1 |
/************************************************************************** |
/************************************************************************** |
2 |
* |
* |
3 |
* XVID MPEG-4 VIDEO CODEC |
* XVID MPEG-4 VIDEO CODEC |
4 |
* image stuff |
* - Image management functions - |
5 |
|
* |
6 |
|
* Copyright(C) 2001-2004 Peter Ross <pross@xvid.org> |
7 |
* |
* |
8 |
* This program is free software; you can redistribute it and/or modify |
* 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 |
* it under the terms of the GNU General Public License as published by |
17 |
* |
* |
18 |
* You should have received a copy of the GNU General Public License |
* You should have received a copy of the GNU General Public License |
19 |
* along with this program; if not, write to the Free Software |
* along with this program; if not, write to the Free Software |
20 |
* Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. |
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA |
|
* |
|
|
*************************************************************************/ |
|
|
|
|
|
/************************************************************************** |
|
21 |
* |
* |
22 |
* History: |
* $Id$ |
23 |
* |
* |
24 |
* 01.05.2002 BFRAME image-based u,v interpolation |
****************************************************************************/ |
|
* 22.04.2002 added some B-frame support |
|
|
* 14.04.2002 added image_dump_yuvpgm(), added image_mad() |
|
|
* XVID_CSP_USER input support |
|
|
* 09.04.2002 PSNR calculations |
|
|
* 06.04.2002 removed interlaced edging from U,V blocks (as per spec) |
|
|
* 26.03.2002 interlacing support (field-based edging in set_edges) |
|
|
* 26.01.2002 rgb555, rgb565 |
|
|
* 07.01.2001 commented u,v interpolation (not required for uv-block-based) |
|
|
* 23.12.2001 removed #ifdefs, added function pointers + init_common() |
|
|
* 22.12.2001 cpu #ifdefs |
|
|
* 19.12.2001 image_dump(); useful for debugging |
|
|
* 6.12.2001 inital version; (c)2001 peter ross <pross@cs.rmit.edu.au> |
|
|
* |
|
|
*************************************************************************/ |
|
25 |
|
|
26 |
#include <stdlib.h> |
#include <stdlib.h> |
27 |
#include <string.h> // memcpy, memset |
#include <string.h> /* memcpy, memset */ |
28 |
#include <math.h> |
#include <math.h> |
|
|
|
29 |
#include "../portab.h" |
#include "../portab.h" |
30 |
#include "../xvid.h" // XVID_CSP_XXX's |
#include "../global.h" /* XVID_CSP_XXX's */ |
31 |
|
#include "../xvid.h" /* XVID_CSP_XXX's */ |
32 |
#include "image.h" |
#include "image.h" |
33 |
#include "colorspace.h" |
#include "colorspace.h" |
34 |
#include "interpolate8x8.h" |
#include "interpolate8x8.h" |
|
#include "../divx4.h" |
|
35 |
#include "../utils/mem_align.h" |
#include "../utils/mem_align.h" |
36 |
|
#include "../motion/sad.h" |
37 |
|
|
38 |
|
#include "font.h" /* XXX: remove later */ |
39 |
|
|
40 |
#define SAFETY 64 |
#define SAFETY 64 |
41 |
#define EDGE_SIZE2 (EDGE_SIZE/2) |
#define EDGE_SIZE2 (EDGE_SIZE/2) |
48 |
{ |
{ |
49 |
const uint32_t edged_width2 = edged_width / 2; |
const uint32_t edged_width2 = edged_width / 2; |
50 |
const uint32_t edged_height2 = edged_height / 2; |
const uint32_t edged_height2 = edged_height / 2; |
|
uint32_t i; |
|
51 |
|
|
52 |
image->y = |
image->y = |
53 |
xvid_malloc(edged_width * (edged_height + 1) + SAFETY, CACHE_LINE); |
xvid_malloc(edged_width * (edged_height + 1) + SAFETY, CACHE_LINE); |
54 |
if (image->y == NULL) { |
if (image->y == NULL) { |
55 |
return -1; |
return -1; |
56 |
} |
} |
57 |
|
memset(image->y, 0, edged_width * (edged_height + 1) + SAFETY); |
|
for (i = 0; i < edged_width * edged_height + SAFETY; i++) { |
|
|
image->y[i] = 0; |
|
|
} |
|
58 |
|
|
59 |
image->u = xvid_malloc(edged_width2 * edged_height2 + SAFETY, CACHE_LINE); |
image->u = xvid_malloc(edged_width2 * edged_height2 + SAFETY, CACHE_LINE); |
60 |
if (image->u == NULL) { |
if (image->u == NULL) { |
61 |
xvid_free(image->y); |
xvid_free(image->y); |
62 |
|
image->y = NULL; |
63 |
return -1; |
return -1; |
64 |
} |
} |
65 |
|
memset(image->u, 0, edged_width2 * edged_height2 + SAFETY); |
66 |
|
|
67 |
image->v = xvid_malloc(edged_width2 * edged_height2 + SAFETY, CACHE_LINE); |
image->v = xvid_malloc(edged_width2 * edged_height2 + SAFETY, CACHE_LINE); |
68 |
if (image->v == NULL) { |
if (image->v == NULL) { |
69 |
xvid_free(image->u); |
xvid_free(image->u); |
70 |
|
image->u = NULL; |
71 |
xvid_free(image->y); |
xvid_free(image->y); |
72 |
|
image->y = NULL; |
73 |
return -1; |
return -1; |
74 |
} |
} |
75 |
|
memset(image->v, 0, edged_width2 * edged_height2 + SAFETY); |
76 |
|
|
77 |
image->y += EDGE_SIZE * edged_width + EDGE_SIZE; |
image->y += EDGE_SIZE * edged_width + EDGE_SIZE; |
78 |
image->u += EDGE_SIZE2 * edged_width2 + EDGE_SIZE2; |
image->u += EDGE_SIZE2 * edged_width2 + EDGE_SIZE2; |
92 |
|
|
93 |
if (image->y) { |
if (image->y) { |
94 |
xvid_free(image->y - (EDGE_SIZE * edged_width + EDGE_SIZE)); |
xvid_free(image->y - (EDGE_SIZE * edged_width + EDGE_SIZE)); |
95 |
|
image->y = NULL; |
96 |
} |
} |
97 |
if (image->u) { |
if (image->u) { |
98 |
xvid_free(image->u - (EDGE_SIZE2 * edged_width2 + EDGE_SIZE2)); |
xvid_free(image->u - (EDGE_SIZE2 * edged_width2 + EDGE_SIZE2)); |
99 |
|
image->u = NULL; |
100 |
} |
} |
101 |
if (image->v) { |
if (image->v) { |
102 |
xvid_free(image->v - (EDGE_SIZE2 * edged_width2 + EDGE_SIZE2)); |
xvid_free(image->v - (EDGE_SIZE2 * edged_width2 + EDGE_SIZE2)); |
103 |
|
image->v = NULL; |
104 |
} |
} |
105 |
} |
} |
106 |
|
|
109 |
image_swap(IMAGE * image1, |
image_swap(IMAGE * image1, |
110 |
IMAGE * image2) |
IMAGE * image2) |
111 |
{ |
{ |
112 |
uint8_t *tmp; |
SWAP(uint8_t*, image1->y, image2->y); |
113 |
|
SWAP(uint8_t*, image1->u, image2->u); |
114 |
tmp = image1->y; |
SWAP(uint8_t*, image1->v, image2->v); |
|
image1->y = image2->y; |
|
|
image2->y = tmp; |
|
|
|
|
|
tmp = image1->u; |
|
|
image1->u = image2->u; |
|
|
image2->u = tmp; |
|
|
|
|
|
tmp = image1->v; |
|
|
image1->v = image2->v; |
|
|
image2->v = tmp; |
|
115 |
} |
} |
116 |
|
|
117 |
|
|
126 |
memcpy(image1->v, image2->v, edged_width * height / 4); |
memcpy(image1->v, image2->v, edged_width * height / 4); |
127 |
} |
} |
128 |
|
|
129 |
|
/* setedges bug was fixed in this BS version */ |
130 |
|
#define SETEDGES_BUG_BEFORE 18 |
131 |
|
|
132 |
void |
void |
133 |
image_setedges(IMAGE * image, |
image_setedges(IMAGE * image, |
135 |
uint32_t edged_height, |
uint32_t edged_height, |
136 |
uint32_t width, |
uint32_t width, |
137 |
uint32_t height, |
uint32_t height, |
138 |
uint32_t interlacing) |
int bs_version) |
139 |
{ |
{ |
140 |
const uint32_t edged_width2 = edged_width / 2; |
const uint32_t edged_width2 = edged_width / 2; |
141 |
const uint32_t width2 = width / 2; |
uint32_t width2; |
142 |
uint32_t i; |
uint32_t i; |
143 |
uint8_t *dst; |
uint8_t *dst; |
144 |
uint8_t *src; |
uint8_t *src; |
145 |
|
|
|
|
|
146 |
dst = image->y - (EDGE_SIZE + EDGE_SIZE * edged_width); |
dst = image->y - (EDGE_SIZE + EDGE_SIZE * edged_width); |
147 |
src = image->y; |
src = image->y; |
148 |
|
|
149 |
|
/* 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 |
for (i = 0; i < EDGE_SIZE; i++) { |
for (i = 0; i < EDGE_SIZE; i++) { |
|
// if interlacing, edges contain top-most data from each field |
|
|
if (interlacing && (i & 1)) { |
|
|
memset(dst, *(src + edged_width), EDGE_SIZE); |
|
|
memcpy(dst + EDGE_SIZE, src + edged_width, width); |
|
|
memset(dst + edged_width - EDGE_SIZE, |
|
|
*(src + edged_width + width - 1), EDGE_SIZE); |
|
|
} else { |
|
159 |
memset(dst, *src, EDGE_SIZE); |
memset(dst, *src, EDGE_SIZE); |
160 |
memcpy(dst + EDGE_SIZE, src, width); |
memcpy(dst + EDGE_SIZE, src, width); |
161 |
memset(dst + edged_width - EDGE_SIZE, *(src + width - 1), |
memset(dst + edged_width - EDGE_SIZE, *(src + width - 1), |
162 |
EDGE_SIZE); |
EDGE_SIZE); |
|
} |
|
163 |
dst += edged_width; |
dst += edged_width; |
164 |
} |
} |
165 |
|
|
172 |
|
|
173 |
src -= edged_width; |
src -= edged_width; |
174 |
for (i = 0; i < EDGE_SIZE; i++) { |
for (i = 0; i < EDGE_SIZE; i++) { |
|
// if interlacing, edges contain bottom-most data from each field |
|
|
if (interlacing && !(i & 1)) { |
|
|
memset(dst, *(src - edged_width), EDGE_SIZE); |
|
|
memcpy(dst + EDGE_SIZE, src - edged_width, width); |
|
|
memset(dst + edged_width - EDGE_SIZE, |
|
|
*(src - edged_width + width - 1), EDGE_SIZE); |
|
|
} else { |
|
175 |
memset(dst, *src, EDGE_SIZE); |
memset(dst, *src, EDGE_SIZE); |
176 |
memcpy(dst + EDGE_SIZE, src, width); |
memcpy(dst + EDGE_SIZE, src, width); |
177 |
memset(dst + edged_width - EDGE_SIZE, *(src + width - 1), |
memset(dst + edged_width - EDGE_SIZE, *(src + width - 1), |
178 |
EDGE_SIZE); |
EDGE_SIZE); |
|
} |
|
179 |
dst += edged_width; |
dst += edged_width; |
180 |
} |
} |
181 |
|
|
182 |
|
|
183 |
//U |
/* U */ |
184 |
dst = image->u - (EDGE_SIZE2 + EDGE_SIZE2 * edged_width2); |
dst = image->u - (EDGE_SIZE2 + EDGE_SIZE2 * edged_width2); |
185 |
src = image->u; |
src = image->u; |
186 |
|
|
208 |
} |
} |
209 |
|
|
210 |
|
|
211 |
// V |
/* V */ |
212 |
dst = image->v - (EDGE_SIZE2 + EDGE_SIZE2 * edged_width2); |
dst = image->v - (EDGE_SIZE2 + EDGE_SIZE2 * edged_width2); |
213 |
src = image->v; |
src = image->v; |
214 |
|
|
236 |
} |
} |
237 |
} |
} |
238 |
|
|
239 |
// bframe encoding requires image-based u,v interpolation |
/* bframe encoding requires image-based u,v interpolation */ |
240 |
void |
void |
241 |
image_interpolate(const IMAGE * refn, |
image_interpolate(const IMAGE * refn, |
242 |
IMAGE * refh, |
IMAGE * refh, |
244 |
IMAGE * refhv, |
IMAGE * refhv, |
245 |
uint32_t edged_width, |
uint32_t edged_width, |
246 |
uint32_t edged_height, |
uint32_t edged_height, |
247 |
|
uint32_t quarterpel, |
248 |
uint32_t rounding) |
uint32_t rounding) |
249 |
{ |
{ |
250 |
const uint32_t offset = EDGE_SIZE * (edged_width + 1); |
const uint32_t offset = EDGE_SIZE2 * (edged_width + 1); /* we only interpolate half of the edge area */ |
251 |
const uint32_t stride_add = 7 * edged_width; |
const uint32_t stride_add = 7 * edged_width; |
252 |
|
#if 0 |
|
#ifdef BFRAMES |
|
253 |
const uint32_t edged_width2 = edged_width / 2; |
const uint32_t edged_width2 = edged_width / 2; |
254 |
const uint32_t edged_height2 = edged_height / 2; |
const uint32_t edged_height2 = edged_height / 2; |
255 |
const uint32_t offset2 = EDGE_SIZE2 * (edged_width2 + 1); |
const uint32_t offset2 = EDGE_SIZE2 * (edged_width2 + 1); |
256 |
const uint32_t stride_add2 = 7 * edged_width2; |
const uint32_t stride_add2 = 7 * edged_width2; |
257 |
#endif |
#endif |
|
|
|
258 |
uint8_t *n_ptr, *h_ptr, *v_ptr, *hv_ptr; |
uint8_t *n_ptr, *h_ptr, *v_ptr, *hv_ptr; |
259 |
uint32_t x, y; |
uint32_t x, y; |
260 |
|
|
262 |
n_ptr = refn->y; |
n_ptr = refn->y; |
263 |
h_ptr = refh->y; |
h_ptr = refh->y; |
264 |
v_ptr = refv->y; |
v_ptr = refv->y; |
|
hv_ptr = refhv->y; |
|
265 |
|
|
266 |
n_ptr -= offset; |
n_ptr -= offset; |
267 |
h_ptr -= offset; |
h_ptr -= offset; |
268 |
v_ptr -= offset; |
v_ptr -= offset; |
269 |
|
|
270 |
|
/* 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 |
|
if(quarterpel) { |
273 |
|
|
274 |
|
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 |
|
|
279 |
|
n_ptr += 8; |
280 |
|
h_ptr += 8; |
281 |
|
v_ptr += 8; |
282 |
|
} |
283 |
|
|
284 |
|
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 |
|
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 |
|
|
296 |
|
for (y = 0; y < (edged_height - EDGE_SIZE); y = y + 8) { |
297 |
|
hv_ptr -= stride_add; |
298 |
|
h_ptr -= stride_add; |
299 |
|
hv_ptr -= EDGE_SIZE; |
300 |
|
h_ptr -= EDGE_SIZE; |
301 |
|
|
302 |
|
for (x = 0; x < (edged_width - EDGE_SIZE); x = x + 8) { |
303 |
|
hv_ptr -= 8; |
304 |
|
h_ptr -= 8; |
305 |
|
interpolate8x8_6tap_lowpass_v(hv_ptr, h_ptr, edged_width, rounding); |
306 |
|
} |
307 |
|
} |
308 |
|
} else { |
309 |
|
|
310 |
|
hv_ptr = refhv->y; |
311 |
hv_ptr -= offset; |
hv_ptr -= offset; |
312 |
|
|
313 |
for (y = 0; y < edged_height; y = y + 8) { |
for (y = 0; y < (edged_height - EDGE_SIZE); y += 8) { |
314 |
for (x = 0; x < edged_width; x = x + 8) { |
for (x = 0; x < (edged_width - EDGE_SIZE); x += 8) { |
315 |
interpolate8x8_halfpel_h(h_ptr, n_ptr, edged_width, rounding); |
interpolate8x8_halfpel_h(h_ptr, n_ptr, edged_width, rounding); |
316 |
interpolate8x8_halfpel_v(v_ptr, n_ptr, edged_width, rounding); |
interpolate8x8_halfpel_v(v_ptr, n_ptr, edged_width, rounding); |
317 |
interpolate8x8_halfpel_hv(hv_ptr, n_ptr, edged_width, rounding); |
interpolate8x8_halfpel_hv(hv_ptr, n_ptr, edged_width, rounding); |
321 |
v_ptr += 8; |
v_ptr += 8; |
322 |
hv_ptr += 8; |
hv_ptr += 8; |
323 |
} |
} |
324 |
|
|
325 |
|
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; |
h_ptr += stride_add; |
331 |
v_ptr += stride_add; |
v_ptr += stride_add; |
332 |
hv_ptr += stride_add; |
hv_ptr += stride_add; |
333 |
n_ptr += stride_add; |
n_ptr += stride_add; |
334 |
} |
} |
335 |
|
} |
336 |
|
/* |
337 |
#ifdef BFRAMES |
#ifdef BFRAMES |
338 |
n_ptr = refn->u; |
n_ptr = refn->u; |
339 |
h_ptr = refh->u; |
h_ptr = refh->u; |
345 |
v_ptr -= offset2; |
v_ptr -= offset2; |
346 |
hv_ptr -= offset2; |
hv_ptr -= offset2; |
347 |
|
|
348 |
for (y = 0; y < edged_height2; y = y + 8) { |
for (y = 0; y < edged_height2; y += 8) { |
349 |
for (x = 0; x < edged_width2; x = x + 8) { |
for (x = 0; x < edged_width2; x += 8) { |
350 |
interpolate8x8_halfpel_h(h_ptr, n_ptr, edged_width2, rounding); |
interpolate8x8_halfpel_h(h_ptr, n_ptr, edged_width2, rounding); |
351 |
interpolate8x8_halfpel_v(v_ptr, n_ptr, edged_width2, rounding); |
interpolate8x8_halfpel_v(v_ptr, n_ptr, edged_width2, rounding); |
352 |
interpolate8x8_halfpel_hv(hv_ptr, n_ptr, edged_width2, rounding); |
interpolate8x8_halfpel_hv(hv_ptr, n_ptr, edged_width2, rounding); |
389 |
n_ptr += stride_add2; |
n_ptr += stride_add2; |
390 |
} |
} |
391 |
#endif |
#endif |
392 |
|
*/ |
393 |
/* |
/* |
394 |
interpolate_halfpel_h( |
interpolate_halfpel_h( |
395 |
refh->y - offset, |
refh->y - offset, |
453 |
} |
} |
454 |
|
|
455 |
|
|
456 |
|
/* |
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 |
|
if (IS_PURE(IMG_Y(y*2 ,x*2 )) && |
477 |
|
IS_PURE(IMG_Y(y*2 ,x*2+1)) && |
478 |
|
IS_PURE(IMG_Y(y*2+1,x*2 )) && |
479 |
|
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 |
|
|
492 |
|
DPRINTF(XVID_DEBUG_DEBUG,"chroma_optimized_pixels = %i/%i\n", pixels, width*height/4); |
493 |
|
} |
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 |
|
static void |
506 |
|
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 |
int |
int |
540 |
image_input(IMAGE * image, |
image_input(IMAGE * image, |
541 |
uint32_t width, |
uint32_t width, |
542 |
int height, |
int height, |
543 |
uint32_t edged_width, |
uint32_t edged_width, |
544 |
uint8_t * src, |
uint8_t * src[4], |
545 |
int csp) |
int src_stride[4], |
546 |
|
int csp, |
547 |
|
int interlacing) |
548 |
{ |
{ |
549 |
|
const int edged_width2 = edged_width/2; |
550 |
/* if (csp & XVID_CSP_VFLIP) |
const int width2 = width/2; |
551 |
{ |
const int height2 = height/2; |
552 |
height = -height; |
#if 0 |
553 |
} |
const int height_signed = (csp & XVID_CSP_VFLIP) ? -height : height; |
554 |
*/ |
#endif |
555 |
|
|
556 |
switch (csp & ~XVID_CSP_VFLIP) { |
switch (csp & ~XVID_CSP_VFLIP) { |
557 |
case XVID_CSP_RGB555: |
case XVID_CSP_RGB555: |
558 |
rgb555_to_yv12(image->y, image->u, image->v, src, width, height, |
safe_packed_conv( |
559 |
edged_width); |
src[0], src_stride[0], image->y, image->u, image->v, |
560 |
return 0; |
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 |
|
|
565 |
case XVID_CSP_RGB565: |
case XVID_CSP_RGB565: |
566 |
rgb565_to_yv12(image->y, image->u, image->v, src, width, height, |
safe_packed_conv( |
567 |
edged_width); |
src[0], src_stride[0], image->y, image->u, image->v, |
568 |
return 0; |
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 |
|
|
573 |
|
|
574 |
case XVID_CSP_RGB24: |
case XVID_CSP_BGR: |
575 |
rgb24_to_yv12(image->y, image->u, image->v, src, width, height, |
safe_packed_conv( |
576 |
edged_width); |
src[0], src_stride[0], image->y, image->u, image->v, |
577 |
return 0; |
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 |
|
|
582 |
case XVID_CSP_RGB32: |
case XVID_CSP_BGRA: |
583 |
rgb32_to_yv12(image->y, image->u, image->v, src, width, height, |
safe_packed_conv( |
584 |
edged_width); |
src[0], src_stride[0], image->y, image->u, image->v, |
585 |
return 0; |
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 |
|
|
590 |
case XVID_CSP_I420: |
case XVID_CSP_ABGR : |
591 |
yuv_to_yv12(image->y, image->u, image->v, src, width, height, |
safe_packed_conv( |
592 |
edged_width); |
src[0], src_stride[0], image->y, image->u, image->v, |
593 |
return 0; |
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 |
|
|
598 |
case XVID_CSP_YV12: /* u/v swapped */ |
case XVID_CSP_RGBA : |
599 |
yuv_to_yv12(image->y, image->v, image->u, src, width, height, |
safe_packed_conv( |
600 |
edged_width); |
src[0], src_stride[0], image->y, image->u, image->v, |
601 |
return 0; |
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 |
|
|
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 |
|
|
614 |
case XVID_CSP_YUY2: |
case XVID_CSP_YUY2: |
615 |
yuyv_to_yv12(image->y, image->u, image->v, src, width, height, |
safe_packed_conv( |
616 |
edged_width); |
src[0], src_stride[0], image->y, image->u, image->v, |
617 |
return 0; |
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 |
|
|
622 |
case XVID_CSP_YVYU: /* u/v swapped */ |
case XVID_CSP_YVYU: /* u/v swapped */ |
623 |
yuyv_to_yv12(image->y, image->v, image->u, src, width, height, |
safe_packed_conv( |
624 |
edged_width); |
src[0], src_stride[0], image->y, image->v, image->u, |
625 |
return 0; |
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 |
|
|
630 |
case XVID_CSP_UYVY: |
case XVID_CSP_UYVY: |
631 |
uyvy_to_yv12(image->y, image->u, image->v, src, width, height, |
safe_packed_conv( |
632 |
edged_width); |
src[0], src_stride[0], image->y, image->u, image->v, |
633 |
return 0; |
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 |
|
|
638 |
case XVID_CSP_USER: |
case XVID_CSP_I420: /* YCbCr == YUV == internal colorspace for MPEG */ |
639 |
user_to_yuv_c(image->y, image->u, image->v, edged_width, |
yv12_to_yv12(image->y, image->u, image->v, edged_width, edged_width2, |
640 |
(DEC_PICTURE *) src, width, height); |
src[0], src[0] + src_stride[0]*height, src[0] + src_stride[0]*height + (src_stride[0]/2)*height2, |
641 |
return 0; |
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 |
|
yv12_to_yv12(image->y, image->v, image->u, edged_width, edged_width2, |
646 |
|
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 |
|
break; |
649 |
|
|
650 |
|
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 |
|
break; |
655 |
|
|
656 |
case XVID_CSP_NULL: |
case XVID_CSP_NULL: |
657 |
break; |
break; |
658 |
|
|
659 |
|
default : |
660 |
|
return -1; |
661 |
} |
} |
662 |
|
|
663 |
return -1; |
|
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 |
|
memset(image->y + i*edged_width + width, |
673 |
|
*(image->y + i*edged_width + width - 1), pad_width); |
674 |
|
} |
675 |
|
for (i = 0; i < height/2; i++) |
676 |
|
{ |
677 |
|
memset(image->u + i*edged_width2 + width2, |
678 |
|
*(image->u + i*edged_width2 + width2 - 1),pad_width/2); |
679 |
|
memset(image->v + i*edged_width2 + width2, |
680 |
|
*(image->v + i*edged_width2 + width2 - 1),pad_width/2); |
681 |
|
} |
682 |
|
} |
683 |
|
|
684 |
|
if (height & 15) |
685 |
|
{ |
686 |
|
int pad_height = 16 - (height&15); |
687 |
|
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 |
} |
} |
711 |
|
|
712 |
|
|
716 |
uint32_t width, |
uint32_t width, |
717 |
int height, |
int height, |
718 |
uint32_t edged_width, |
uint32_t edged_width, |
719 |
uint8_t * dst, |
uint8_t * dst[4], |
720 |
uint32_t dst_stride, |
int dst_stride[4], |
721 |
int csp) |
int csp, |
722 |
|
int interlacing) |
723 |
{ |
{ |
724 |
if (csp & XVID_CSP_VFLIP) { |
const int edged_width2 = edged_width/2; |
725 |
height = -height; |
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 |
switch (csp & ~XVID_CSP_VFLIP) { |
switch (csp & ~XVID_CSP_VFLIP) { |
734 |
case XVID_CSP_RGB555: |
case XVID_CSP_RGB555: |
735 |
yv12_to_rgb555(dst, dst_stride, image->y, image->u, image->v, |
safe_packed_conv( |
736 |
edged_width, edged_width / 2, width, height); |
dst[0], dst_stride[0], image->y, image->u, image->v, |
737 |
|
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 |
return 0; |
return 0; |
741 |
|
|
742 |
case XVID_CSP_RGB565: |
case XVID_CSP_RGB565: |
743 |
yv12_to_rgb565(dst, dst_stride, image->y, image->u, image->v, |
safe_packed_conv( |
744 |
edged_width, edged_width / 2, width, height); |
dst[0], dst_stride[0], image->y, image->u, image->v, |
745 |
return 0; |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
746 |
|
interlacing?yv12_to_rgb565i :yv12_to_rgb565, |
747 |
case XVID_CSP_RGB24: |
interlacing?yv12_to_rgb565i_c:yv12_to_rgb565_c, 2); |
748 |
yv12_to_rgb24(dst, dst_stride, image->y, image->u, image->v, |
return 0; |
749 |
edged_width, edged_width / 2, width, height); |
|
750 |
return 0; |
case XVID_CSP_BGR: |
751 |
|
safe_packed_conv( |
752 |
case XVID_CSP_RGB32: |
dst[0], dst_stride[0], image->y, image->u, image->v, |
753 |
yv12_to_rgb32(dst, dst_stride, image->y, image->u, image->v, |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
754 |
edged_width, edged_width / 2, width, height); |
interlacing?yv12_to_bgri :yv12_to_bgr, |
755 |
return 0; |
interlacing?yv12_to_bgri_c:yv12_to_bgr_c, 3); |
756 |
|
return 0; |
757 |
case XVID_CSP_I420: |
|
758 |
yv12_to_yuv(dst, dst_stride, image->y, image->u, image->v, edged_width, |
case XVID_CSP_BGRA: |
759 |
edged_width / 2, width, height); |
safe_packed_conv( |
760 |
return 0; |
dst[0], dst_stride[0], image->y, image->u, image->v, |
761 |
|
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
762 |
case XVID_CSP_YV12: // u,v swapped |
interlacing?yv12_to_bgrai :yv12_to_bgra, |
763 |
yv12_to_yuv(dst, dst_stride, image->y, image->v, image->u, edged_width, |
interlacing?yv12_to_bgrai_c:yv12_to_bgra_c, 4); |
764 |
edged_width / 2, width, height); |
return 0; |
765 |
|
|
766 |
|
case XVID_CSP_ABGR: |
767 |
|
safe_packed_conv( |
768 |
|
dst[0], dst_stride[0], image->y, image->u, image->v, |
769 |
|
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 |
|
return 0; |
773 |
|
|
774 |
|
case XVID_CSP_RGBA: |
775 |
|
safe_packed_conv( |
776 |
|
dst[0], dst_stride[0], image->y, image->u, image->v, |
777 |
|
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 |
|
return 0; |
781 |
|
|
782 |
|
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; |
return 0; |
789 |
|
|
790 |
case XVID_CSP_YUY2: |
case XVID_CSP_YUY2: |
791 |
yv12_to_yuyv(dst, dst_stride, image->y, image->u, image->v, |
safe_packed_conv( |
792 |
edged_width, edged_width / 2, width, height); |
dst[0], dst_stride[0], image->y, image->u, image->v, |
793 |
|
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 |
return 0; |
return 0; |
797 |
|
|
798 |
case XVID_CSP_YVYU: // u,v swapped |
case XVID_CSP_YVYU: /* u,v swapped */ |
799 |
yv12_to_yuyv(dst, dst_stride, image->y, image->v, image->u, |
safe_packed_conv( |
800 |
edged_width, edged_width / 2, width, height); |
dst[0], dst_stride[0], image->y, image->v, image->u, |
801 |
|
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 |
return 0; |
return 0; |
805 |
|
|
806 |
case XVID_CSP_UYVY: |
case XVID_CSP_UYVY: |
807 |
yv12_to_uyvy(dst, dst_stride, image->y, image->u, image->v, |
safe_packed_conv( |
808 |
edged_width, edged_width / 2, width, height); |
dst[0], dst_stride[0], image->y, image->u, image->v, |
809 |
return 0; |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
810 |
|
interlacing?yv12_to_uyvyi :yv12_to_uyvy, |
811 |
case XVID_CSP_USER: |
interlacing?yv12_to_uyvyi_c:yv12_to_uyvy_c, 2); |
812 |
((DEC_PICTURE *) dst)->y = image->y; |
return 0; |
813 |
((DEC_PICTURE *) dst)->u = image->u; |
|
814 |
((DEC_PICTURE *) dst)->v = image->v; |
case XVID_CSP_I420: /* YCbCr == YUV == internal colorspace for MPEG */ |
815 |
((DEC_PICTURE *) dst)->stride_y = edged_width; |
yv12_to_yv12(dst[0], dst[0] + dst_stride[0]*height, dst[0] + dst_stride[0]*height + (dst_stride[0]/2)*height2, |
816 |
((DEC_PICTURE *) dst)->stride_uv = edged_width / 2; |
dst_stride[0], dst_stride[0]/2, |
817 |
|
image->y, image->u, image->v, edged_width, edged_width2, |
818 |
|
width, height, (csp & XVID_CSP_VFLIP)); |
819 |
|
return 0; |
820 |
|
|
821 |
|
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 |
|
image->y, image->v, image->u, edged_width, edged_width2, |
825 |
|
width, height, (csp & XVID_CSP_VFLIP)); |
826 |
|
return 0; |
827 |
|
|
828 |
|
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 |
|
return 0; |
834 |
|
|
835 |
|
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; |
return 0; |
843 |
|
|
844 |
case XVID_CSP_NULL: |
case XVID_CSP_NULL: |
845 |
case XVID_CSP_EXTERN: |
case XVID_CSP_SLICE: |
846 |
return 0; |
return 0; |
847 |
|
|
848 |
} |
} |
882 |
return psnr_y; |
return psnr_y; |
883 |
} |
} |
884 |
|
|
885 |
/* |
|
886 |
|
float sse_to_PSNR(long sse, int pixels) |
887 |
|
{ |
888 |
|
if (sse==0) |
889 |
|
return 99.99F; |
890 |
|
|
891 |
|
return 48.131F - 10*(float)log10((float)sse/(float)(pixels)); /* log10(255*255)=4.8131 */ |
892 |
|
|
893 |
|
} |
894 |
|
|
895 |
|
long plane_sse(uint8_t *orig, |
896 |
|
uint8_t *recon, |
897 |
|
uint16_t stride, |
898 |
|
uint16_t width, |
899 |
|
uint16_t height) |
900 |
|
{ |
901 |
|
int y, bwidth, bheight; |
902 |
|
long sse = 0; |
903 |
|
|
904 |
|
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 |
|
for (x = 0; x < width; x++) { |
944 |
|
int diff; |
945 |
|
diff = *(orig + x) - *(recon + x); |
946 |
|
sse += diff * diff; |
947 |
|
} |
948 |
|
orig += stride; |
949 |
|
recon += stride; |
950 |
|
} |
951 |
|
|
952 |
|
return (sse); |
953 |
|
} |
954 |
|
|
955 |
|
#if 0 |
956 |
|
|
957 |
#include <stdio.h> |
#include <stdio.h> |
958 |
#include <string.h> |
#include <string.h> |
976 |
} |
} |
977 |
|
|
978 |
|
|
979 |
// dump image+edges to yuv pgm files |
/* dump image+edges to yuv pgm files */ |
980 |
|
|
981 |
int image_dump(IMAGE * image, uint32_t edged_width, uint32_t edged_height, char * path, int number) |
int image_dump(IMAGE * image, uint32_t edged_width, uint32_t edged_height, char * path, int number) |
982 |
{ |
{ |
999 |
|
|
1000 |
return 0; |
return 0; |
1001 |
} |
} |
1002 |
*/ |
#endif |
1003 |
|
|
1004 |
|
|
1005 |
|
|
1048 |
} |
} |
1049 |
|
|
1050 |
|
|
|
#define ABS(X) (((X)>0)?(X):-(X)) |
|
1051 |
float |
float |
1052 |
image_mad(const IMAGE * img1, |
image_mad(const IMAGE * img1, |
1053 |
const IMAGE * img2, |
const IMAGE * img2, |
1064 |
|
|
1065 |
for (y = 0; y < height; y++) |
for (y = 0; y < height; y++) |
1066 |
for (x = 0; x < width; x++) |
for (x = 0; x < width; x++) |
1067 |
sum += ABS(img1->y[x + y * stride] - img2->y[x + y * stride]); |
sum += abs(img1->y[x + y * stride] - img2->y[x + y * stride]); |
1068 |
|
|
1069 |
for (y = 0; y < height2; y++) |
for (y = 0; y < height2; y++) |
1070 |
for (x = 0; x < width2; x++) |
for (x = 0; x < width2; x++) |
1071 |
sum += ABS(img1->u[x + y * stride2] - img2->u[x + y * stride2]); |
sum += abs(img1->u[x + y * stride2] - img2->u[x + y * stride2]); |
1072 |
|
|
1073 |
for (y = 0; y < height2; y++) |
for (y = 0; y < height2; y++) |
1074 |
for (x = 0; x < width2; x++) |
for (x = 0; x < width2; x++) |
1075 |
sum += ABS(img1->v[x + y * stride2] - img2->v[x + y * stride2]); |
sum += abs(img1->v[x + y * stride2] - img2->v[x + y * stride2]); |
1076 |
|
|
1077 |
return (float) sum / (width * height * 3 / 2); |
return (float) sum / (width * height * 3 / 2); |
1078 |
} |
} |
1079 |
|
|
1080 |
void |
void |
1081 |
output_slice(IMAGE * cur, int std, int width, XVID_DEC_PICTURE* out_frm, int mbx, int mby,int mbl) { |
output_slice(IMAGE * cur, int stride, int width, xvid_image_t* out_frm, int mbx, int mby,int mbl) { |
1082 |
uint8_t *dY,*dU,*dV,*sY,*sU,*sV; |
uint8_t *dY,*dU,*dV,*sY,*sU,*sV; |
1083 |
int std2 = std >> 1; |
int stride2 = stride >> 1; |
1084 |
int w = mbl << 4, w2,i; |
int w = mbl << 4, w2,i; |
1085 |
|
|
1086 |
if(w > width) |
if(w > width) |
1087 |
w = width; |
w = width; |
1088 |
w2 = w >> 1; |
w2 = w >> 1; |
1089 |
|
|
1090 |
dY = (uint8_t*)out_frm->y + (mby << 4) * out_frm->stride_y + (mbx << 4); |
dY = (uint8_t*)out_frm->plane[0] + (mby << 4) * out_frm->stride[0] + (mbx << 4); |
1091 |
dU = (uint8_t*)out_frm->u + (mby << 3) * out_frm->stride_u + (mbx << 3); |
dU = (uint8_t*)out_frm->plane[1] + (mby << 3) * out_frm->stride[1] + (mbx << 3); |
1092 |
dV = (uint8_t*)out_frm->v + (mby << 3) * out_frm->stride_v + (mbx << 3); |
dV = (uint8_t*)out_frm->plane[2] + (mby << 3) * out_frm->stride[2] + (mbx << 3); |
1093 |
sY = cur->y + (mby << 4) * std + (mbx << 4); |
sY = cur->y + (mby << 4) * stride + (mbx << 4); |
1094 |
sU = cur->u + (mby << 3) * std2 + (mbx << 3); |
sU = cur->u + (mby << 3) * stride2 + (mbx << 3); |
1095 |
sV = cur->v + (mby << 3) * std2 + (mbx << 3); |
sV = cur->v + (mby << 3) * stride2 + (mbx << 3); |
1096 |
|
|
1097 |
for(i = 0 ; i < 16 ; i++) { |
for(i = 0 ; i < 16 ; i++) { |
1098 |
memcpy(dY,sY,w); |
memcpy(dY,sY,w); |
1099 |
dY += out_frm->stride_y; |
dY += out_frm->stride[0]; |
1100 |
sY += std; |
sY += stride; |
1101 |
} |
} |
1102 |
for(i = 0 ; i < 8 ; i++) { |
for(i = 0 ; i < 8 ; i++) { |
1103 |
memcpy(dU,sU,w2); |
memcpy(dU,sU,w2); |
1104 |
dU += out_frm->stride_u; |
dU += out_frm->stride[1]; |
1105 |
sU += std2; |
sU += stride2; |
1106 |
} |
} |
1107 |
for(i = 0 ; i < 8 ; i++) { |
for(i = 0 ; i < 8 ; i++) { |
1108 |
memcpy(dV,sV,w2); |
memcpy(dV,sV,w2); |
1109 |
dV += out_frm->stride_v; |
dV += out_frm->stride[2]; |
1110 |
sV += std2; |
sV += stride2; |
1111 |
|
} |
1112 |
|
} |
1113 |
|
|
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 |
} |
} |