing) | ||
706 : | image_printf(image, edged_width, height, 5,5, "[i]"); | ||
707 : | image_dump_yuvpgm(image, edged_width, ((width+15)/16)*16, ((height+15)/16)*16, "\\encode.pgm"); | ||
708 : | */ | ||
709 : | return 0; | ||
710 : | albeu | 315 | } |
711 : | |||
712 : | |||
713 : | |||
714 : | int | ||
715 : | image_output(IMAGE * image, | ||
716 : | uint32_t width, | ||
717 : | int height, | ||
718 : | uint32_t edged_width, | ||
719 : | edgomez | 1382 | uint8_t * dst[4], |
720 : | uint32_t dst_stride[4], | ||
721 : | edgomez | 851 | int csp, |
722 : | int interlacing) | ||
723 : | albeu | 315 | { |
724 : | edgomez | 851 | const int edged_width2 = edged_width/2; |
725 : | int height2 = height/2; | ||
726 : | |||
727 : | /* | ||
728 : | if (interlacing) | ||
729 : | image_printf(image, edged_width, height, 5,100, "[i]=%i,%i",width,height); | ||
730 : | image_dump_yuvpgm(image, edged_width, width, height, "\\decode.pgm"); | ||
731 : | */ | ||
732 : | |||
733 : | albeu | 315 | switch (csp & ~XVID_CSP_VFLIP) { |
734 : | case XVID_CSP_RGB555: | ||
735 : | edgomez | 851 | safe_packed_conv( |
736 : | edgomez | 1382 | dst[0], dst_stride[0], image->y, image->u, image->v, |
737 : | edgomez | 851 | edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
738 : | interlacing?yv12_to_rgb555i :yv12_to_rgb555, | ||
739 : | interlacing?yv12_to_rgb555i_c:yv12_to_rgb555_c, 2); | ||
740 : | albeu | 315 | return 0; |
741 : | |||
742 : | case XVID_CSP_RGB565: | ||
743 : | edgomez | 851 | safe_packed_conv( |
744 : | edgomez | 1382 | dst[0], dst_stride[0], image->y, image->u, image->v, |
745 : | edgomez | 851 | edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
746 : | interlacing?yv12_to_rgb565i :yv12_to_rgb565, | ||
747 : | interlacing?yv12_to_rgb565i_c:yv12_to_rgb565_c, 2); | ||
748 : | albeu | 315 | return 0; |
749 : | |||
750 : | edgomez | 1382 | case XVID_CSP_BGR: |
751 : | edgomez | 851 | safe_packed_conv( |
752 : | edgomez | 1382 | dst[0], dst_stride[0], image->y, image->u, image->v, |
753 : | edgomez | 851 | edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
754 : | interlacing?yv12_to_bgri :yv12_to_bgr, | ||
755 : | interlacing?yv12_to_bgri_c:yv12_to_bgr_c, 3); | ||
756 : | albeu | 315 | return 0; |
757 : | |||
758 : | edgomez | 1382 | case XVID_CSP_BGRA: |
759 : | edgomez | 851 | safe_packed_conv( |
760 : | edgomez | 1382 | dst[0], dst_stride[0], image->y, image->u, image->v, |
761 : | edgomez | 851 | edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
762 : | interlacing?yv12_to_bgrai :yv12_to_bgra, | ||
763 : | interlacing?yv12_to_bgrai_c:yv12_to_bgra_c, 4); | ||
764 : | albeu | 315 | return 0; |
765 : | |||
766 : | edgomez | 851 | case XVID_CSP_ABGR: |
767 : | safe_packed_conv( | ||
768 : | edgomez | 1382 | dst[0], dst_stride[0], image->y, image->u, image->v, |
769 : | edgomez | 851 | edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
770 : | interlacing?yv12_to_abgri :yv12_to_abgr, | ||
771 : | interlacing?yv12_to_abgri_c:yv12_to_abgr_c, 4); | ||
772 : | albeu | 315 | return 0; |
773 : | |||
774 : | edgomez | 851 | case XVID_CSP_RGBA: |
775 : | safe_packed_conv( | ||
776 : | edgomez | 1382 | dst[0], dst_stride[0], image->y, image->u, image->v, |
777 : | edgomez | 851 | edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
778 : | interlacing?yv12_to_rgbai :yv12_to_rgba, | ||
779 : | interlacing?yv12_to_rgbai_c:yv12_to_rgba_c, 4); | ||
780 : | albeu | 315 | return 0; |
781 : | |||
782 : | edgomez | 1382 | case XVID_CSP_ARGB: |
783 : | safe_packed_conv( | ||
784 : | dst[0], dst_stride[0], image->y, image->u, image->v, | ||
785 : | edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), | ||
786 : | interlacing?yv12_to_argbi :yv12_to_argb, | ||
787 : | interlacing?yv12_to_argbi_c:yv12_to_argb_c, 4); | ||
788 : | return 0; | ||
789 : | |||
790 : | albeu | 315 | case XVID_CSP_YUY2: |
791 : | edgomez | 851 | safe_packed_conv( |
792 : | edgomez | 1382 | dst[0], dst_stride[0], image->y, image->u, image->v, |
793 : | edgomez | 851 | edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
794 : | interlacing?yv12_to_yuyvi :yv12_to_yuyv, | ||
795 : | interlacing?yv12_to_yuyvi_c:yv12_to_yuyv_c, 2); | ||
796 : | albeu | 315 | return 0; |
797 : | |||
798 : | edgomez | 1382 | case XVID_CSP_YVYU: /* u,v swapped */ |
799 : | edgomez | 851 | safe_packed_conv( |
800 : | edgomez | 1382 | dst[0], dst_stride[0], image->y, image->v, image->u, |
801 : | edgomez | 851 | edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
802 : | interlacing?yv12_to_yuyvi :yv12_to_yuyv, | ||
803 : | interlacing?yv12_to_yuyvi_c:yv12_to_yuyv_c, 2); | ||
804 : | albeu | 315 | return 0; |
805 : | |||
806 : | case XVID_CSP_UYVY: | ||
807 : | edgomez | 851 | safe_packed_conv( |
808 : | edgomez | 1382 | dst[0], dst_stride[0], image->y, image->u, image->v, |
809 : | edgomez | 851 | edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
810 : | interlacing?yv12_to_uyvyi :yv12_to_uyvy, | ||
811 : | interlacing?yv12_to_uyvyi_c:yv12_to_uyvy_c, 2); | ||
812 : | albeu | 315 | return 0; |
813 : | |||
814 : | edgomez | 1382 | case XVID_CSP_I420: /* YCbCr == YUV == internal colorspace for MPEG */ |
815 : | yv12_to_yv12(dst[0], dst[0] + dst_stride[0]*height, dst[0] + dst_stride[0]*height + (dst_stride[0]/2)*height2, | ||
816 : | dst_stride[0], dst_stride[0]/2, | ||
817 : | edgomez | 851 | image->y, image->u, image->v, edged_width, edged_width2, |
818 : | width, height, (csp & XVID_CSP_VFLIP)); | ||
819 : | return 0; | ||
820 : | |||
821 : | edgomez | 1382 | case XVID_CSP_YV12: /* YCrCb == YVU == U and V plane swapped */ |
822 : | yv12_to_yv12(dst[0], dst[0] + dst_stride[0]*height, dst[0] + dst_stride[0]*height + (dst_stride[0]/2)*height2, | ||
823 : | dst_stride[0], dst_stride[0]/2, | ||
824 : | edgomez | 851 | image->y, image->v, image->u, edged_width, edged_width2, |
825 : | width, height, (csp & XVID_CSP_VFLIP)); | ||
826 : | return 0; | ||
827 : | |||
828 : | edgomez | 1382 | case XVID_CSP_PLANAR: /* YCbCr with arbitrary pointers and different strides for Y and UV */ |
829 : | yv12_to_yv12(dst[0], dst[1], dst[2], | ||
830 : | dst_stride[0], dst_stride[1], /* v: dst_stride[2] not yet supported */ | ||
831 : | image->y, image->u, image->v, edged_width, edged_width2, | ||
832 : | width, height, (csp & XVID_CSP_VFLIP)); | ||
833 : | albeu | 315 | return 0; |
834 : | |||
835 : | edgomez | 1382 | case XVID_CSP_INTERNAL : |
836 : | dst[0] = image->y; | ||
837 : | dst[1] = image->u; | ||
838 : | dst[2] = image->v; | ||
839 : | dst_stride[0] = edged_width; | ||
840 : | dst_stride[1] = edged_width/2; | ||
841 : | dst_stride[2] = edged_width/2; | ||
842 : | return 0; | ||
843 : | |||
844 : | albeu | 315 | case XVID_CSP_NULL: |
845 : | edgomez | 1382 | case XVID_CSP_SLICE: |
846 : | albeu | 315 | return 0; |
847 : | |||
848 : | } | ||
849 : | |||
850 : | return -1; | ||
851 : | } | ||
852 : | |||
853 : | float | ||
854 : | image_psnr(IMAGE * orig_image, | ||
855 : | IMAGE * recon_image, | ||
856 : | uint16_t stride, | ||
857 : | uint16_t width, | ||
858 : | uint16_t height) | ||
859 : | { | ||
860 : | int32_t diff, x, y, quad = 0; | ||
861 : | uint8_t *orig = orig_image->y; | ||
862 : | uint8_t *recon = recon_image->y; | ||
863 : | float psnr_y; | ||
864 : | |||
865 : | for (y = 0; y < height; y++) { | ||
866 : | for (x = 0; x < width; x++) { | ||
867 : | diff = *(orig + x) - *(recon + x); | ||
868 : | quad += diff * diff; | ||
869 : | } | ||
870 : | orig += stride; | ||
871 : | recon += stride; | ||
872 : | } | ||
873 : | |||
874 : | psnr_y = (float) quad / (float) (width * height); | ||
875 : | |||
876 : | if (psnr_y) { | ||
877 : | psnr_y = (float) (255 * 255) / psnr_y; | ||
878 : | psnr_y = 10 * (float) log10(psnr_y); | ||
879 : | } else | ||
880 : | psnr_y = (float) 99.99; | ||
881 : | |||
882 : | return psnr_y; | ||
883 : | } | ||
884 : | |||
885 : | |||
886 : | edgomez | 851 | float sse_to_PSNR(long sse, int pixels) |
887 : | { | ||
888 : | if (sse==0) | ||
889 : | return 99.99F; | ||
890 : | |||
891 : | edgomez | 1382 | return 48.131F - 10*(float)log10((float)sse/(float)(pixels)); /* log10(255*255)=4.8131 */ |
892 : | edgomez | 851 | |
893 : | } | ||
894 : | |||
895 : | edgomez | 1424 | long plane_sse(uint8_t *orig, |
896 : | uint8_t *recon, | ||
897 : | uint16_t stride, | ||
898 : | uint16_t width, | ||
899 : | uint16_t height) | ||
900 : | edgomez | 851 | { |
901 : | edgomez | 1424 | int y, bwidth, bheight; |
902 : | long sse = 0; | ||
903 : | edgomez | 851 | |
904 : | edgomez | 1424 | bwidth = width & (~0x07); |
905 : | bheight = height & (~0x07); | ||
906 : | |||
907 : | /* Compute the 8x8 integer part */ | ||
908 : | for (y = 0; y<bheight; y += 8) { | ||
909 : | int x; | ||
910 : | |||
911 : | /* Compute sse for the band */ | ||
912 : | for (x = 0; x<bwidth; x += 8) | ||
913 : | sse += sse8_8bit(orig + x, recon + x, stride); | ||
914 : | |||
915 : | /* remaining pixels of the 8 pixels high band */ | ||
916 : | for (x = bwidth; x < width; x++) { | ||
917 : | int diff; | ||
918 : | diff = *(orig + 0*stride + x) - *(recon + 0*stride + x); | ||
919 : | sse += diff * diff; | ||
920 : | diff = *(orig + 1*stride + x) - *(recon + 1*stride + x); | ||
921 : | sse += diff * diff; | ||
922 : | diff = *(orig + 2*stride + x) - *(recon + 2*stride + x); | ||
923 : | sse += diff * diff; | ||
924 : | diff = *(orig + 3*stride + x) - *(recon + 3*stride + x); | ||
925 : | sse += diff * diff; | ||
926 : | diff = *(orig + 4*stride + x) - *(recon + 4*stride + x); | ||
927 : | sse += diff * diff; | ||
928 : | diff = *(orig + 5*stride + x) - *(recon + 5*stride + x); | ||
929 : | sse += diff * diff; | ||
930 : | diff = *(orig + 6*stride + x) - *(recon + 6*stride + x); | ||
931 : | sse += diff * diff; | ||
932 : | diff = *(orig + 7*stride + x) - *(recon + 7*stride + x); | ||
933 : | sse += diff * diff; | ||
934 : | } | ||
935 : | |||
936 : | orig += 8*stride; | ||
937 : | recon += 8*stride; | ||
938 : | } | ||
939 : | |||
940 : | /* Compute the down rectangle sse */ | ||
941 : | for (y = bheight; y < height; y++) { | ||
942 : | int x; | ||
943 : | edgomez | 851 | for (x = 0; x < width; x++) { |
944 : | edgomez | 1424 | int diff; |
945 : | edgomez | 851 | diff = *(orig + x) - *(recon + x); |
946 : | sse += diff * diff; | ||
947 : | } | ||
948 : | orig += stride; | ||
949 : | recon += stride; | ||
950 : | } | ||
951 : | edgomez | 1424 | |
952 : | return (sse); | ||
953 : | edgomez | 851 | } |
954 : | |||
955 : | edgomez | 1382 | #if 0 |
956 : | edgomez | 851 | |
957 : | albeu | 315 | #include <stdio.h> |
958 : | #include <string.h> | ||
959 : | |||
960 : | int image_dump_pgm(uint8_t * bmp, uint32_t width, uint32_t height, char * filename) | ||
961 : | { | ||
962 : | FILE * f; | ||
963 : | char hdr[1024]; | ||
964 : | edgomez | 1382 | |
965 : | albeu | 315 | f = fopen(filename, "wb"); |
966 : | if ( f == NULL) | ||
967 : | { | ||
968 : | return -1; | ||
969 : | } | ||
970 : | sprintf(hdr, "P5\n#xvid\n%i %i\n255\n", width, height); | ||
971 : | fwrite(hdr, strlen(hdr), 1, f); | ||
972 : | fwrite(bmp, width, height, f); | ||
973 : | fclose(f); | ||
974 : | |||
975 : | return 0; | ||
976 : | } | ||
977 : | |||
978 : | |||
979 : | edgomez | 1382 | /* dump image+edges to yuv pgm files */ |
980 : | albeu | 315 | |
981 : | int image_dump(IMAGE * image, uint32_t edged_width, uint32_t edged_height, char * path, int number) | ||
982 : | { | ||
983 : | char filename[1024]; | ||
984 : | |||
985 : | sprintf(filename, "%s_%i_%c.pgm", path, number, 'y'); | ||
986 : | image_dump_pgm( | ||
987 : | image->y - (EDGE_SIZE * edged_width + EDGE_SIZE), | ||
988 : | edged_width, edged_height, filename); | ||
989 : | |||
990 : | sprintf(filename, "%s_%i_%c.pgm", path, number, 'u'); | ||
991 : | image_dump_pgm( | ||
992 : | image->u - (EDGE_SIZE2 * edged_width / 2 + EDGE_SIZE2), | ||
993 : | edged_width / 2, edged_height / 2, filename); | ||
994 : | |||
995 : | sprintf(filename, "%s_%i_%c.pgm", path, number, 'v'); | ||
996 : | image_dump_pgm( | ||
997 : | image->v - (EDGE_SIZE2 * edged_width / 2 + EDGE_SIZE2), | ||
998 : | edged_width / 2, edged_height / 2, filename); | ||
999 : | |||
1000 : | return 0; | ||
1001 : | } | ||
1002 : | edgomez | 1382 | #endif |
1003 : | albeu | 315 | |
1004 : | |||
1005 : | |||
1006 : | /* dump image to yuvpgm file */ | ||
1007 : | |||
1008 : | #include <stdio.h> | ||
1009 : | |||
1010 : | int | ||
1011 : | image_dump_yuvpgm(const IMAGE * image, | ||
1012 : | const uint32_t edged_width, | ||
1013 : | const uint32_t width, | ||
1014 : | const uint32_t height, | ||
1015 : | char *filename) | ||
1016 : | { | ||
1017 : | FILE *f; | ||
1018 : | char hdr[1024]; | ||
1019 : | uint32_t i; | ||
1020 : | uint8_t *bmp1; | ||
1021 : | uint8_t *bmp2; | ||
1022 : | |||
1023 : | |||
1024 : | f = fopen(filename, "wb"); | ||
1025 : | if (f == NULL) { | ||
1026 : | return -1; | ||
1027 : | } | ||
1028 : | sprintf(hdr, "P5\n#xvid\n%i %i\n255\n", width, (3 * height) / 2); | ||
1029 : | fwrite(hdr, strlen(hdr), 1, f); | ||
1030 : | |||
1031 : | bmp1 = image->y; | ||
1032 : | for (i = 0; i < height; i++) { | ||
1033 : | fwrite(bmp1, width, 1, f); | ||
1034 : | bmp1 += edged_width; | ||
1035 : | } | ||
1036 : | |||
1037 : | bmp1 = image->u; | ||
1038 : | bmp2 = image->v; | ||
1039 : | for (i = 0; i < height / 2; i++) { | ||
1040 : | fwrite(bmp1, width / 2, 1, f); | ||
1041 : | fwrite(bmp2, width / 2, 1, f); | ||
1042 : | bmp1 += edged_width / 2; | ||
1043 : | bmp2 += edged_width / 2; | ||
1044 : | } | ||
1045 : | |||
1046 : | fclose(f); | ||
1047 : | return 0; | ||
1048 : | } | ||
1049 : | |||
1050 : | |||
1051 : | float | ||
1052 : | image_mad(const IMAGE * img1, | ||
1053 : | const IMAGE * img2, | ||
1054 : | uint32_t stride, | ||
1055 : | uint32_t width, | ||
1056 : | uint32_t height) | ||
1057 : | { | ||
1058 : | const uint32_t stride2 = stride / 2; | ||
1059 : | const uint32_t width2 = width / 2; | ||
1060 : | const uint32_t height2 = height / 2; | ||
1061 : | |||
1062 : | uint32_t x, y; | ||
1063 : | uint32_t sum = 0; | ||
1064 : | |||
1065 : | for (y = 0; y < height; y++) | ||
1066 : | for (x = 0; x < width; x++) | ||
1067 : | edgomez | 1382 | sum += abs(img1->y[x + y * stride] - img2->y[x + y * stride]); |
1068 : | albeu | 315 | |
1069 : | for (y = 0; y < height2; y++) | ||
1070 : | for (x = 0; x < width2; x++) | ||
1071 : | edgomez | 1382 | sum += abs(img1->u[x + y * stride2] - img2->u[x + y * stride2]); |
1072 : | albeu | 315 | |
1073 : | for (y = 0; y < height2; y++) | ||
1074 : | for (x = 0; x < width2; x++) | ||
1075 : | edgomez | 1382 | sum += abs(img1->v[x + y * stride2] - img2->v[x + y * stride2]); |
1076 : | albeu | 315 | |
1077 : | return (float) sum / (width * height * 3 / 2); | ||
1078 : | } | ||
1079 : | |||
1080 : | void | ||
1081 : | edgomez | 1382 | output_slice(IMAGE * cur, int std, int width, xvid_image_t* out_frm, int mbx, int mby,int mbl) { |
1082 : | albeu | 315 | uint8_t *dY,*dU,*dV,*sY,*sU,*sV; |
1083 : | int std2 = std >> 1; | ||
1084 : | int w = mbl << 4, w2,i; | ||
1085 : | |||
1086 : | if(w > width) | ||
1087 : | w = width; | ||
1088 : | w2 = w >> 1; | ||
1089 : | suxen_drol | 323 | |
1090 : | edgomez | 1382 | dY = (uint8_t*)out_frm->plane[0] + (mby << 4) * out_frm->stride[0] + (mbx << 4); |
1091 : | dU = (uint8_t*)out_frm->plane[1] + (mby << 3) * out_frm->stride[1] + (mbx << 3); | ||
1092 : | dV = (uint8_t*)out_frm->plane[2] + (mby << 3) * out_frm->stride[2] + (mbx << 3); | ||
1093 : | albeu | 315 | sY = cur->y + (mby << 4) * std + (mbx << 4); |
1094 : | sU = cur->u + (mby << 3) * std2 + (mbx << 3); | ||
1095 : | sV = cur->v + (mby << 3) * std2 + (mbx << 3); | ||
1096 : | |||
1097 : | for(i = 0 ; i < 16 ; i++) { | ||
1098 : | memcpy(dY,sY,w); | ||
1099 : | edgomez | 1382 | dY += out_frm->stride[0]; |
1100 : | suxen_drol | 323 | sY += std; |
1101 : | } | ||
1102 : | albeu | 315 | for(i = 0 ; i < 8 ; i++) { |
1103 : | memcpy(dU,sU,w2); | ||
1104 : | edgomez | 1382 | dU += out_frm->stride[1]; |
1105 : | albeu | 315 | sU += std2; |
1106 : | } | ||
1107 : | for(i = 0 ; i < 8 ; i++) { | ||
1108 : | memcpy(dV,sV,w2); | ||
1109 : | edgomez | 1382 | dV += out_frm->stride[2]; |
1110 : | albeu | 315 | sV += std2; |
1111 : | } | ||
1112 : | } | ||
1113 : | edgomez | 851 | |
1114 : | |||
1115 : | void | ||
1116 : | image_clear(IMAGE * img, int width, int height, int edged_width, | ||
1117 : | int y, int u, int v) | ||
1118 : | { | ||
1119 : | uint8_t * p; | ||
1120 : | int i; | ||
1121 : | |||
1122 : | p = img->y; | ||
1123 : | for (i = 0; i < height; i++) { | ||
1124 : | memset(p, y, width); | ||
1125 : | p += edged_width; | ||
1126 : | } | ||
1127 : | |||
1128 : | p = img->u; | ||
1129 : | for (i = 0; i < height/2; i++) { | ||
1130 : | memset(p, u, width/2); | ||
1131 : | p += edged_width/2; | ||
1132 : | } | ||
1133 : | |||
1134 : | p = img->v; | ||
1135 : | for (i = 0; i < height/2; i++) { | ||
1136 : | memset(p, v, width/2); | ||
1137 : | p += edged_width/2; | ||
1138 : | } | ||
1139 : | } |
No admin address has been configured | ViewVC Help |
Powered by ViewVC 1.0.4 |