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 |
|
|
30 |
#include "../portab.h" |
#include "../portab.h" |
31 |
#include "../xvid.h" // XVID_CSP_XXX's |
#include "../global.h" /* XVID_CSP_XXX's */ |
32 |
|
#include "../xvid.h" /* XVID_CSP_XXX's */ |
33 |
#include "image.h" |
#include "image.h" |
34 |
#include "colorspace.h" |
#include "colorspace.h" |
35 |
#include "interpolate8x8.h" |
#include "interpolate8x8.h" |
36 |
#include "../divx4.h" |
#include "reduced.h" |
37 |
#include "../utils/mem_align.h" |
#include "../utils/mem_align.h" |
38 |
|
#include "../motion/sad.h" |
39 |
|
|
40 |
|
#include "font.h" /* XXX: remove later */ |
41 |
|
|
42 |
#define SAFETY 64 |
#define SAFETY 64 |
43 |
#define EDGE_SIZE2 (EDGE_SIZE/2) |
#define EDGE_SIZE2 (EDGE_SIZE/2) |
50 |
{ |
{ |
51 |
const uint32_t edged_width2 = edged_width / 2; |
const uint32_t edged_width2 = edged_width / 2; |
52 |
const uint32_t edged_height2 = edged_height / 2; |
const uint32_t edged_height2 = edged_height / 2; |
|
uint32_t i; |
|
53 |
|
|
54 |
image->y = |
image->y = |
55 |
xvid_malloc(edged_width * (edged_height + 1) + SAFETY, CACHE_LINE); |
xvid_malloc(edged_width * (edged_height + 1) + SAFETY, CACHE_LINE); |
56 |
if (image->y == NULL) { |
if (image->y == NULL) { |
57 |
return -1; |
return -1; |
58 |
} |
} |
59 |
|
memset(image->y, 0, edged_width * (edged_height + 1) + SAFETY); |
|
for (i = 0; i < edged_width * edged_height + SAFETY; i++) { |
|
|
image->y[i] = 0; |
|
|
} |
|
60 |
|
|
61 |
image->u = xvid_malloc(edged_width2 * edged_height2 + SAFETY, CACHE_LINE); |
image->u = xvid_malloc(edged_width2 * edged_height2 + SAFETY, CACHE_LINE); |
62 |
if (image->u == NULL) { |
if (image->u == NULL) { |
63 |
xvid_free(image->y); |
xvid_free(image->y); |
64 |
|
image->y = NULL; |
65 |
return -1; |
return -1; |
66 |
} |
} |
67 |
|
memset(image->u, 0, edged_width2 * edged_height2 + SAFETY); |
68 |
|
|
69 |
image->v = xvid_malloc(edged_width2 * edged_height2 + SAFETY, CACHE_LINE); |
image->v = xvid_malloc(edged_width2 * edged_height2 + SAFETY, CACHE_LINE); |
70 |
if (image->v == NULL) { |
if (image->v == NULL) { |
71 |
xvid_free(image->u); |
xvid_free(image->u); |
72 |
|
image->u = NULL; |
73 |
xvid_free(image->y); |
xvid_free(image->y); |
74 |
|
image->y = NULL; |
75 |
return -1; |
return -1; |
76 |
} |
} |
77 |
|
memset(image->v, 0, edged_width2 * edged_height2 + SAFETY); |
78 |
|
|
79 |
image->y += EDGE_SIZE * edged_width + EDGE_SIZE; |
image->y += EDGE_SIZE * edged_width + EDGE_SIZE; |
80 |
image->u += EDGE_SIZE2 * edged_width2 + EDGE_SIZE2; |
image->u += EDGE_SIZE2 * edged_width2 + EDGE_SIZE2; |
94 |
|
|
95 |
if (image->y) { |
if (image->y) { |
96 |
xvid_free(image->y - (EDGE_SIZE * edged_width + EDGE_SIZE)); |
xvid_free(image->y - (EDGE_SIZE * edged_width + EDGE_SIZE)); |
97 |
|
image->y = NULL; |
98 |
} |
} |
99 |
if (image->u) { |
if (image->u) { |
100 |
xvid_free(image->u - (EDGE_SIZE2 * edged_width2 + EDGE_SIZE2)); |
xvid_free(image->u - (EDGE_SIZE2 * edged_width2 + EDGE_SIZE2)); |
101 |
|
image->u = NULL; |
102 |
} |
} |
103 |
if (image->v) { |
if (image->v) { |
104 |
xvid_free(image->v - (EDGE_SIZE2 * edged_width2 + EDGE_SIZE2)); |
xvid_free(image->v - (EDGE_SIZE2 * edged_width2 + EDGE_SIZE2)); |
105 |
|
image->v = NULL; |
106 |
} |
} |
107 |
} |
} |
108 |
|
|
111 |
image_swap(IMAGE * image1, |
image_swap(IMAGE * image1, |
112 |
IMAGE * image2) |
IMAGE * image2) |
113 |
{ |
{ |
114 |
uint8_t *tmp; |
SWAP(uint8_t*, image1->y, image2->y); |
115 |
|
SWAP(uint8_t*, image1->u, image2->u); |
116 |
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; |
|
117 |
} |
} |
118 |
|
|
119 |
|
|
128 |
memcpy(image1->v, image2->v, edged_width * height / 4); |
memcpy(image1->v, image2->v, edged_width * height / 4); |
129 |
} |
} |
130 |
|
|
131 |
|
/* setedges bug was fixed in this BS version */ |
132 |
|
#define SETEDGES_BUG_BEFORE 18 |
133 |
|
|
134 |
void |
void |
135 |
image_setedges(IMAGE * image, |
image_setedges(IMAGE * image, |
137 |
uint32_t edged_height, |
uint32_t edged_height, |
138 |
uint32_t width, |
uint32_t width, |
139 |
uint32_t height, |
uint32_t height, |
140 |
uint32_t interlacing) |
int bs_version) |
141 |
{ |
{ |
142 |
const uint32_t edged_width2 = edged_width / 2; |
const uint32_t edged_width2 = edged_width / 2; |
143 |
const uint32_t width2 = width / 2; |
uint32_t width2; |
144 |
uint32_t i; |
uint32_t i; |
145 |
uint8_t *dst; |
uint8_t *dst; |
146 |
uint8_t *src; |
uint8_t *src; |
147 |
|
|
|
|
|
148 |
dst = image->y - (EDGE_SIZE + EDGE_SIZE * edged_width); |
dst = image->y - (EDGE_SIZE + EDGE_SIZE * edged_width); |
149 |
src = image->y; |
src = image->y; |
150 |
|
|
151 |
|
/* According to the Standard Clause 7.6.4, padding is done starting at 16 |
152 |
|
* pixel width and height multiples. This was not respected in old xvids */ |
153 |
|
if (bs_version == 0 || bs_version >= SETEDGES_BUG_BEFORE) { |
154 |
|
width = (width+15)&~15; |
155 |
|
height = (height+15)&~15; |
156 |
|
} |
157 |
|
|
158 |
|
width2 = width/2; |
159 |
|
|
160 |
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 {*/ |
|
161 |
memset(dst, *src, EDGE_SIZE); |
memset(dst, *src, EDGE_SIZE); |
162 |
memcpy(dst + EDGE_SIZE, src, width); |
memcpy(dst + EDGE_SIZE, src, width); |
163 |
memset(dst + edged_width - EDGE_SIZE, *(src + width - 1), |
memset(dst + edged_width - EDGE_SIZE, *(src + width - 1), |
164 |
EDGE_SIZE); |
EDGE_SIZE); |
|
/*}*/ |
|
165 |
dst += edged_width; |
dst += edged_width; |
166 |
} |
} |
167 |
|
|
174 |
|
|
175 |
src -= edged_width; |
src -= edged_width; |
176 |
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 {*/ |
|
177 |
memset(dst, *src, EDGE_SIZE); |
memset(dst, *src, EDGE_SIZE); |
178 |
memcpy(dst + EDGE_SIZE, src, width); |
memcpy(dst + EDGE_SIZE, src, width); |
179 |
memset(dst + edged_width - EDGE_SIZE, *(src + width - 1), |
memset(dst + edged_width - EDGE_SIZE, *(src + width - 1), |
180 |
EDGE_SIZE); |
EDGE_SIZE); |
|
/*}*/ |
|
181 |
dst += edged_width; |
dst += edged_width; |
182 |
} |
} |
183 |
|
|
184 |
|
|
185 |
//U |
/* U */ |
186 |
dst = image->u - (EDGE_SIZE2 + EDGE_SIZE2 * edged_width2); |
dst = image->u - (EDGE_SIZE2 + EDGE_SIZE2 * edged_width2); |
187 |
src = image->u; |
src = image->u; |
188 |
|
|
210 |
} |
} |
211 |
|
|
212 |
|
|
213 |
// V |
/* V */ |
214 |
dst = image->v - (EDGE_SIZE2 + EDGE_SIZE2 * edged_width2); |
dst = image->v - (EDGE_SIZE2 + EDGE_SIZE2 * edged_width2); |
215 |
src = image->v; |
src = image->v; |
216 |
|
|
238 |
} |
} |
239 |
} |
} |
240 |
|
|
241 |
// bframe encoding requires image-based u,v interpolation |
/* bframe encoding requires image-based u,v interpolation */ |
242 |
void |
void |
243 |
image_interpolate(const IMAGE * refn, |
image_interpolate(const IMAGE * refn, |
244 |
IMAGE * refh, |
IMAGE * refh, |
246 |
IMAGE * refhv, |
IMAGE * refhv, |
247 |
uint32_t edged_width, |
uint32_t edged_width, |
248 |
uint32_t edged_height, |
uint32_t edged_height, |
249 |
|
uint32_t quarterpel, |
250 |
uint32_t rounding) |
uint32_t rounding) |
251 |
{ |
{ |
252 |
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 */ |
253 |
const uint32_t stride_add = 7 * edged_width; |
const uint32_t stride_add = 7 * edged_width; |
254 |
|
#if 0 |
|
#ifdef BFRAMES |
|
255 |
const uint32_t edged_width2 = edged_width / 2; |
const uint32_t edged_width2 = edged_width / 2; |
256 |
const uint32_t edged_height2 = edged_height / 2; |
const uint32_t edged_height2 = edged_height / 2; |
257 |
const uint32_t offset2 = EDGE_SIZE2 * (edged_width2 + 1); |
const uint32_t offset2 = EDGE_SIZE2 * (edged_width2 + 1); |
258 |
const uint32_t stride_add2 = 7 * edged_width2; |
const uint32_t stride_add2 = 7 * edged_width2; |
259 |
#endif |
#endif |
|
|
|
260 |
uint8_t *n_ptr, *h_ptr, *v_ptr, *hv_ptr; |
uint8_t *n_ptr, *h_ptr, *v_ptr, *hv_ptr; |
261 |
uint32_t x, y; |
uint32_t x, y; |
262 |
|
|
264 |
n_ptr = refn->y; |
n_ptr = refn->y; |
265 |
h_ptr = refh->y; |
h_ptr = refh->y; |
266 |
v_ptr = refv->y; |
v_ptr = refv->y; |
|
hv_ptr = refhv->y; |
|
267 |
|
|
268 |
n_ptr -= offset; |
n_ptr -= offset; |
269 |
h_ptr -= offset; |
h_ptr -= offset; |
270 |
v_ptr -= offset; |
v_ptr -= offset; |
271 |
|
|
272 |
|
/* Note we initialize the hv pointer later, as we can optimize code a bit |
273 |
|
* doing it down to up in quarterpel and up to down in halfpel */ |
274 |
|
if(quarterpel) { |
275 |
|
|
276 |
|
for (y = 0; y < (edged_height - EDGE_SIZE); y += 8) { |
277 |
|
for (x = 0; x < (edged_width - EDGE_SIZE); x += 8) { |
278 |
|
interpolate8x8_6tap_lowpass_h(h_ptr, n_ptr, edged_width, rounding); |
279 |
|
interpolate8x8_6tap_lowpass_v(v_ptr, n_ptr, edged_width, rounding); |
280 |
|
|
281 |
|
n_ptr += 8; |
282 |
|
h_ptr += 8; |
283 |
|
v_ptr += 8; |
284 |
|
} |
285 |
|
|
286 |
|
n_ptr += EDGE_SIZE; |
287 |
|
h_ptr += EDGE_SIZE; |
288 |
|
v_ptr += EDGE_SIZE; |
289 |
|
|
290 |
|
h_ptr += stride_add; |
291 |
|
v_ptr += stride_add; |
292 |
|
n_ptr += stride_add; |
293 |
|
} |
294 |
|
|
295 |
|
h_ptr = refh->y + (edged_height - EDGE_SIZE - EDGE_SIZE2)*edged_width - EDGE_SIZE2; |
296 |
|
hv_ptr = refhv->y + (edged_height - EDGE_SIZE - EDGE_SIZE2)*edged_width - EDGE_SIZE2; |
297 |
|
|
298 |
|
for (y = 0; y < (edged_height - EDGE_SIZE); y = y + 8) { |
299 |
|
hv_ptr -= stride_add; |
300 |
|
h_ptr -= stride_add; |
301 |
|
hv_ptr -= EDGE_SIZE; |
302 |
|
h_ptr -= EDGE_SIZE; |
303 |
|
|
304 |
|
for (x = 0; x < (edged_width - EDGE_SIZE); x = x + 8) { |
305 |
|
hv_ptr -= 8; |
306 |
|
h_ptr -= 8; |
307 |
|
interpolate8x8_6tap_lowpass_v(hv_ptr, h_ptr, edged_width, rounding); |
308 |
|
} |
309 |
|
} |
310 |
|
} else { |
311 |
|
|
312 |
|
hv_ptr = refhv->y; |
313 |
hv_ptr -= offset; |
hv_ptr -= offset; |
314 |
|
|
315 |
for (y = 0; y < edged_height; y = y + 8) { |
for (y = 0; y < (edged_height - EDGE_SIZE); y += 8) { |
316 |
for (x = 0; x < edged_width; x = x + 8) { |
for (x = 0; x < (edged_width - EDGE_SIZE); x += 8) { |
317 |
interpolate8x8_halfpel_h(h_ptr, n_ptr, edged_width, rounding); |
interpolate8x8_halfpel_h(h_ptr, n_ptr, edged_width, rounding); |
318 |
interpolate8x8_halfpel_v(v_ptr, n_ptr, edged_width, rounding); |
interpolate8x8_halfpel_v(v_ptr, n_ptr, edged_width, rounding); |
319 |
interpolate8x8_halfpel_hv(hv_ptr, n_ptr, edged_width, rounding); |
interpolate8x8_halfpel_hv(hv_ptr, n_ptr, edged_width, rounding); |
323 |
v_ptr += 8; |
v_ptr += 8; |
324 |
hv_ptr += 8; |
hv_ptr += 8; |
325 |
} |
} |
326 |
|
|
327 |
|
h_ptr += EDGE_SIZE; |
328 |
|
v_ptr += EDGE_SIZE; |
329 |
|
hv_ptr += EDGE_SIZE; |
330 |
|
n_ptr += EDGE_SIZE; |
331 |
|
|
332 |
h_ptr += stride_add; |
h_ptr += stride_add; |
333 |
v_ptr += stride_add; |
v_ptr += stride_add; |
334 |
hv_ptr += stride_add; |
hv_ptr += stride_add; |
335 |
n_ptr += stride_add; |
n_ptr += stride_add; |
336 |
} |
} |
337 |
|
} |
338 |
|
/* |
339 |
#ifdef BFRAMES |
#ifdef BFRAMES |
340 |
n_ptr = refn->u; |
n_ptr = refn->u; |
341 |
h_ptr = refh->u; |
h_ptr = refh->u; |
347 |
v_ptr -= offset2; |
v_ptr -= offset2; |
348 |
hv_ptr -= offset2; |
hv_ptr -= offset2; |
349 |
|
|
350 |
for (y = 0; y < edged_height2; y = y + 8) { |
for (y = 0; y < edged_height2; y += 8) { |
351 |
for (x = 0; x < edged_width2; x = x + 8) { |
for (x = 0; x < edged_width2; x += 8) { |
352 |
interpolate8x8_halfpel_h(h_ptr, n_ptr, edged_width2, rounding); |
interpolate8x8_halfpel_h(h_ptr, n_ptr, edged_width2, rounding); |
353 |
interpolate8x8_halfpel_v(v_ptr, n_ptr, edged_width2, rounding); |
interpolate8x8_halfpel_v(v_ptr, n_ptr, edged_width2, rounding); |
354 |
interpolate8x8_halfpel_hv(hv_ptr, n_ptr, edged_width2, rounding); |
interpolate8x8_halfpel_hv(hv_ptr, n_ptr, edged_width2, rounding); |
391 |
n_ptr += stride_add2; |
n_ptr += stride_add2; |
392 |
} |
} |
393 |
#endif |
#endif |
394 |
|
*/ |
395 |
/* |
/* |
396 |
interpolate_halfpel_h( |
interpolate_halfpel_h( |
397 |
refh->y - offset, |
refh->y - offset, |
455 |
} |
} |
456 |
|
|
457 |
|
|
458 |
|
/* |
459 |
|
chroma optimize filter, invented by mf |
460 |
|
a chroma pixel is average from the surrounding pixels, when the |
461 |
|
correpsonding luma pixels are pure black or white. |
462 |
|
*/ |
463 |
|
|
464 |
|
void |
465 |
|
image_chroma_optimize(IMAGE * img, int width, int height, int edged_width) |
466 |
|
{ |
467 |
|
int x,y; |
468 |
|
int pixels = 0; |
469 |
|
|
470 |
|
for (y = 1; y < height/2 - 1; y++) |
471 |
|
for (x = 1; x < width/2 - 1; x++) |
472 |
|
{ |
473 |
|
#define IS_PURE(a) ((a)<=16||(a)>=235) |
474 |
|
#define IMG_Y(Y,X) img->y[(Y)*edged_width + (X)] |
475 |
|
#define IMG_U(Y,X) img->u[(Y)*edged_width/2 + (X)] |
476 |
|
#define IMG_V(Y,X) img->v[(Y)*edged_width/2 + (X)] |
477 |
|
|
478 |
|
if (IS_PURE(IMG_Y(y*2 ,x*2 )) && |
479 |
|
IS_PURE(IMG_Y(y*2 ,x*2+1)) && |
480 |
|
IS_PURE(IMG_Y(y*2+1,x*2 )) && |
481 |
|
IS_PURE(IMG_Y(y*2+1,x*2+1))) |
482 |
|
{ |
483 |
|
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; |
484 |
|
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; |
485 |
|
pixels++; |
486 |
|
} |
487 |
|
|
488 |
|
#undef IS_PURE |
489 |
|
#undef IMG_Y |
490 |
|
#undef IMG_U |
491 |
|
#undef IMG_V |
492 |
|
} |
493 |
|
|
494 |
|
DPRINTF(XVID_DEBUG_DEBUG,"chroma_optimized_pixels = %i/%i\n", pixels, width*height/4); |
495 |
|
} |
496 |
|
|
497 |
|
|
498 |
|
|
499 |
|
|
500 |
|
|
501 |
|
/* |
502 |
|
perform safe packed colorspace conversion, by splitting |
503 |
|
the image up into an optimized area (pixel width divisible by 16), |
504 |
|
and two unoptimized/plain-c areas (pixel width divisible by 2) |
505 |
|
*/ |
506 |
|
|
507 |
|
static void |
508 |
|
safe_packed_conv(uint8_t * x_ptr, int x_stride, |
509 |
|
uint8_t * y_ptr, uint8_t * u_ptr, uint8_t * v_ptr, |
510 |
|
int y_stride, int uv_stride, |
511 |
|
int width, int height, int vflip, |
512 |
|
packedFunc * func_opt, packedFunc func_c, int size) |
513 |
|
{ |
514 |
|
int width_opt, width_c; |
515 |
|
|
516 |
|
if (func_opt != func_c && x_stride < size*((width+15)/16)*16) |
517 |
|
{ |
518 |
|
width_opt = width & (~15); |
519 |
|
width_c = width - width_opt; |
520 |
|
} |
521 |
|
else |
522 |
|
{ |
523 |
|
width_opt = width; |
524 |
|
width_c = 0; |
525 |
|
} |
526 |
|
|
527 |
|
func_opt(x_ptr, x_stride, |
528 |
|
y_ptr, u_ptr, v_ptr, y_stride, uv_stride, |
529 |
|
width_opt, height, vflip); |
530 |
|
|
531 |
|
if (width_c) |
532 |
|
{ |
533 |
|
func_c(x_ptr + size*width_opt, x_stride, |
534 |
|
y_ptr + width_opt, u_ptr + width_opt/2, v_ptr + width_opt/2, |
535 |
|
y_stride, uv_stride, width_c, height, vflip); |
536 |
|
} |
537 |
|
} |
538 |
|
|
539 |
|
|
540 |
|
|
541 |
int |
int |
542 |
image_input(IMAGE * image, |
image_input(IMAGE * image, |
543 |
uint32_t width, |
uint32_t width, |
544 |
int height, |
int height, |
545 |
uint32_t edged_width, |
uint32_t edged_width, |
546 |
uint8_t * src, |
uint8_t * src[4], |
547 |
int csp) |
int src_stride[4], |
548 |
|
int csp, |
549 |
|
int interlacing) |
550 |
{ |
{ |
551 |
|
const int edged_width2 = edged_width/2; |
552 |
/* if (csp & XVID_CSP_VFLIP) |
const int width2 = width/2; |
553 |
{ |
const int height2 = height/2; |
554 |
height = -height; |
#if 0 |
555 |
} |
const int height_signed = (csp & XVID_CSP_VFLIP) ? -height : height; |
556 |
*/ |
#endif |
557 |
|
|
558 |
switch (csp & ~XVID_CSP_VFLIP) { |
switch (csp & ~XVID_CSP_VFLIP) { |
559 |
case XVID_CSP_RGB555: |
case XVID_CSP_RGB555: |
560 |
rgb555_to_yv12(image->y, image->u, image->v, src, width, height, |
safe_packed_conv( |
561 |
edged_width); |
src[0], src_stride[0], image->y, image->u, image->v, |
562 |
return 0; |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
563 |
|
interlacing?rgb555i_to_yv12 :rgb555_to_yv12, |
564 |
|
interlacing?rgb555i_to_yv12_c:rgb555_to_yv12_c, 2); |
565 |
|
break; |
566 |
|
|
567 |
case XVID_CSP_RGB565: |
case XVID_CSP_RGB565: |
568 |
rgb565_to_yv12(image->y, image->u, image->v, src, width, height, |
safe_packed_conv( |
569 |
edged_width); |
src[0], src_stride[0], image->y, image->u, image->v, |
570 |
return 0; |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
571 |
|
interlacing?rgb565i_to_yv12 :rgb565_to_yv12, |
572 |
|
interlacing?rgb565i_to_yv12_c:rgb565_to_yv12_c, 2); |
573 |
|
break; |
574 |
|
|
575 |
|
|
576 |
case XVID_CSP_RGB24: |
case XVID_CSP_BGR: |
577 |
rgb24_to_yv12(image->y, image->u, image->v, src, width, height, |
safe_packed_conv( |
578 |
edged_width); |
src[0], src_stride[0], image->y, image->u, image->v, |
579 |
return 0; |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
580 |
|
interlacing?bgri_to_yv12 :bgr_to_yv12, |
581 |
|
interlacing?bgri_to_yv12_c:bgr_to_yv12_c, 3); |
582 |
|
break; |
583 |
|
|
584 |
case XVID_CSP_RGB32: |
case XVID_CSP_BGRA: |
585 |
rgb32_to_yv12(image->y, image->u, image->v, src, width, height, |
safe_packed_conv( |
586 |
edged_width); |
src[0], src_stride[0], image->y, image->u, image->v, |
587 |
return 0; |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
588 |
|
interlacing?bgrai_to_yv12 :bgra_to_yv12, |
589 |
|
interlacing?bgrai_to_yv12_c:bgra_to_yv12_c, 4); |
590 |
|
break; |
591 |
|
|
592 |
case XVID_CSP_I420: |
case XVID_CSP_ABGR : |
593 |
yuv_to_yv12(image->y, image->u, image->v, src, width, height, |
safe_packed_conv( |
594 |
edged_width); |
src[0], src_stride[0], image->y, image->u, image->v, |
595 |
return 0; |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
596 |
|
interlacing?abgri_to_yv12 :abgr_to_yv12, |
597 |
|
interlacing?abgri_to_yv12_c:abgr_to_yv12_c, 4); |
598 |
|
break; |
599 |
|
|
600 |
case XVID_CSP_YV12: /* u/v swapped */ |
case XVID_CSP_RGBA : |
601 |
yuv_to_yv12(image->y, image->v, image->u, src, width, height, |
safe_packed_conv( |
602 |
edged_width); |
src[0], src_stride[0], image->y, image->u, image->v, |
603 |
return 0; |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
604 |
|
interlacing?rgbai_to_yv12 :rgba_to_yv12, |
605 |
|
interlacing?rgbai_to_yv12_c:rgba_to_yv12_c, 4); |
606 |
|
break; |
607 |
|
|
608 |
|
case XVID_CSP_ARGB: |
609 |
|
safe_packed_conv( |
610 |
|
src[0], src_stride[0], image->y, image->u, image->v, |
611 |
|
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
612 |
|
interlacing?argbi_to_yv12 : argb_to_yv12, |
613 |
|
interlacing?argbi_to_yv12_c: argb_to_yv12_c, 4); |
614 |
|
break; |
615 |
|
|
616 |
case XVID_CSP_YUY2: |
case XVID_CSP_YUY2: |
617 |
yuyv_to_yv12(image->y, image->u, image->v, src, width, height, |
safe_packed_conv( |
618 |
edged_width); |
src[0], src_stride[0], image->y, image->u, image->v, |
619 |
return 0; |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
620 |
|
interlacing?yuyvi_to_yv12 :yuyv_to_yv12, |
621 |
|
interlacing?yuyvi_to_yv12_c:yuyv_to_yv12_c, 2); |
622 |
|
break; |
623 |
|
|
624 |
case XVID_CSP_YVYU: /* u/v swapped */ |
case XVID_CSP_YVYU: /* u/v swapped */ |
625 |
yuyv_to_yv12(image->y, image->v, image->u, src, width, height, |
safe_packed_conv( |
626 |
edged_width); |
src[0], src_stride[0], image->y, image->v, image->u, |
627 |
return 0; |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
628 |
|
interlacing?yuyvi_to_yv12 :yuyv_to_yv12, |
629 |
|
interlacing?yuyvi_to_yv12_c:yuyv_to_yv12_c, 2); |
630 |
|
break; |
631 |
|
|
632 |
case XVID_CSP_UYVY: |
case XVID_CSP_UYVY: |
633 |
uyvy_to_yv12(image->y, image->u, image->v, src, width, height, |
safe_packed_conv( |
634 |
edged_width); |
src[0], src_stride[0], image->y, image->u, image->v, |
635 |
return 0; |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
636 |
|
interlacing?uyvyi_to_yv12 :uyvy_to_yv12, |
637 |
|
interlacing?uyvyi_to_yv12_c:uyvy_to_yv12_c, 2); |
638 |
|
break; |
639 |
|
|
640 |
case XVID_CSP_USER: |
case XVID_CSP_I420: /* YCbCr == YUV == internal colorspace for MPEG */ |
641 |
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, |
642 |
(DEC_PICTURE *) src, width, height); |
src[0], src[0] + src_stride[0]*height, src[0] + src_stride[0]*height + (src_stride[0]/2)*height2, |
643 |
return 0; |
src_stride[0], src_stride[0]/2, width, height, (csp & XVID_CSP_VFLIP)); |
644 |
|
break; |
645 |
|
|
646 |
|
case XVID_CSP_YV12: /* YCrCb == YVA == U and V plane swapped */ |
647 |
|
yv12_to_yv12(image->y, image->v, image->u, edged_width, edged_width2, |
648 |
|
src[0], src[0] + src_stride[0]*height, src[0] + src_stride[0]*height + (src_stride[0]/2)*height2, |
649 |
|
src_stride[0], src_stride[0]/2, width, height, (csp & XVID_CSP_VFLIP)); |
650 |
|
break; |
651 |
|
|
652 |
|
case XVID_CSP_PLANAR: /* YCbCr with arbitrary pointers and different strides for Y and UV */ |
653 |
|
yv12_to_yv12(image->y, image->u, image->v, edged_width, edged_width2, |
654 |
|
src[0], src[1], src[2], src_stride[0], src_stride[1], /* v: dst_stride[2] not yet supported */ |
655 |
|
width, height, (csp & XVID_CSP_VFLIP)); |
656 |
|
break; |
657 |
|
|
658 |
case XVID_CSP_NULL: |
case XVID_CSP_NULL: |
659 |
break; |
break; |
660 |
|
|
661 |
|
default : |
662 |
|
return -1; |
663 |
} |
} |
664 |
|
|
665 |
return -1; |
|
666 |
|
/* pad out image when the width and/or height is not a multiple of 16 */ |
667 |
|
|
668 |
|
if (width & 15) |
669 |
|
{ |
670 |
|
int i; |
671 |
|
int pad_width = 16 - (width&15); |
672 |
|
for (i = 0; i < height; i++) |
673 |
|
{ |
674 |
|
memset(image->y + i*edged_width + width, |
675 |
|
*(image->y + i*edged_width + width - 1), pad_width); |
676 |
|
} |
677 |
|
for (i = 0; i < height/2; i++) |
678 |
|
{ |
679 |
|
memset(image->u + i*edged_width2 + width2, |
680 |
|
*(image->u + i*edged_width2 + width2 - 1),pad_width/2); |
681 |
|
memset(image->v + i*edged_width2 + width2, |
682 |
|
*(image->v + i*edged_width2 + width2 - 1),pad_width/2); |
683 |
|
} |
684 |
|
} |
685 |
|
|
686 |
|
if (height & 15) |
687 |
|
{ |
688 |
|
int pad_height = 16 - (height&15); |
689 |
|
int length = ((width+15)/16)*16; |
690 |
|
int i; |
691 |
|
for (i = 0; i < pad_height; i++) |
692 |
|
{ |
693 |
|
memcpy(image->y + (height+i)*edged_width, |
694 |
|
image->y + (height-1)*edged_width,length); |
695 |
|
} |
696 |
|
|
697 |
|
for (i = 0; i < pad_height/2; i++) |
698 |
|
{ |
699 |
|
memcpy(image->u + (height2+i)*edged_width2, |
700 |
|
image->u + (height2-1)*edged_width2,length/2); |
701 |
|
memcpy(image->v + (height2+i)*edged_width2, |
702 |
|
image->v + (height2-1)*edged_width2,length/2); |
703 |
|
} |
704 |
|
} |
705 |
|
|
706 |
|
/* |
707 |
|
if (interlacing) |
708 |
|
image_printf(image, edged_width, height, 5,5, "[i]"); |
709 |
|
image_dump_yuvpgm(image, edged_width, ((width+15)/16)*16, ((height+15)/16)*16, "\\encode.pgm"); |
710 |
|
*/ |
711 |
|
return 0; |
712 |
} |
} |
713 |
|
|
714 |
|
|
718 |
uint32_t width, |
uint32_t width, |
719 |
int height, |
int height, |
720 |
uint32_t edged_width, |
uint32_t edged_width, |
721 |
uint8_t * dst, |
uint8_t * dst[4], |
722 |
uint32_t dst_stride, |
uint32_t dst_stride[4], |
723 |
int csp) |
int csp, |
724 |
|
int interlacing) |
725 |
{ |
{ |
726 |
if (csp & XVID_CSP_VFLIP) { |
const int edged_width2 = edged_width/2; |
727 |
height = -height; |
int height2 = height/2; |
728 |
} |
|
729 |
|
/* |
730 |
|
if (interlacing) |
731 |
|
image_printf(image, edged_width, height, 5,100, "[i]=%i,%i",width,height); |
732 |
|
image_dump_yuvpgm(image, edged_width, width, height, "\\decode.pgm"); |
733 |
|
*/ |
734 |
|
|
735 |
switch (csp & ~XVID_CSP_VFLIP) { |
switch (csp & ~XVID_CSP_VFLIP) { |
736 |
case XVID_CSP_RGB555: |
case XVID_CSP_RGB555: |
737 |
yv12_to_rgb555(dst, dst_stride, image->y, image->u, image->v, |
safe_packed_conv( |
738 |
edged_width, edged_width / 2, width, height); |
dst[0], dst_stride[0], image->y, image->u, image->v, |
739 |
|
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
740 |
|
interlacing?yv12_to_rgb555i :yv12_to_rgb555, |
741 |
|
interlacing?yv12_to_rgb555i_c:yv12_to_rgb555_c, 2); |
742 |
return 0; |
return 0; |
743 |
|
|
744 |
case XVID_CSP_RGB565: |
case XVID_CSP_RGB565: |
745 |
yv12_to_rgb565(dst, dst_stride, image->y, image->u, image->v, |
safe_packed_conv( |
746 |
edged_width, edged_width / 2, width, height); |
dst[0], dst_stride[0], image->y, image->u, image->v, |
747 |
return 0; |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
748 |
|
interlacing?yv12_to_rgb565i :yv12_to_rgb565, |
749 |
case XVID_CSP_RGB24: |
interlacing?yv12_to_rgb565i_c:yv12_to_rgb565_c, 2); |
750 |
yv12_to_rgb24(dst, dst_stride, image->y, image->u, image->v, |
return 0; |
751 |
edged_width, edged_width / 2, width, height); |
|
752 |
return 0; |
case XVID_CSP_BGR: |
753 |
|
safe_packed_conv( |
754 |
case XVID_CSP_RGB32: |
dst[0], dst_stride[0], image->y, image->u, image->v, |
755 |
yv12_to_rgb32(dst, dst_stride, image->y, image->u, image->v, |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
756 |
edged_width, edged_width / 2, width, height); |
interlacing?yv12_to_bgri :yv12_to_bgr, |
757 |
return 0; |
interlacing?yv12_to_bgri_c:yv12_to_bgr_c, 3); |
758 |
|
return 0; |
759 |
case XVID_CSP_I420: |
|
760 |
yv12_to_yuv(dst, dst_stride, image->y, image->u, image->v, edged_width, |
case XVID_CSP_BGRA: |
761 |
edged_width / 2, width, height); |
safe_packed_conv( |
762 |
return 0; |
dst[0], dst_stride[0], image->y, image->u, image->v, |
763 |
|
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
764 |
case XVID_CSP_YV12: // u,v swapped |
interlacing?yv12_to_bgrai :yv12_to_bgra, |
765 |
yv12_to_yuv(dst, dst_stride, image->y, image->v, image->u, edged_width, |
interlacing?yv12_to_bgrai_c:yv12_to_bgra_c, 4); |
766 |
edged_width / 2, width, height); |
return 0; |
767 |
|
|
768 |
|
case XVID_CSP_ABGR: |
769 |
|
safe_packed_conv( |
770 |
|
dst[0], dst_stride[0], image->y, image->u, image->v, |
771 |
|
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
772 |
|
interlacing?yv12_to_abgri :yv12_to_abgr, |
773 |
|
interlacing?yv12_to_abgri_c:yv12_to_abgr_c, 4); |
774 |
|
return 0; |
775 |
|
|
776 |
|
case XVID_CSP_RGBA: |
777 |
|
safe_packed_conv( |
778 |
|
dst[0], dst_stride[0], image->y, image->u, image->v, |
779 |
|
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
780 |
|
interlacing?yv12_to_rgbai :yv12_to_rgba, |
781 |
|
interlacing?yv12_to_rgbai_c:yv12_to_rgba_c, 4); |
782 |
|
return 0; |
783 |
|
|
784 |
|
case XVID_CSP_ARGB: |
785 |
|
safe_packed_conv( |
786 |
|
dst[0], dst_stride[0], image->y, image->u, image->v, |
787 |
|
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
788 |
|
interlacing?yv12_to_argbi :yv12_to_argb, |
789 |
|
interlacing?yv12_to_argbi_c:yv12_to_argb_c, 4); |
790 |
return 0; |
return 0; |
791 |
|
|
792 |
case XVID_CSP_YUY2: |
case XVID_CSP_YUY2: |
793 |
yv12_to_yuyv(dst, dst_stride, image->y, image->u, image->v, |
safe_packed_conv( |
794 |
edged_width, edged_width / 2, width, height); |
dst[0], dst_stride[0], image->y, image->u, image->v, |
795 |
|
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
796 |
|
interlacing?yv12_to_yuyvi :yv12_to_yuyv, |
797 |
|
interlacing?yv12_to_yuyvi_c:yv12_to_yuyv_c, 2); |
798 |
return 0; |
return 0; |
799 |
|
|
800 |
case XVID_CSP_YVYU: // u,v swapped |
case XVID_CSP_YVYU: /* u,v swapped */ |
801 |
yv12_to_yuyv(dst, dst_stride, image->y, image->v, image->u, |
safe_packed_conv( |
802 |
edged_width, edged_width / 2, width, height); |
dst[0], dst_stride[0], image->y, image->v, image->u, |
803 |
|
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
804 |
|
interlacing?yv12_to_yuyvi :yv12_to_yuyv, |
805 |
|
interlacing?yv12_to_yuyvi_c:yv12_to_yuyv_c, 2); |
806 |
return 0; |
return 0; |
807 |
|
|
808 |
case XVID_CSP_UYVY: |
case XVID_CSP_UYVY: |
809 |
yv12_to_uyvy(dst, dst_stride, image->y, image->u, image->v, |
safe_packed_conv( |
810 |
edged_width, edged_width / 2, width, height); |
dst[0], dst_stride[0], image->y, image->u, image->v, |
811 |
return 0; |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
812 |
|
interlacing?yv12_to_uyvyi :yv12_to_uyvy, |
813 |
case XVID_CSP_USER: |
interlacing?yv12_to_uyvyi_c:yv12_to_uyvy_c, 2); |
814 |
((DEC_PICTURE *) dst)->y = image->y; |
return 0; |
815 |
((DEC_PICTURE *) dst)->u = image->u; |
|
816 |
((DEC_PICTURE *) dst)->v = image->v; |
case XVID_CSP_I420: /* YCbCr == YUV == internal colorspace for MPEG */ |
817 |
((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, |
818 |
((DEC_PICTURE *) dst)->stride_uv = edged_width / 2; |
dst_stride[0], dst_stride[0]/2, |
819 |
|
image->y, image->u, image->v, edged_width, edged_width2, |
820 |
|
width, height, (csp & XVID_CSP_VFLIP)); |
821 |
|
return 0; |
822 |
|
|
823 |
|
case XVID_CSP_YV12: /* YCrCb == YVU == U and V plane swapped */ |
824 |
|
yv12_to_yv12(dst[0], dst[0] + dst_stride[0]*height, dst[0] + dst_stride[0]*height + (dst_stride[0]/2)*height2, |
825 |
|
dst_stride[0], dst_stride[0]/2, |
826 |
|
image->y, image->v, image->u, edged_width, edged_width2, |
827 |
|
width, height, (csp & XVID_CSP_VFLIP)); |
828 |
|
return 0; |
829 |
|
|
830 |
|
case XVID_CSP_PLANAR: /* YCbCr with arbitrary pointers and different strides for Y and UV */ |
831 |
|
yv12_to_yv12(dst[0], dst[1], dst[2], |
832 |
|
dst_stride[0], dst_stride[1], /* v: dst_stride[2] not yet supported */ |
833 |
|
image->y, image->u, image->v, edged_width, edged_width2, |
834 |
|
width, height, (csp & XVID_CSP_VFLIP)); |
835 |
|
return 0; |
836 |
|
|
837 |
|
case XVID_CSP_INTERNAL : |
838 |
|
dst[0] = image->y; |
839 |
|
dst[1] = image->u; |
840 |
|
dst[2] = image->v; |
841 |
|
dst_stride[0] = edged_width; |
842 |
|
dst_stride[1] = edged_width/2; |
843 |
|
dst_stride[2] = edged_width/2; |
844 |
return 0; |
return 0; |
845 |
|
|
846 |
case XVID_CSP_NULL: |
case XVID_CSP_NULL: |
847 |
case XVID_CSP_EXTERN: |
case XVID_CSP_SLICE: |
848 |
return 0; |
return 0; |
849 |
|
|
850 |
} |
} |
884 |
return psnr_y; |
return psnr_y; |
885 |
} |
} |
886 |
|
|
887 |
/* |
|
888 |
|
float sse_to_PSNR(long sse, int pixels) |
889 |
|
{ |
890 |
|
if (sse==0) |
891 |
|
return 99.99F; |
892 |
|
|
893 |
|
return 48.131F - 10*(float)log10((float)sse/(float)(pixels)); /* log10(255*255)=4.8131 */ |
894 |
|
|
895 |
|
} |
896 |
|
|
897 |
|
long plane_sse(uint8_t *orig, |
898 |
|
uint8_t *recon, |
899 |
|
uint16_t stride, |
900 |
|
uint16_t width, |
901 |
|
uint16_t height) |
902 |
|
{ |
903 |
|
int y, bwidth, bheight; |
904 |
|
long sse = 0; |
905 |
|
|
906 |
|
bwidth = width & (~0x07); |
907 |
|
bheight = height & (~0x07); |
908 |
|
|
909 |
|
/* Compute the 8x8 integer part */ |
910 |
|
for (y = 0; y<bheight; y += 8) { |
911 |
|
int x; |
912 |
|
|
913 |
|
/* Compute sse for the band */ |
914 |
|
for (x = 0; x<bwidth; x += 8) |
915 |
|
sse += sse8_8bit(orig + x, recon + x, stride); |
916 |
|
|
917 |
|
/* remaining pixels of the 8 pixels high band */ |
918 |
|
for (x = bwidth; x < width; x++) { |
919 |
|
int diff; |
920 |
|
diff = *(orig + 0*stride + x) - *(recon + 0*stride + x); |
921 |
|
sse += diff * diff; |
922 |
|
diff = *(orig + 1*stride + x) - *(recon + 1*stride + x); |
923 |
|
sse += diff * diff; |
924 |
|
diff = *(orig + 2*stride + x) - *(recon + 2*stride + x); |
925 |
|
sse += diff * diff; |
926 |
|
diff = *(orig + 3*stride + x) - *(recon + 3*stride + x); |
927 |
|
sse += diff * diff; |
928 |
|
diff = *(orig + 4*stride + x) - *(recon + 4*stride + x); |
929 |
|
sse += diff * diff; |
930 |
|
diff = *(orig + 5*stride + x) - *(recon + 5*stride + x); |
931 |
|
sse += diff * diff; |
932 |
|
diff = *(orig + 6*stride + x) - *(recon + 6*stride + x); |
933 |
|
sse += diff * diff; |
934 |
|
diff = *(orig + 7*stride + x) - *(recon + 7*stride + x); |
935 |
|
sse += diff * diff; |
936 |
|
} |
937 |
|
|
938 |
|
orig += 8*stride; |
939 |
|
recon += 8*stride; |
940 |
|
} |
941 |
|
|
942 |
|
/* Compute the down rectangle sse */ |
943 |
|
for (y = bheight; y < height; y++) { |
944 |
|
int x; |
945 |
|
for (x = 0; x < width; x++) { |
946 |
|
int diff; |
947 |
|
diff = *(orig + x) - *(recon + x); |
948 |
|
sse += diff * diff; |
949 |
|
} |
950 |
|
orig += stride; |
951 |
|
recon += stride; |
952 |
|
} |
953 |
|
|
954 |
|
return (sse); |
955 |
|
} |
956 |
|
|
957 |
|
#if 0 |
958 |
|
|
959 |
#include <stdio.h> |
#include <stdio.h> |
960 |
#include <string.h> |
#include <string.h> |
978 |
} |
} |
979 |
|
|
980 |
|
|
981 |
// dump image+edges to yuv pgm files |
/* dump image+edges to yuv pgm files */ |
982 |
|
|
983 |
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) |
984 |
{ |
{ |
1001 |
|
|
1002 |
return 0; |
return 0; |
1003 |
} |
} |
1004 |
*/ |
#endif |
1005 |
|
|
1006 |
|
|
1007 |
|
|
1050 |
} |
} |
1051 |
|
|
1052 |
|
|
|
#define ABS(X) (((X)>0)?(X):-(X)) |
|
1053 |
float |
float |
1054 |
image_mad(const IMAGE * img1, |
image_mad(const IMAGE * img1, |
1055 |
const IMAGE * img2, |
const IMAGE * img2, |
1066 |
|
|
1067 |
for (y = 0; y < height; y++) |
for (y = 0; y < height; y++) |
1068 |
for (x = 0; x < width; x++) |
for (x = 0; x < width; x++) |
1069 |
sum += ABS(img1->y[x + y * stride] - img2->y[x + y * stride]); |
sum += abs(img1->y[x + y * stride] - img2->y[x + y * stride]); |
1070 |
|
|
1071 |
for (y = 0; y < height2; y++) |
for (y = 0; y < height2; y++) |
1072 |
for (x = 0; x < width2; x++) |
for (x = 0; x < width2; x++) |
1073 |
sum += ABS(img1->u[x + y * stride2] - img2->u[x + y * stride2]); |
sum += abs(img1->u[x + y * stride2] - img2->u[x + y * stride2]); |
1074 |
|
|
1075 |
for (y = 0; y < height2; y++) |
for (y = 0; y < height2; y++) |
1076 |
for (x = 0; x < width2; x++) |
for (x = 0; x < width2; x++) |
1077 |
sum += ABS(img1->v[x + y * stride2] - img2->v[x + y * stride2]); |
sum += abs(img1->v[x + y * stride2] - img2->v[x + y * stride2]); |
1078 |
|
|
1079 |
return (float) sum / (width * height * 3 / 2); |
return (float) sum / (width * height * 3 / 2); |
1080 |
} |
} |
1081 |
|
|
1082 |
void |
void |
1083 |
output_slice(IMAGE * cur, int std, int width, XVID_DEC_PICTURE* out_frm, int mbx, int mby,int mbl) { |
output_slice(IMAGE * cur, int std, int width, xvid_image_t* out_frm, int mbx, int mby,int mbl) { |
1084 |
uint8_t *dY,*dU,*dV,*sY,*sU,*sV; |
uint8_t *dY,*dU,*dV,*sY,*sU,*sV; |
1085 |
int std2 = std >> 1; |
int std2 = std >> 1; |
1086 |
int w = mbl << 4, w2,i; |
int w = mbl << 4, w2,i; |
1089 |
w = width; |
w = width; |
1090 |
w2 = w >> 1; |
w2 = w >> 1; |
1091 |
|
|
1092 |
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); |
1093 |
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); |
1094 |
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); |
1095 |
sY = cur->y + (mby << 4) * std + (mbx << 4); |
sY = cur->y + (mby << 4) * std + (mbx << 4); |
1096 |
sU = cur->u + (mby << 3) * std2 + (mbx << 3); |
sU = cur->u + (mby << 3) * std2 + (mbx << 3); |
1097 |
sV = cur->v + (mby << 3) * std2 + (mbx << 3); |
sV = cur->v + (mby << 3) * std2 + (mbx << 3); |
1098 |
|
|
1099 |
for(i = 0 ; i < 16 ; i++) { |
for(i = 0 ; i < 16 ; i++) { |
1100 |
memcpy(dY,sY,w); |
memcpy(dY,sY,w); |
1101 |
dY += out_frm->stride_y; |
dY += out_frm->stride[0]; |
1102 |
sY += std; |
sY += std; |
1103 |
} |
} |
1104 |
for(i = 0 ; i < 8 ; i++) { |
for(i = 0 ; i < 8 ; i++) { |
1105 |
memcpy(dU,sU,w2); |
memcpy(dU,sU,w2); |
1106 |
dU += out_frm->stride_u; |
dU += out_frm->stride[1]; |
1107 |
sU += std2; |
sU += std2; |
1108 |
} |
} |
1109 |
for(i = 0 ; i < 8 ; i++) { |
for(i = 0 ; i < 8 ; i++) { |
1110 |
memcpy(dV,sV,w2); |
memcpy(dV,sV,w2); |
1111 |
dV += out_frm->stride_v; |
dV += out_frm->stride[2]; |
1112 |
sV += std2; |
sV += std2; |
1113 |
} |
} |
1114 |
} |
} |
1115 |
|
|
1116 |
|
|
1117 |
|
void |
1118 |
|
image_clear(IMAGE * img, int width, int height, int edged_width, |
1119 |
|
int y, int u, int v) |
1120 |
|
{ |
1121 |
|
uint8_t * p; |
1122 |
|
int i; |
1123 |
|
|
1124 |
|
p = img->y; |
1125 |
|
for (i = 0; i < height; i++) { |
1126 |
|
memset(p, y, width); |
1127 |
|
p += edged_width; |
1128 |
|
} |
1129 |
|
|
1130 |
|
p = img->u; |
1131 |
|
for (i = 0; i < height/2; i++) { |
1132 |
|
memset(p, u, width/2); |
1133 |
|
p += edged_width/2; |
1134 |
|
} |
1135 |
|
|
1136 |
|
p = img->v; |
1137 |
|
for (i = 0; i < height/2; i++) { |
1138 |
|
memset(p, v, width/2); |
1139 |
|
p += edged_width/2; |
1140 |
|
} |
1141 |
|
} |
1142 |
|
|
1143 |
|
|
1144 |
|
/* reduced resolution deblocking filter |
1145 |
|
block = block size (16=rrv, 8=full resolution) |
1146 |
|
flags = XVID_DEC_YDEBLOCK|XVID_DEC_UVDEBLOCK |
1147 |
|
*/ |
1148 |
|
void |
1149 |
|
image_deblock_rrv(IMAGE * img, int edged_width, |
1150 |
|
const MACROBLOCK * mbs, int mb_width, int mb_height, int mb_stride, |
1151 |
|
int block, int flags) |
1152 |
|
{ |
1153 |
|
const int edged_width2 = edged_width /2; |
1154 |
|
const int nblocks = block / 8; /* skals code uses 8pixel block uints */ |
1155 |
|
int i,j; |
1156 |
|
|
1157 |
|
/* luma: j,i in block units */ |
1158 |
|
|
1159 |
|
for (j = 1; j < mb_height*2; j++) /* horizontal deblocking */ |
1160 |
|
for (i = 0; i < mb_width*2; i++) |
1161 |
|
{ |
1162 |
|
if (mbs[(j-1)/2*mb_stride + (i/2)].mode != MODE_NOT_CODED || |
1163 |
|
mbs[(j+0)/2*mb_stride + (i/2)].mode != MODE_NOT_CODED) |
1164 |
|
{ |
1165 |
|
hfilter_31(img->y + (j*block - 1)*edged_width + i*block, |
1166 |
|
img->y + (j*block + 0)*edged_width + i*block, nblocks); |
1167 |
|
} |
1168 |
|
} |
1169 |
|
|
1170 |
|
for (j = 0; j < mb_height*2; j++) /* vertical deblocking */ |
1171 |
|
for (i = 1; i < mb_width*2; i++) |
1172 |
|
{ |
1173 |
|
if (mbs[(j/2)*mb_stride + (i-1)/2].mode != MODE_NOT_CODED || |
1174 |
|
mbs[(j/2)*mb_stride + (i+0)/2].mode != MODE_NOT_CODED) |
1175 |
|
{ |
1176 |
|
vfilter_31(img->y + (j*block)*edged_width + i*block - 1, |
1177 |
|
img->y + (j*block)*edged_width + i*block + 0, |
1178 |
|
edged_width, nblocks); |
1179 |
|
} |
1180 |
|
} |
1181 |
|
|
1182 |
|
|
1183 |
|
|
1184 |
|
/* chroma */ |
1185 |
|
|
1186 |
|
for (j = 1; j < mb_height; j++) /* horizontal deblocking */ |
1187 |
|
for (i = 0; i < mb_width; i++) |
1188 |
|
{ |
1189 |
|
if (mbs[(j-1)*mb_stride + i].mode != MODE_NOT_CODED || |
1190 |
|
mbs[(j+0)*mb_stride + i].mode != MODE_NOT_CODED) |
1191 |
|
{ |
1192 |
|
hfilter_31(img->u + (j*block - 1)*edged_width2 + i*block, |
1193 |
|
img->u + (j*block + 0)*edged_width2 + i*block, nblocks); |
1194 |
|
hfilter_31(img->v + (j*block - 1)*edged_width2 + i*block, |
1195 |
|
img->v + (j*block + 0)*edged_width2 + i*block, nblocks); |
1196 |
|
} |
1197 |
|
} |
1198 |
|
|
1199 |
|
for (j = 0; j < mb_height; j++) /* vertical deblocking */ |
1200 |
|
for (i = 1; i < mb_width; i++) |
1201 |
|
{ |
1202 |
|
if (mbs[j*mb_stride + i - 1].mode != MODE_NOT_CODED || |
1203 |
|
mbs[j*mb_stride + i + 0].mode != MODE_NOT_CODED) |
1204 |
|
{ |
1205 |
|
vfilter_31(img->u + (j*block)*edged_width2 + i*block - 1, |
1206 |
|
img->u + (j*block)*edged_width2 + i*block + 0, |
1207 |
|
edged_width2, nblocks); |
1208 |
|
vfilter_31(img->v + (j*block)*edged_width2 + i*block - 1, |
1209 |
|
img->v + (j*block)*edged_width2 + i*block + 0, |
1210 |
|
edged_width2, nblocks); |
1211 |
|
} |
1212 |
|
} |
1213 |
|
|
1214 |
|
|
1215 |
|
} |
1216 |
|
|