Parent Directory | Revision Log
Revision 1856 - (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 : | Isibaar | 1856 | * $Id: image.c,v 1.44 2009-03-30 14:40:05 Isibaar 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 : | Skal | 1733 | #include "../utils/emms.h" |
38 : | albeu | 315 | |
39 : | edgomez | 1382 | #include "font.h" |
40 : | edgomez | 851 | |
41 : | albeu | 315 | #define SAFETY 64 |
42 : | #define EDGE_SIZE2 (EDGE_SIZE/2) | ||
43 : | |||
44 : | |||
45 : | int32_t | ||
46 : | image_create(IMAGE * image, | ||
47 : | uint32_t edged_width, | ||
48 : | uint32_t edged_height) | ||
49 : | { | ||
50 : | const uint32_t edged_width2 = edged_width / 2; | ||
51 : | const uint32_t edged_height2 = edged_height / 2; | ||
52 : | |||
53 : | image->y = | ||
54 : | xvid_malloc(edged_width * (edged_height + 1) + SAFETY, CACHE_LINE); | ||
55 : | if (image->y == NULL) { | ||
56 : | return -1; | ||
57 : | } | ||
58 : | edgomez | 1382 | memset(image->y, 0, edged_width * (edged_height + 1) + SAFETY); |
59 : | albeu | 315 | |
60 : | image->u = xvid_malloc(edged_width2 * edged_height2 + SAFETY, CACHE_LINE); | ||
61 : | if (image->u == NULL) { | ||
62 : | xvid_free(image->y); | ||
63 : | edgomez | 1382 | image->y = NULL; |
64 : | albeu | 315 | return -1; |
65 : | } | ||
66 : | edgomez | 1382 | memset(image->u, 0, edged_width2 * edged_height2 + SAFETY); |
67 : | |||
68 : | albeu | 315 | image->v = xvid_malloc(edged_width2 * edged_height2 + SAFETY, CACHE_LINE); |
69 : | if (image->v == NULL) { | ||
70 : | xvid_free(image->u); | ||
71 : | edgomez | 1382 | image->u = NULL; |
72 : | albeu | 315 | xvid_free(image->y); |
73 : | edgomez | 1382 | image->y = NULL; |
74 : | albeu | 315 | return -1; |
75 : | } | ||
76 : | edgomez | 1382 | memset(image->v, 0, edged_width2 * edged_height2 + SAFETY); |
77 : | albeu | 315 | |
78 : | image->y += EDGE_SIZE * edged_width + EDGE_SIZE; | ||
79 : | image->u += EDGE_SIZE2 * edged_width2 + EDGE_SIZE2; | ||
80 : | image->v += EDGE_SIZE2 * edged_width2 + EDGE_SIZE2; | ||
81 : | |||
82 : | return 0; | ||
83 : | } | ||
84 : | |||
85 : | |||
86 : | |||
87 : | void | ||
88 : | image_destroy(IMAGE * image, | ||
89 : | uint32_t edged_width, | ||
90 : | uint32_t edged_height) | ||
91 : | { | ||
92 : | const uint32_t edged_width2 = edged_width / 2; | ||
93 : | |||
94 : | if (image->y) { | ||
95 : | xvid_free(image->y - (EDGE_SIZE * edged_width + EDGE_SIZE)); | ||
96 : | edgomez | 1382 | image->y = NULL; |
97 : | albeu | 315 | } |
98 : | if (image->u) { | ||
99 : | xvid_free(image->u - (EDGE_SIZE2 * edged_width2 + EDGE_SIZE2)); | ||
100 : | edgomez | 1382 | image->u = NULL; |
101 : | albeu | 315 | } |
102 : | if (image->v) { | ||
103 : | xvid_free(image->v - (EDGE_SIZE2 * edged_width2 + EDGE_SIZE2)); | ||
104 : | edgomez | 1382 | image->v = NULL; |
105 : | albeu | 315 | } |
106 : | } | ||
107 : | |||
108 : | |||
109 : | void | ||
110 : | image_swap(IMAGE * image1, | ||
111 : | IMAGE * image2) | ||
112 : | { | ||
113 : | edgomez | 1382 | SWAP(uint8_t*, image1->y, image2->y); |
114 : | SWAP(uint8_t*, image1->u, image2->u); | ||
115 : | SWAP(uint8_t*, image1->v, image2->v); | ||
116 : | albeu | 315 | } |
117 : | |||
118 : | |||
119 : | void | ||
120 : | image_copy(IMAGE * image1, | ||
121 : | IMAGE * image2, | ||
122 : | uint32_t edged_width, | ||
123 : | uint32_t height) | ||
124 : | { | ||
125 : | memcpy(image1->y, image2->y, edged_width * height); | ||
126 : | memcpy(image1->u, image2->u, edged_width * height / 4); | ||
127 : | memcpy(image1->v, image2->v, edged_width * height / 4); | ||
128 : | } | ||
129 : | |||
130 : | edgomez | 1382 | /* setedges bug was fixed in this BS version */ |
131 : | #define SETEDGES_BUG_BEFORE 18 | ||
132 : | albeu | 315 | |
133 : | void | ||
134 : | image_setedges(IMAGE * image, | ||
135 : | uint32_t edged_width, | ||
136 : | uint32_t edged_height, | ||
137 : | uint32_t width, | ||
138 : | edgomez | 1382 | uint32_t height, |
139 : | int bs_version) | ||
140 : | albeu | 315 | { |
141 : | const uint32_t edged_width2 = edged_width / 2; | ||
142 : | edgomez | 1382 | uint32_t width2; |
143 : | albeu | 315 | uint32_t i; |
144 : | uint8_t *dst; | ||
145 : | uint8_t *src; | ||
146 : | |||
147 : | dst = image->y - (EDGE_SIZE + EDGE_SIZE * edged_width); | ||
148 : | src = image->y; | ||
149 : | |||
150 : | edgomez | 1382 | /* According to the Standard Clause 7.6.4, padding is done starting at 16 |
151 : | * pixel width and height multiples. This was not respected in old xvids */ | ||
152 : | if (bs_version == 0 || bs_version >= SETEDGES_BUG_BEFORE) { | ||
153 : | width = (width+15)&~15; | ||
154 : | height = (height+15)&~15; | ||
155 : | } | ||
156 : | |||
157 : | width2 = width/2; | ||
158 : | |||
159 : | albeu | 315 | for (i = 0; i < EDGE_SIZE; i++) { |
160 : | edgomez | 851 | memset(dst, *src, EDGE_SIZE); |
161 : | memcpy(dst + EDGE_SIZE, src, width); | ||
162 : | memset(dst + edged_width - EDGE_SIZE, *(src + width - 1), | ||
163 : | EDGE_SIZE); | ||
164 : | albeu | 315 | dst += edged_width; |
165 : | } | ||
166 : | |||
167 : | for (i = 0; i < height; i++) { | ||
168 : | memset(dst, *src, EDGE_SIZE); | ||
169 : | memset(dst + edged_width - EDGE_SIZE, src[width - 1], EDGE_SIZE); | ||
170 : | dst += edged_width; | ||
171 : | src += edged_width; | ||
172 : | } | ||
173 : | |||
174 : | src -= edged_width; | ||
175 : | for (i = 0; i < EDGE_SIZE; i++) { | ||
176 : | edgomez | 851 | memset(dst, *src, EDGE_SIZE); |
177 : | memcpy(dst + EDGE_SIZE, src, width); | ||
178 : | memset(dst + edged_width - EDGE_SIZE, *(src + width - 1), | ||
179 : | albeu | 315 | EDGE_SIZE); |
180 : | dst += edged_width; | ||
181 : | } | ||
182 : | |||
183 : | |||
184 : | edgomez | 1382 | /* U */ |
185 : | albeu | 315 | dst = image->u - (EDGE_SIZE2 + EDGE_SIZE2 * edged_width2); |
186 : | src = image->u; | ||
187 : | |||
188 : | for (i = 0; i < EDGE_SIZE2; i++) { | ||
189 : | memset(dst, *src, EDGE_SIZE2); | ||
190 : | memcpy(dst + EDGE_SIZE2, src, width2); | ||
191 : | memset(dst + edged_width2 - EDGE_SIZE2, *(src + width2 - 1), | ||
192 : | EDGE_SIZE2); | ||
193 : | dst += edged_width2; | ||
194 : | } | ||
195 : | |||
196 : | for (i = 0; i < height / 2; i++) { | ||
197 : | memset(dst, *src, EDGE_SIZE2); | ||
198 : | memset(dst + edged_width2 - EDGE_SIZE2, src[width2 - 1], EDGE_SIZE2); | ||
199 : | dst += edged_width2; | ||
200 : | src += edged_width2; | ||
201 : | } | ||
202 : | src -= edged_width2; | ||
203 : | for (i = 0; i < EDGE_SIZE2; i++) { | ||
204 : | memset(dst, *src, EDGE_SIZE2); | ||
205 : | memcpy(dst + EDGE_SIZE2, src, width2); | ||
206 : | memset(dst + edged_width2 - EDGE_SIZE2, *(src + width2 - 1), | ||
207 : | EDGE_SIZE2); | ||
208 : | dst += edged_width2; | ||
209 : | } | ||
210 : | |||
211 : | |||
212 : | edgomez | 1382 | /* V */ |
213 : | albeu | 315 | dst = image->v - (EDGE_SIZE2 + EDGE_SIZE2 * edged_width2); |
214 : | src = image->v; | ||
215 : | |||
216 : | for (i = 0; i < EDGE_SIZE2; i++) { | ||
217 : | memset(dst, *src, EDGE_SIZE2); | ||
218 : | memcpy(dst + EDGE_SIZE2, src, width2); | ||
219 : | memset(dst + edged_width2 - EDGE_SIZE2, *(src + width2 - 1), | ||
220 : | EDGE_SIZE2); | ||
221 : | dst += edged_width2; | ||
222 : | } | ||
223 : | |||
224 : | for (i = 0; i < height / 2; i++) { | ||
225 : | memset(dst, *src, EDGE_SIZE2); | ||
226 : | memset(dst + edged_width2 - EDGE_SIZE2, src[width2 - 1], EDGE_SIZE2); | ||
227 : | dst += edged_width2; | ||
228 : | src += edged_width2; | ||
229 : | } | ||
230 : | src -= edged_width2; | ||
231 : | for (i = 0; i < EDGE_SIZE2; i++) { | ||
232 : | memset(dst, *src, EDGE_SIZE2); | ||
233 : | memcpy(dst + EDGE_SIZE2, src, width2); | ||
234 : | memset(dst + edged_width2 - EDGE_SIZE2, *(src + width2 - 1), | ||
235 : | EDGE_SIZE2); | ||
236 : | dst += edged_width2; | ||
237 : | } | ||
238 : | } | ||
239 : | |||
240 : | void | ||
241 : | syskin | 1665 | image_interpolate(const uint8_t * refn, |
242 : | uint8_t * refh, | ||
243 : | uint8_t * refv, | ||
244 : | uint8_t * refhv, | ||
245 : | albeu | 315 | 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 : | syskin | 1664 | |
253 : | syskin | 1665 | uint8_t *n_ptr; |
254 : | uint8_t *h_ptr, *v_ptr, *hv_ptr; | ||
255 : | albeu | 315 | uint32_t x, y; |
256 : | |||
257 : | syskin | 1665 | n_ptr = (uint8_t*)refn; |
258 : | h_ptr = refh; | ||
259 : | syskin | 1666 | v_ptr = refv; |
260 : | albeu | 315 | |
261 : | n_ptr -= offset; | ||
262 : | h_ptr -= offset; | ||
263 : | v_ptr -= offset; | ||
264 : | |||
265 : | edgomez | 1382 | /* Note we initialize the hv pointer later, as we can optimize code a bit |
266 : | * doing it down to up in quarterpel and up to down in halfpel */ | ||
267 : | edgomez | 851 | if(quarterpel) { |
268 : | edgomez | 1382 | |
269 : | edgomez | 851 | for (y = 0; y < (edged_height - EDGE_SIZE); y += 8) { |
270 : | for (x = 0; x < (edged_width - EDGE_SIZE); x += 8) { | ||
271 : | interpolate8x8_6tap_lowpass_h(h_ptr, n_ptr, edged_width, rounding); | ||
272 : | interpolate8x8_6tap_lowpass_v(v_ptr, n_ptr, edged_width, rounding); | ||
273 : | albeu | 315 | |
274 : | edgomez | 851 | n_ptr += 8; |
275 : | h_ptr += 8; | ||
276 : | v_ptr += 8; | ||
277 : | } | ||
278 : | edgomez | 1382 | |
279 : | edgomez | 851 | n_ptr += EDGE_SIZE; |
280 : | h_ptr += EDGE_SIZE; | ||
281 : | v_ptr += EDGE_SIZE; | ||
282 : | |||
283 : | h_ptr += stride_add; | ||
284 : | v_ptr += stride_add; | ||
285 : | n_ptr += stride_add; | ||
286 : | } | ||
287 : | |||
288 : | syskin | 1665 | h_ptr = refh + (edged_height - EDGE_SIZE - EDGE_SIZE2)*edged_width - EDGE_SIZE2; |
289 : | hv_ptr = refhv + (edged_height - EDGE_SIZE - EDGE_SIZE2)*edged_width - EDGE_SIZE2; | ||
290 : | edgomez | 851 | |
291 : | for (y = 0; y < (edged_height - EDGE_SIZE); y = y + 8) { | ||
292 : | edgomez | 1382 | hv_ptr -= stride_add; |
293 : | h_ptr -= stride_add; | ||
294 : | hv_ptr -= EDGE_SIZE; | ||
295 : | h_ptr -= EDGE_SIZE; | ||
296 : | |||
297 : | edgomez | 851 | for (x = 0; x < (edged_width - EDGE_SIZE); x = x + 8) { |
298 : | edgomez | 1382 | hv_ptr -= 8; |
299 : | h_ptr -= 8; | ||
300 : | edgomez | 851 | interpolate8x8_6tap_lowpass_v(hv_ptr, h_ptr, edged_width, rounding); |
301 : | } | ||
302 : | } | ||
303 : | edgomez | 1382 | } else { |
304 : | edgomez | 851 | |
305 : | syskin | 1665 | hv_ptr = refhv; |
306 : | edgomez | 1382 | hv_ptr -= offset; |
307 : | |||
308 : | edgomez | 851 | for (y = 0; y < (edged_height - EDGE_SIZE); y += 8) { |
309 : | for (x = 0; x < (edged_width - EDGE_SIZE); x += 8) { | ||
310 : | interpolate8x8_halfpel_h(h_ptr, n_ptr, edged_width, rounding); | ||
311 : | interpolate8x8_halfpel_v(v_ptr, n_ptr, edged_width, rounding); | ||
312 : | interpolate8x8_halfpel_hv(hv_ptr, n_ptr, edged_width, rounding); | ||
313 : | |||
314 : | n_ptr += 8; | ||
315 : | h_ptr += 8; | ||
316 : | v_ptr += 8; | ||
317 : | hv_ptr += 8; | ||
318 : | } | ||
319 : | edgomez | 1382 | |
320 : | edgomez | 851 | h_ptr += EDGE_SIZE; |
321 : | v_ptr += EDGE_SIZE; | ||
322 : | hv_ptr += EDGE_SIZE; | ||
323 : | n_ptr += EDGE_SIZE; | ||
324 : | |||
325 : | h_ptr += stride_add; | ||
326 : | v_ptr += stride_add; | ||
327 : | hv_ptr += stride_add; | ||
328 : | n_ptr += stride_add; | ||
329 : | } | ||
330 : | } | ||
331 : | albeu | 315 | } |
332 : | |||
333 : | |||
334 : | edgomez | 851 | /* |
335 : | chroma optimize filter, invented by mf | ||
336 : | a chroma pixel is average from the surrounding pixels, when the | ||
337 : | correpsonding luma pixels are pure black or white. | ||
338 : | */ | ||
339 : | |||
340 : | void | ||
341 : | image_chroma_optimize(IMAGE * img, int width, int height, int edged_width) | ||
342 : | { | ||
343 : | int x,y; | ||
344 : | int pixels = 0; | ||
345 : | |||
346 : | for (y = 1; y < height/2 - 1; y++) | ||
347 : | for (x = 1; x < width/2 - 1; x++) | ||
348 : | { | ||
349 : | #define IS_PURE(a) ((a)<=16||(a)>=235) | ||
350 : | #define IMG_Y(Y,X) img->y[(Y)*edged_width + (X)] | ||
351 : | #define IMG_U(Y,X) img->u[(Y)*edged_width/2 + (X)] | ||
352 : | #define IMG_V(Y,X) img->v[(Y)*edged_width/2 + (X)] | ||
353 : | |||
354 : | edgomez | 1382 | if (IS_PURE(IMG_Y(y*2 ,x*2 )) && |
355 : | edgomez | 851 | IS_PURE(IMG_Y(y*2 ,x*2+1)) && |
356 : | edgomez | 1382 | IS_PURE(IMG_Y(y*2+1,x*2 )) && |
357 : | edgomez | 851 | IS_PURE(IMG_Y(y*2+1,x*2+1))) |
358 : | { | ||
359 : | 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; | ||
360 : | 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; | ||
361 : | pixels++; | ||
362 : | } | ||
363 : | |||
364 : | #undef IS_PURE | ||
365 : | #undef IMG_Y | ||
366 : | #undef IMG_U | ||
367 : | #undef IMG_V | ||
368 : | } | ||
369 : | edgomez | 1382 | |
370 : | DPRINTF(XVID_DEBUG_DEBUG,"chroma_optimized_pixels = %i/%i\n", pixels, width*height/4); | ||
371 : | edgomez | 851 | } |
372 : | |||
373 : | |||
374 : | |||
375 : | |||
376 : | |||
377 : | /* | ||
378 : | perform safe packed colorspace conversion, by splitting | ||
379 : | the image up into an optimized area (pixel width divisible by 16), | ||
380 : | and two unoptimized/plain-c areas (pixel width divisible by 2) | ||
381 : | */ | ||
382 : | |||
383 : | edgomez | 1382 | static void |
384 : | edgomez | 851 | safe_packed_conv(uint8_t * x_ptr, int x_stride, |
385 : | uint8_t * y_ptr, uint8_t * u_ptr, uint8_t * v_ptr, | ||
386 : | int y_stride, int uv_stride, | ||
387 : | int width, int height, int vflip, | ||
388 : | Isibaar | 1815 | packedFunc * func_opt, packedFunc func_c, |
389 : | int size, int interlacing) | ||
390 : | edgomez | 851 | { |
391 : | Isibaar | 1815 | int width_opt, width_c, height_opt; |
392 : | edgomez | 851 | |
393 : | Isibaar | 1856 | if (width<0 || width==1 || height==1) return; /* forget about it */ |
394 : | Isibaar | 1815 | |
395 : | edgomez | 851 | if (func_opt != func_c && x_stride < size*((width+15)/16)*16) |
396 : | { | ||
397 : | width_opt = width & (~15); | ||
398 : | Isibaar | 1815 | width_c = (width - width_opt) & (~1); |
399 : | edgomez | 851 | } |
400 : | Isibaar | 1856 | else if (func_opt != func_c && !(width&1) && (size==3)) |
401 : | edgomez | 851 | { |
402 : | Isibaar | 1856 | /* MMX reads 4 bytes per pixel for RGB/BGR */ |
403 : | width_opt = width - 2; | ||
404 : | width_c = 2; | ||
405 : | } | ||
406 : | else { | ||
407 : | Isibaar | 1815 | /* Enforce the width to be divisable by two. */ |
408 : | width_opt = width & (~1); | ||
409 : | edgomez | 851 | width_c = 0; |
410 : | } | ||
411 : | |||
412 : | Isibaar | 1815 | /* packed conversions require height to be divisable by 2 |
413 : | (or even by 4 for interlaced conversion) */ | ||
414 : | if (interlacing) | ||
415 : | height_opt = height & (~3); | ||
416 : | else | ||
417 : | height_opt = height & (~1); | ||
418 : | |||
419 : | edgomez | 851 | func_opt(x_ptr, x_stride, |
420 : | y_ptr, u_ptr, v_ptr, y_stride, uv_stride, | ||
421 : | Isibaar | 1815 | width_opt, height_opt, vflip); |
422 : | edgomez | 851 | |
423 : | if (width_c) | ||
424 : | { | ||
425 : | func_c(x_ptr + size*width_opt, x_stride, | ||
426 : | y_ptr + width_opt, u_ptr + width_opt/2, v_ptr + width_opt/2, | ||
427 : | Isibaar | 1815 | y_stride, uv_stride, width_c, height_opt, vflip); |
428 : | edgomez | 851 | } |
429 : | } | ||
430 : | |||
431 : | |||
432 : | |||
433 : | albeu | 315 | int |
434 : | image_input(IMAGE * image, | ||
435 : | uint32_t width, | ||
436 : | int height, | ||
437 : | uint32_t edged_width, | ||
438 : | edgomez | 1382 | uint8_t * src[4], |
439 : | int src_stride[4], | ||
440 : | edgomez | 851 | int csp, |
441 : | int interlacing) | ||
442 : | albeu | 315 | { |
443 : | edgomez | 851 | const int edged_width2 = edged_width/2; |
444 : | const int width2 = width/2; | ||
445 : | const int height2 = height/2; | ||
446 : | edgomez | 1382 | #if 0 |
447 : | const int height_signed = (csp & XVID_CSP_VFLIP) ? -height : height; | ||
448 : | #endif | ||
449 : | albeu | 315 | |
450 : | switch (csp & ~XVID_CSP_VFLIP) { | ||
451 : | case XVID_CSP_RGB555: | ||
452 : | edgomez | 851 | safe_packed_conv( |
453 : | edgomez | 1382 | src[0], src_stride[0], image->y, image->u, image->v, |
454 : | edgomez | 851 | edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
455 : | interlacing?rgb555i_to_yv12 :rgb555_to_yv12, | ||
456 : | Isibaar | 1815 | interlacing?rgb555i_to_yv12_c:rgb555_to_yv12_c, 2, interlacing); |
457 : | edgomez | 851 | break; |
458 : | albeu | 315 | |
459 : | case XVID_CSP_RGB565: | ||
460 : | edgomez | 851 | safe_packed_conv( |
461 : | edgomez | 1382 | src[0], src_stride[0], image->y, image->u, image->v, |
462 : | edgomez | 851 | edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
463 : | interlacing?rgb565i_to_yv12 :rgb565_to_yv12, | ||
464 : | Isibaar | 1815 | interlacing?rgb565i_to_yv12_c:rgb565_to_yv12_c, 2, interlacing); |
465 : | edgomez | 851 | break; |
466 : | albeu | 315 | |
467 : | |||
468 : | edgomez | 1382 | case XVID_CSP_BGR: |
469 : | edgomez | 851 | safe_packed_conv( |
470 : | edgomez | 1382 | src[0], src_stride[0], image->y, image->u, image->v, |
471 : | edgomez | 851 | edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
472 : | interlacing?bgri_to_yv12 :bgr_to_yv12, | ||
473 : | Isibaar | 1815 | interlacing?bgri_to_yv12_c:bgr_to_yv12_c, 3, interlacing); |
474 : | edgomez | 851 | break; |
475 : | albeu | 315 | |
476 : | edgomez | 1382 | case XVID_CSP_BGRA: |
477 : | edgomez | 851 | safe_packed_conv( |
478 : | edgomez | 1382 | src[0], src_stride[0], image->y, image->u, image->v, |
479 : | edgomez | 851 | edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
480 : | interlacing?bgrai_to_yv12 :bgra_to_yv12, | ||
481 : | Isibaar | 1815 | interlacing?bgrai_to_yv12_c:bgra_to_yv12_c, 4, interlacing); |
482 : | edgomez | 851 | break; |
483 : | albeu | 315 | |
484 : | edgomez | 851 | case XVID_CSP_ABGR : |
485 : | safe_packed_conv( | ||
486 : | edgomez | 1382 | src[0], src_stride[0], image->y, image->u, image->v, |
487 : | edgomez | 851 | edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
488 : | interlacing?abgri_to_yv12 :abgr_to_yv12, | ||
489 : | Isibaar | 1815 | interlacing?abgri_to_yv12_c:abgr_to_yv12_c, 4, interlacing); |
490 : | edgomez | 851 | break; |
491 : | albeu | 315 | |
492 : | chl | 1759 | case XVID_CSP_RGB: |
493 : | safe_packed_conv( | ||
494 : | src[0], src_stride[0], image->y, image->u, image->v, | ||
495 : | edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), | ||
496 : | interlacing?rgbi_to_yv12 :rgb_to_yv12, | ||
497 : | Isibaar | 1815 | interlacing?rgbi_to_yv12_c:rgb_to_yv12_c, 3, interlacing); |
498 : | chl | 1759 | break; |
499 : | |||
500 : | edgomez | 851 | case XVID_CSP_RGBA : |
501 : | safe_packed_conv( | ||
502 : | edgomez | 1382 | src[0], src_stride[0], image->y, image->u, image->v, |
503 : | edgomez | 851 | edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
504 : | interlacing?rgbai_to_yv12 :rgba_to_yv12, | ||
505 : | Isibaar | 1815 | interlacing?rgbai_to_yv12_c:rgba_to_yv12_c, 4, interlacing); |
506 : | edgomez | 851 | break; |
507 : | edgomez | 1382 | |
508 : | case XVID_CSP_ARGB: | ||
509 : | safe_packed_conv( | ||
510 : | src[0], src_stride[0], image->y, image->u, image->v, | ||
511 : | edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), | ||
512 : | interlacing?argbi_to_yv12 : argb_to_yv12, | ||
513 : | Isibaar | 1815 | interlacing?argbi_to_yv12_c: argb_to_yv12_c, 4, interlacing); |
514 : | edgomez | 1382 | break; |
515 : | albeu | 315 | |
516 : | case XVID_CSP_YUY2: | ||
517 : | edgomez | 851 | safe_packed_conv( |
518 : | edgomez | 1382 | src[0], src_stride[0], image->y, image->u, image->v, |
519 : | edgomez | 851 | edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
520 : | interlacing?yuyvi_to_yv12 :yuyv_to_yv12, | ||
521 : | Isibaar | 1815 | interlacing?yuyvi_to_yv12_c:yuyv_to_yv12_c, 2, interlacing); |
522 : | edgomez | 851 | break; |
523 : | albeu | 315 | |
524 : | case XVID_CSP_YVYU: /* u/v swapped */ | ||
525 : | edgomez | 851 | safe_packed_conv( |
526 : | edgomez | 1382 | src[0], src_stride[0], image->y, image->v, image->u, |
527 : | edgomez | 851 | edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
528 : | interlacing?yuyvi_to_yv12 :yuyv_to_yv12, | ||
529 : | Isibaar | 1815 | interlacing?yuyvi_to_yv12_c:yuyv_to_yv12_c, 2, interlacing); |
530 : | edgomez | 851 | break; |
531 : | albeu | 315 | |
532 : | case XVID_CSP_UYVY: | ||
533 : | edgomez | 851 | safe_packed_conv( |
534 : | edgomez | 1382 | src[0], src_stride[0], image->y, image->u, image->v, |
535 : | edgomez | 851 | edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
536 : | interlacing?uyvyi_to_yv12 :uyvy_to_yv12, | ||
537 : | Isibaar | 1815 | interlacing?uyvyi_to_yv12_c:uyvy_to_yv12_c, 2, interlacing); |
538 : | edgomez | 851 | break; |
539 : | albeu | 315 | |
540 : | edgomez | 1382 | case XVID_CSP_I420: /* YCbCr == YUV == internal colorspace for MPEG */ |
541 : | edgomez | 851 | yv12_to_yv12(image->y, image->u, image->v, edged_width, edged_width2, |
542 : | edgomez | 1382 | src[0], src[0] + src_stride[0]*height, src[0] + src_stride[0]*height + (src_stride[0]/2)*height2, |
543 : | src_stride[0], src_stride[0]/2, width, height, (csp & XVID_CSP_VFLIP)); | ||
544 : | break; | ||
545 : | |||
546 : | case XVID_CSP_YV12: /* YCrCb == YVA == U and V plane swapped */ | ||
547 : | edgomez | 851 | yv12_to_yv12(image->y, image->v, image->u, edged_width, edged_width2, |
548 : | edgomez | 1382 | src[0], src[0] + src_stride[0]*height, src[0] + src_stride[0]*height + (src_stride[0]/2)*height2, |
549 : | src_stride[0], src_stride[0]/2, width, height, (csp & XVID_CSP_VFLIP)); | ||
550 : | edgomez | 851 | break; |
551 : | |||
552 : | edgomez | 1382 | case XVID_CSP_PLANAR: /* YCbCr with arbitrary pointers and different strides for Y and UV */ |
553 : | yv12_to_yv12(image->y, image->u, image->v, edged_width, edged_width2, | ||
554 : | src[0], src[1], src[2], src_stride[0], src_stride[1], /* v: dst_stride[2] not yet supported */ | ||
555 : | width, height, (csp & XVID_CSP_VFLIP)); | ||
556 : | edgomez | 851 | break; |
557 : | albeu | 315 | |
558 : | case XVID_CSP_NULL: | ||
559 : | break; | ||
560 : | |||
561 : | edgomez | 851 | default : |
562 : | return -1; | ||
563 : | albeu | 315 | } |
564 : | |||
565 : | edgomez | 851 | |
566 : | /* pad out image when the width and/or height is not a multiple of 16 */ | ||
567 : | |||
568 : | if (width & 15) | ||
569 : | { | ||
570 : | int i; | ||
571 : | int pad_width = 16 - (width&15); | ||
572 : | for (i = 0; i < height; i++) | ||
573 : | { | ||
574 : | edgomez | 1382 | memset(image->y + i*edged_width + width, |
575 : | edgomez | 851 | *(image->y + i*edged_width + width - 1), pad_width); |
576 : | } | ||
577 : | for (i = 0; i < height/2; i++) | ||
578 : | { | ||
579 : | edgomez | 1382 | memset(image->u + i*edged_width2 + width2, |
580 : | edgomez | 851 | *(image->u + i*edged_width2 + width2 - 1),pad_width/2); |
581 : | edgomez | 1382 | memset(image->v + i*edged_width2 + width2, |
582 : | edgomez | 851 | *(image->v + i*edged_width2 + width2 - 1),pad_width/2); |
583 : | } | ||
584 : | } | ||
585 : | |||
586 : | if (height & 15) | ||
587 : | { | ||
588 : | edgomez | 1382 | int pad_height = 16 - (height&15); |
589 : | edgomez | 851 | int length = ((width+15)/16)*16; |
590 : | int i; | ||
591 : | for (i = 0; i < pad_height; i++) | ||
592 : | { | ||
593 : | memcpy(image->y + (height+i)*edged_width, | ||
594 : | image->y + (height-1)*edged_width,length); | ||
595 : | } | ||
596 : | |||
597 : | for (i = 0; i < pad_height/2; i++) | ||
598 : | { | ||
599 : | memcpy(image->u + (height2+i)*edged_width2, | ||
600 : | image->u + (height2-1)*edged_width2,length/2); | ||
601 : | memcpy(image->v + (height2+i)*edged_width2, | ||
602 : | image->v + (height2-1)*edged_width2,length/2); | ||
603 : | } | ||
604 : | } | ||
605 : | |||
606 : | /* | ||
607 : | if (interlacing) | ||
608 : | image_printf(image, edged_width, height, 5,5, "[i]"); | ||
609 : | image_dump_yuvpgm(image, edged_width, ((width+15)/16)*16, ((height+15)/16)*16, "\\encode.pgm"); | ||
610 : | */ | ||
611 : | return 0; | ||
612 : | albeu | 315 | } |
613 : | |||
614 : | |||
615 : | |||
616 : | int | ||
617 : | image_output(IMAGE * image, | ||
618 : | uint32_t width, | ||
619 : | int height, | ||
620 : | uint32_t edged_width, | ||
621 : | edgomez | 1382 | uint8_t * dst[4], |
622 : | Skal | 1617 | int dst_stride[4], |
623 : | edgomez | 851 | int csp, |
624 : | int interlacing) | ||
625 : | albeu | 315 | { |
626 : | edgomez | 851 | const int edged_width2 = edged_width/2; |
627 : | int height2 = height/2; | ||
628 : | |||
629 : | /* | ||
630 : | if (interlacing) | ||
631 : | image_printf(image, edged_width, height, 5,100, "[i]=%i,%i",width,height); | ||
632 : | image_dump_yuvpgm(image, edged_width, width, height, "\\decode.pgm"); | ||
633 : | */ | ||
634 : | |||
635 : | albeu | 315 | switch (csp & ~XVID_CSP_VFLIP) { |
636 : | case XVID_CSP_RGB555: | ||
637 : | edgomez | 851 | safe_packed_conv( |
638 : | edgomez | 1382 | dst[0], dst_stride[0], image->y, image->u, image->v, |
639 : | edgomez | 851 | edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
640 : | interlacing?yv12_to_rgb555i :yv12_to_rgb555, | ||
641 : | Isibaar | 1815 | interlacing?yv12_to_rgb555i_c:yv12_to_rgb555_c, 2, interlacing); |
642 : | albeu | 315 | return 0; |
643 : | |||
644 : | case XVID_CSP_RGB565: | ||
645 : | edgomez | 851 | safe_packed_conv( |
646 : | edgomez | 1382 | dst[0], dst_stride[0], image->y, image->u, image->v, |
647 : | edgomez | 851 | edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
648 : | interlacing?yv12_to_rgb565i :yv12_to_rgb565, | ||
649 : | Isibaar | 1815 | interlacing?yv12_to_rgb565i_c:yv12_to_rgb565_c, 2, interlacing); |
650 : | albeu | 315 | return 0; |
651 : | |||
652 : | edgomez | 1382 | case XVID_CSP_BGR: |
653 : | edgomez | 851 | safe_packed_conv( |
654 : | edgomez | 1382 | dst[0], dst_stride[0], image->y, image->u, image->v, |
655 : | edgomez | 851 | edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
656 : | interlacing?yv12_to_bgri :yv12_to_bgr, | ||
657 : | Isibaar | 1815 | interlacing?yv12_to_bgri_c:yv12_to_bgr_c, 3, interlacing); |
658 : | albeu | 315 | return 0; |
659 : | |||
660 : | edgomez | 1382 | case XVID_CSP_BGRA: |
661 : | edgomez | 851 | safe_packed_conv( |
662 : | edgomez | 1382 | dst[0], dst_stride[0], image->y, image->u, image->v, |
663 : | edgomez | 851 | edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
664 : | interlacing?yv12_to_bgrai :yv12_to_bgra, | ||
665 : | Isibaar | 1815 | interlacing?yv12_to_bgrai_c:yv12_to_bgra_c, 4, interlacing); |
666 : | albeu | 315 | return 0; |
667 : | |||
668 : | edgomez | 851 | case XVID_CSP_ABGR: |
669 : | safe_packed_conv( | ||
670 : | edgomez | 1382 | dst[0], dst_stride[0], image->y, image->u, image->v, |
671 : | edgomez | 851 | edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
672 : | interlacing?yv12_to_abgri :yv12_to_abgr, | ||
673 : | Isibaar | 1815 | interlacing?yv12_to_abgri_c:yv12_to_abgr_c, 4, interlacing); |
674 : | albeu | 315 | return 0; |
675 : | |||
676 : | chl | 1759 | case XVID_CSP_RGB: |
677 : | safe_packed_conv( | ||
678 : | dst[0], dst_stride[0], image->y, image->u, image->v, | ||
679 : | edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), | ||
680 : | interlacing?yv12_to_rgbi :yv12_to_rgb, | ||
681 : | Isibaar | 1815 | interlacing?yv12_to_rgbi_c:yv12_to_rgb_c, 3, interlacing); |
682 : | chl | 1759 | return 0; |
683 : | |||
684 : | edgomez | 851 | case XVID_CSP_RGBA: |
685 : | safe_packed_conv( | ||
686 : | edgomez | 1382 | dst[0], dst_stride[0], image->y, image->u, image->v, |
687 : | edgomez | 851 | edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
688 : | interlacing?yv12_to_rgbai :yv12_to_rgba, | ||
689 : | Isibaar | 1815 | interlacing?yv12_to_rgbai_c:yv12_to_rgba_c, 4, interlacing); |
690 : | albeu | 315 | return 0; |
691 : | |||
692 : | edgomez | 1382 | case XVID_CSP_ARGB: |
693 : | safe_packed_conv( | ||
694 : | dst[0], dst_stride[0], image->y, image->u, image->v, | ||
695 : | edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), | ||
696 : | interlacing?yv12_to_argbi :yv12_to_argb, | ||
697 : | Isibaar | 1815 | interlacing?yv12_to_argbi_c:yv12_to_argb_c, 4, interlacing); |
698 : | edgomez | 1382 | return 0; |
699 : | |||
700 : | albeu | 315 | case XVID_CSP_YUY2: |
701 : | edgomez | 851 | safe_packed_conv( |
702 : | edgomez | 1382 | dst[0], dst_stride[0], image->y, image->u, image->v, |
703 : | edgomez | 851 | edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
704 : | interlacing?yv12_to_yuyvi :yv12_to_yuyv, | ||
705 : | Isibaar | 1815 | interlacing?yv12_to_yuyvi_c:yv12_to_yuyv_c, 2, interlacing); |
706 : | albeu | 315 | return 0; |
707 : | |||
708 : | edgomez | 1382 | case XVID_CSP_YVYU: /* u,v swapped */ |
709 : | edgomez | 851 | safe_packed_conv( |
710 : | edgomez | 1382 | dst[0], dst_stride[0], image->y, image->v, image->u, |
711 : | edgomez | 851 | edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
712 : | interlacing?yv12_to_yuyvi :yv12_to_yuyv, | ||
713 : | Isibaar | 1815 | interlacing?yv12_to_yuyvi_c:yv12_to_yuyv_c, 2, interlacing); |
714 : | albeu | 315 | return 0; |
715 : | |||
716 : | case XVID_CSP_UYVY: | ||
717 : | edgomez | 851 | safe_packed_conv( |
718 : | edgomez | 1382 | dst[0], dst_stride[0], image->y, image->u, image->v, |
719 : | edgomez | 851 | edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
720 : | interlacing?yv12_to_uyvyi :yv12_to_uyvy, | ||
721 : | Isibaar | 1815 | interlacing?yv12_to_uyvyi_c:yv12_to_uyvy_c, 2, interlacing); |
722 : | albeu | 315 | return 0; |
723 : | |||
724 : | edgomez | 1382 | case XVID_CSP_I420: /* YCbCr == YUV == internal colorspace for MPEG */ |
725 : | yv12_to_yv12(dst[0], dst[0] + dst_stride[0]*height, dst[0] + dst_stride[0]*height + (dst_stride[0]/2)*height2, | ||
726 : | dst_stride[0], dst_stride[0]/2, | ||
727 : | edgomez | 851 | image->y, image->u, image->v, edged_width, edged_width2, |
728 : | width, height, (csp & XVID_CSP_VFLIP)); | ||
729 : | return 0; | ||
730 : | |||
731 : | edgomez | 1382 | case XVID_CSP_YV12: /* YCrCb == YVU == U and V plane swapped */ |
732 : | yv12_to_yv12(dst[0], dst[0] + dst_stride[0]*height, dst[0] + dst_stride[0]*height + (dst_stride[0]/2)*height2, | ||
733 : | dst_stride[0], dst_stride[0]/2, | ||
734 : | edgomez | 851 | image->y, image->v, image->u, edged_width, edged_width2, |
735 : | width, height, (csp & XVID_CSP_VFLIP)); | ||
736 : | return 0; | ||
737 : | |||
738 : | edgomez | 1382 | case XVID_CSP_PLANAR: /* YCbCr with arbitrary pointers and different strides for Y and UV */ |
739 : | yv12_to_yv12(dst[0], dst[1], dst[2], | ||
740 : | dst_stride[0], dst_stride[1], /* v: dst_stride[2] not yet supported */ | ||
741 : | image->y, image->u, image->v, edged_width, edged_width2, | ||
742 : | width, height, (csp & XVID_CSP_VFLIP)); | ||
743 : | albeu | 315 | return 0; |
744 : | |||
745 : | edgomez | 1382 | case XVID_CSP_INTERNAL : |
746 : | dst[0] = image->y; | ||
747 : | dst[1] = image->u; | ||
748 : | dst[2] = image->v; | ||
749 : | dst_stride[0] = edged_width; | ||
750 : | dst_stride[1] = edged_width/2; | ||
751 : | dst_stride[2] = edged_width/2; | ||
752 : | return 0; | ||
753 : | |||
754 : | albeu | 315 | case XVID_CSP_NULL: |
755 : | edgomez | 1382 | case XVID_CSP_SLICE: |
756 : | albeu | 315 | return 0; |
757 : | |||
758 : | } | ||
759 : | |||
760 : | return -1; | ||
761 : | } | ||
762 : | |||
763 : | float | ||
764 : | image_psnr(IMAGE * orig_image, | ||
765 : | IMAGE * recon_image, | ||
766 : | uint16_t stride, | ||
767 : | uint16_t width, | ||
768 : | uint16_t height) | ||
769 : | { | ||
770 : | int32_t diff, x, y, quad = 0; | ||
771 : | uint8_t *orig = orig_image->y; | ||
772 : | uint8_t *recon = recon_image->y; | ||
773 : | float psnr_y; | ||
774 : | |||
775 : | for (y = 0; y < height; y++) { | ||
776 : | for (x = 0; x < width; x++) { | ||
777 : | diff = *(orig + x) - *(recon + x); | ||
778 : | quad += diff * diff; | ||
779 : | } | ||
780 : | orig += stride; | ||
781 : | recon += stride; | ||
782 : | } | ||
783 : | |||
784 : | psnr_y = (float) quad / (float) (width * height); | ||
785 : | |||
786 : | if (psnr_y) { | ||
787 : | psnr_y = (float) (255 * 255) / psnr_y; | ||
788 : | psnr_y = 10 * (float) log10(psnr_y); | ||
789 : | } else | ||
790 : | psnr_y = (float) 99.99; | ||
791 : | |||
792 : | return psnr_y; | ||
793 : | } | ||
794 : | |||
795 : | |||
796 : | edgomez | 851 | float sse_to_PSNR(long sse, int pixels) |
797 : | { | ||
798 : | if (sse==0) | ||
799 : | return 99.99F; | ||
800 : | |||
801 : | edgomez | 1382 | return 48.131F - 10*(float)log10((float)sse/(float)(pixels)); /* log10(255*255)=4.8131 */ |
802 : | edgomez | 851 | |
803 : | } | ||
804 : | |||
805 : | edgomez | 1424 | long plane_sse(uint8_t *orig, |
806 : | uint8_t *recon, | ||
807 : | uint16_t stride, | ||
808 : | uint16_t width, | ||
809 : | uint16_t height) | ||
810 : | edgomez | 851 | { |
811 : | edgomez | 1424 | int y, bwidth, bheight; |
812 : | long sse = 0; | ||
813 : | edgomez | 851 | |
814 : | edgomez | 1424 | bwidth = width & (~0x07); |
815 : | bheight = height & (~0x07); | ||
816 : | |||
817 : | /* Compute the 8x8 integer part */ | ||
818 : | for (y = 0; y<bheight; y += 8) { | ||
819 : | int x; | ||
820 : | |||
821 : | /* Compute sse for the band */ | ||
822 : | for (x = 0; x<bwidth; x += 8) | ||
823 : | sse += sse8_8bit(orig + x, recon + x, stride); | ||
824 : | |||
825 : | /* remaining pixels of the 8 pixels high band */ | ||
826 : | for (x = bwidth; x < width; x++) { | ||
827 : | int diff; | ||
828 : | diff = *(orig + 0*stride + x) - *(recon + 0*stride + x); | ||
829 : | sse += diff * diff; | ||
830 : | diff = *(orig + 1*stride + x) - *(recon + 1*stride + x); | ||
831 : | sse += diff * diff; | ||
832 : | diff = *(orig + 2*stride + x) - *(recon + 2*stride + x); | ||
833 : | sse += diff * diff; | ||
834 : | diff = *(orig + 3*stride + x) - *(recon + 3*stride + x); | ||
835 : | sse += diff * diff; | ||
836 : | diff = *(orig + 4*stride + x) - *(recon + 4*stride + x); | ||
837 : | sse += diff * diff; | ||
838 : | diff = *(orig + 5*stride + x) - *(recon + 5*stride + x); | ||
839 : | sse += diff * diff; | ||
840 : | diff = *(orig + 6*stride + x) - *(recon + 6*stride + x); | ||
841 : | sse += diff * diff; | ||
842 : | diff = *(orig + 7*stride + x) - *(recon + 7*stride + x); | ||
843 : | sse += diff * diff; | ||
844 : | } | ||
845 : | |||
846 : | orig += 8*stride; | ||
847 : | recon += 8*stride; | ||
848 : | } | ||
849 : | |||
850 : | /* Compute the down rectangle sse */ | ||
851 : | for (y = bheight; y < height; y++) { | ||
852 : | int x; | ||
853 : | edgomez | 851 | for (x = 0; x < width; x++) { |
854 : | edgomez | 1424 | int diff; |
855 : | edgomez | 851 | diff = *(orig + x) - *(recon + x); |
856 : | sse += diff * diff; | ||
857 : | } | ||
858 : | orig += stride; | ||
859 : | recon += stride; | ||
860 : | } | ||
861 : | edgomez | 1424 | |
862 : | return (sse); | ||
863 : | edgomez | 851 | } |
864 : | |||
865 : | edgomez | 1382 | #if 0 |
866 : | edgomez | 851 | |
867 : | albeu | 315 | #include <stdio.h> |
868 : | #include <string.h> | ||
869 : | |||
870 : | int image_dump_pgm(uint8_t * bmp, uint32_t width, uint32_t height, char * filename) | ||
871 : | { | ||
872 : | FILE * f; | ||
873 : | char hdr[1024]; | ||
874 : | edgomez | 1382 | |
875 : | albeu | 315 | f = fopen(filename, "wb"); |
876 : | if ( f == NULL) | ||
877 : | { | ||
878 : | return -1; | ||
879 : | } | ||
880 : | sprintf(hdr, "P5\n#xvid\n%i %i\n255\n", width, height); | ||
881 : | fwrite(hdr, strlen(hdr), 1, f); | ||
882 : | fwrite(bmp, width, height, f); | ||
883 : | fclose(f); | ||
884 : | |||
885 : | return 0; | ||
886 : | } | ||
887 : | |||
888 : | |||
889 : | edgomez | 1382 | /* dump image+edges to yuv pgm files */ |
890 : | albeu | 315 | |
891 : | int image_dump(IMAGE * image, uint32_t edged_width, uint32_t edged_height, char * path, int number) | ||
892 : | { | ||
893 : | char filename[1024]; | ||
894 : | |||
895 : | sprintf(filename, "%s_%i_%c.pgm", path, number, 'y'); | ||
896 : | image_dump_pgm( | ||
897 : | image->y - (EDGE_SIZE * edged_width + EDGE_SIZE), | ||
898 : | edged_width, edged_height, filename); | ||
899 : | |||
900 : | sprintf(filename, "%s_%i_%c.pgm", path, number, 'u'); | ||
901 : | image_dump_pgm( | ||
902 : | image->u - (EDGE_SIZE2 * edged_width / 2 + EDGE_SIZE2), | ||
903 : | edged_width / 2, edged_height / 2, filename); | ||
904 : | |||
905 : | sprintf(filename, "%s_%i_%c.pgm", path, number, 'v'); | ||
906 : | image_dump_pgm( | ||
907 : | image->v - (EDGE_SIZE2 * edged_width / 2 + EDGE_SIZE2), | ||
908 : | edged_width / 2, edged_height / 2, filename); | ||
909 : | |||
910 : | return 0; | ||
911 : | } | ||
912 : | edgomez | 1382 | #endif |
913 : | albeu | 315 | |
914 : | |||
915 : | |||
916 : | /* dump image to yuvpgm file */ | ||
917 : | |||
918 : | #include <stdio.h> | ||
919 : | |||
920 : | int | ||
921 : | image_dump_yuvpgm(const IMAGE * image, | ||
922 : | const uint32_t edged_width, | ||
923 : | const uint32_t width, | ||
924 : | const uint32_t height, | ||
925 : | char *filename) | ||
926 : | { | ||
927 : | FILE *f; | ||
928 : | char hdr[1024]; | ||
929 : | uint32_t i; | ||
930 : | uint8_t *bmp1; | ||
931 : | uint8_t *bmp2; | ||
932 : | |||
933 : | |||
934 : | f = fopen(filename, "wb"); | ||
935 : | if (f == NULL) { | ||
936 : | return -1; | ||
937 : | } | ||
938 : | sprintf(hdr, "P5\n#xvid\n%i %i\n255\n", width, (3 * height) / 2); | ||
939 : | fwrite(hdr, strlen(hdr), 1, f); | ||
940 : | |||
941 : | bmp1 = image->y; | ||
942 : | for (i = 0; i < height; i++) { | ||
943 : | fwrite(bmp1, width, 1, f); | ||
944 : | bmp1 += edged_width; | ||
945 : | } | ||
946 : | |||
947 : | bmp1 = image->u; | ||
948 : | bmp2 = image->v; | ||
949 : | for (i = 0; i < height / 2; i++) { | ||
950 : | fwrite(bmp1, width / 2, 1, f); | ||
951 : | fwrite(bmp2, width / 2, 1, f); | ||
952 : | bmp1 += edged_width / 2; | ||
953 : | bmp2 += edged_width / 2; | ||
954 : | } | ||
955 : | |||
956 : | fclose(f); | ||
957 : | return 0; | ||
958 : | } | ||
959 : | |||
960 : | |||
961 : | float | ||
962 : | image_mad(const IMAGE * img1, | ||
963 : | const IMAGE * img2, | ||
964 : | uint32_t stride, | ||
965 : | uint32_t width, | ||
966 : | uint32_t height) | ||
967 : | { | ||
968 : | const uint32_t stride2 = stride / 2; | ||
969 : | const uint32_t width2 = width / 2; | ||
970 : | const uint32_t height2 = height / 2; | ||
971 : | |||
972 : | uint32_t x, y; | ||
973 : | uint32_t sum = 0; | ||
974 : | |||
975 : | for (y = 0; y < height; y++) | ||
976 : | for (x = 0; x < width; x++) | ||
977 : | edgomez | 1382 | sum += abs(img1->y[x + y * stride] - img2->y[x + y * stride]); |
978 : | albeu | 315 | |
979 : | for (y = 0; y < height2; y++) | ||
980 : | for (x = 0; x < width2; x++) | ||
981 : | edgomez | 1382 | sum += abs(img1->u[x + y * stride2] - img2->u[x + y * stride2]); |
982 : | albeu | 315 | |
983 : | for (y = 0; y < height2; y++) | ||
984 : | for (x = 0; x < width2; x++) | ||
985 : | edgomez | 1382 | sum += abs(img1->v[x + y * stride2] - img2->v[x + y * stride2]); |
986 : | albeu | 315 | |
987 : | return (float) sum / (width * height * 3 / 2); | ||
988 : | } | ||
989 : | |||
990 : | void | ||
991 : | suxen_drol | 1631 | output_slice(IMAGE * cur, int stride, int width, xvid_image_t* out_frm, int mbx, int mby,int mbl) { |
992 : | albeu | 315 | uint8_t *dY,*dU,*dV,*sY,*sU,*sV; |
993 : | suxen_drol | 1631 | int stride2 = stride >> 1; |
994 : | albeu | 315 | int w = mbl << 4, w2,i; |
995 : | |||
996 : | if(w > width) | ||
997 : | w = width; | ||
998 : | w2 = w >> 1; | ||
999 : | suxen_drol | 323 | |
1000 : | edgomez | 1382 | dY = (uint8_t*)out_frm->plane[0] + (mby << 4) * out_frm->stride[0] + (mbx << 4); |
1001 : | dU = (uint8_t*)out_frm->plane[1] + (mby << 3) * out_frm->stride[1] + (mbx << 3); | ||
1002 : | dV = (uint8_t*)out_frm->plane[2] + (mby << 3) * out_frm->stride[2] + (mbx << 3); | ||
1003 : | suxen_drol | 1631 | sY = cur->y + (mby << 4) * stride + (mbx << 4); |
1004 : | sU = cur->u + (mby << 3) * stride2 + (mbx << 3); | ||
1005 : | sV = cur->v + (mby << 3) * stride2 + (mbx << 3); | ||
1006 : | albeu | 315 | |
1007 : | for(i = 0 ; i < 16 ; i++) { | ||
1008 : | memcpy(dY,sY,w); | ||
1009 : | edgomez | 1382 | dY += out_frm->stride[0]; |
1010 : | suxen_drol | 1631 | sY += stride; |
1011 : | suxen_drol | 323 | } |
1012 : | albeu | 315 | for(i = 0 ; i < 8 ; i++) { |
1013 : | memcpy(dU,sU,w2); | ||
1014 : | edgomez | 1382 | dU += out_frm->stride[1]; |
1015 : | suxen_drol | 1631 | sU += stride2; |
1016 : | albeu | 315 | } |
1017 : | for(i = 0 ; i < 8 ; i++) { | ||
1018 : | memcpy(dV,sV,w2); | ||
1019 : | edgomez | 1382 | dV += out_frm->stride[2]; |
1020 : | suxen_drol | 1631 | sV += stride2; |
1021 : | albeu | 315 | } |
1022 : | } | ||
1023 : | edgomez | 851 | |
1024 : | |||
1025 : | void | ||
1026 : | image_clear(IMAGE * img, int width, int height, int edged_width, | ||
1027 : | int y, int u, int v) | ||
1028 : | { | ||
1029 : | uint8_t * p; | ||
1030 : | int i; | ||
1031 : | |||
1032 : | p = img->y; | ||
1033 : | for (i = 0; i < height; i++) { | ||
1034 : | memset(p, y, width); | ||
1035 : | p += edged_width; | ||
1036 : | } | ||
1037 : | |||
1038 : | p = img->u; | ||
1039 : | for (i = 0; i < height/2; i++) { | ||
1040 : | memset(p, u, width/2); | ||
1041 : | p += edged_width/2; | ||
1042 : | } | ||
1043 : | |||
1044 : | p = img->v; | ||
1045 : | for (i = 0; i < height/2; i++) { | ||
1046 : | memset(p, v, width/2); | ||
1047 : | p += edged_width/2; | ||
1048 : | } | ||
1049 : | } | ||
1050 : | Skal | 1733 | |
1051 : | /****************************************************************************/ | ||
1052 : | |||
1053 : | Isibaar | 1734 | static void (*deintl_core)(uint8_t *, int width, int height, const int stride) = 0; |
1054 : | extern void xvid_deinterlace_sse(uint8_t *, int width, int height, const int stride); | ||
1055 : | Skal | 1733 | |
1056 : | #define CLIP_255(x) ( ((x)&~255) ? ((-(x)) >> (8*sizeof((x))-1))&0xff : (x) ) | ||
1057 : | |||
1058 : | Isibaar | 1734 | static void deinterlace_c(uint8_t *pix, int width, int height, const int bps) |
1059 : | Skal | 1733 | { |
1060 : | pix += bps; | ||
1061 : | while(width-->0) | ||
1062 : | { | ||
1063 : | int p1 = pix[-bps]; | ||
1064 : | int p2 = pix[0]; | ||
1065 : | int p0 = p2; | ||
1066 : | int j = (height>>1) - 1; | ||
1067 : | int V; | ||
1068 : | unsigned char *P = pix++; | ||
1069 : | while(j-->0) | ||
1070 : | { | ||
1071 : | const int p3 = P[ bps]; | ||
1072 : | const int p4 = P[2*bps]; | ||
1073 : | V = ((p1+p3+1)>>1) + ((p2 - ((p0+p4+1)>>1)) >> 2); | ||
1074 : | P[0] = CLIP_255( V ); | ||
1075 : | p0 = p2; | ||
1076 : | p1 = p3; | ||
1077 : | p2 = p4; | ||
1078 : | P += 2*bps; | ||
1079 : | } | ||
1080 : | V = ((p1+p1+1)>>1) + ((p2 - ((p0+p2+1)>>1)) >> 2); | ||
1081 : | P[0] = CLIP_255( V ); | ||
1082 : | } | ||
1083 : | } | ||
1084 : | #undef CLIP_255 | ||
1085 : | |||
1086 : | int xvid_image_deinterlace(xvid_image_t* img, int width, int height, int bottom_first) | ||
1087 : | { | ||
1088 : | if (height&1) | ||
1089 : | return 0; | ||
1090 : | if (img->csp!=XVID_CSP_PLANAR && img->csp!=XVID_CSP_I420 && img->csp!=XVID_CSP_YV12) | ||
1091 : | return 0; /* not yet supported */ | ||
1092 : | if (deintl_core==0) { | ||
1093 : | deintl_core = deinterlace_c; | ||
1094 : | Isibaar | 1795 | #if defined(ARCH_IS_IA32) || defined(ARCH_IS_X86_64) |
1095 : | Skal | 1736 | { |
1096 : | int cpu_flags = check_cpu_features(); | ||
1097 : | if (cpu_flags & XVID_CPU_MMX) | ||
1098 : | deintl_core = xvid_deinterlace_sse; | ||
1099 : | } | ||
1100 : | Skal | 1733 | #endif |
1101 : | } | ||
1102 : | if (!bottom_first) { | ||
1103 : | deintl_core(img->plane[0], width, height, img->stride[0]); | ||
1104 : | deintl_core(img->plane[1], width>>1, height>>1, img->stride[1]); | ||
1105 : | deintl_core(img->plane[2], width>>1, height>>1, img->stride[2]); | ||
1106 : | } | ||
1107 : | else { | ||
1108 : | Isibaar | 1734 | deintl_core((uint8_t *)img->plane[0] + ( height -1)*img->stride[0], width, height, -img->stride[0]); |
1109 : | deintl_core((uint8_t *)img->plane[1] + ((height>>1)-1)*img->stride[1], width>>1, height>>1, -img->stride[1]); | ||
1110 : | deintl_core((uint8_t *)img->plane[2] + ((height>>1)-1)*img->stride[2], width>>1, height>>1, -img->stride[2]); | ||
1111 : | Skal | 1733 | } |
1112 : | emms(); | ||
1113 : | |||
1114 : | return 1; | ||
1115 : | } | ||
1116 : |
No admin address has been configured | ViewVC Help |
Powered by ViewVC 1.0.4 |