1 |
/************************************************************************** |
/************************************************************************** |
2 |
* |
* |
3 |
* XVID MPEG-4 VIDEO CODEC |
* XVID MPEG-4 VIDEO CODEC |
4 |
* image stuff |
* - Image management functions - |
5 |
* |
* |
6 |
* This program is an implementation of a part of one or more MPEG-4 |
* Copyright(C) 2001-2004 Peter Ross <pross@xvid.org> |
|
* Video tools as specified in ISO/IEC 14496-2 standard. Those intending |
|
|
* to use this software module in hardware or software products are |
|
|
* advised that its use may infringe existing patents or copyrights, and |
|
|
* any such use would be at such party's own risk. The original |
|
|
* developer of this software module and his/her company, and subsequent |
|
|
* editors and their companies, will have no liability for use of this |
|
|
* software or modifications or derivatives thereof. |
|
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 |
*************************************************************************/ |
* $Id: image.c,v 1.34 2005-12-17 12:04:52 syskin Exp $ |
|
|
|
|
/************************************************************************** |
|
|
* |
|
|
* History: |
|
23 |
* |
* |
24 |
* 05.10.2002 support for interpolated images in qpel mode - Isibaar |
****************************************************************************/ |
|
* 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 - Isibaar |
|
|
* 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 "../global.h" // XVID_CSP_XXX's |
#include "../global.h" /* XVID_CSP_XXX's */ |
31 |
#include "../xvid.h" // XVID_CSP_XXX's |
#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 "reduced.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 |
#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, |
134 |
uint32_t edged_width, |
uint32_t edged_width, |
135 |
uint32_t edged_height, |
uint32_t edged_height, |
136 |
uint32_t width, |
uint32_t width, |
137 |
uint32_t height) |
uint32_t height, |
138 |
|
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++) { |
159 |
memset(dst, *src, EDGE_SIZE); |
memset(dst, *src, EDGE_SIZE); |
160 |
memcpy(dst + EDGE_SIZE, src, width); |
memcpy(dst + EDGE_SIZE, src, 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 |
|
|
|
// bframe encoding requires image-based u,v interpolation |
|
239 |
void |
void |
240 |
image_interpolate(const IMAGE * refn, |
image_interpolate(const uint8_t * refn, |
241 |
IMAGE * refh, |
uint8_t * refh, |
242 |
IMAGE * refv, |
uint8_t * refv, |
243 |
IMAGE * refhv, |
uint8_t * refhv, |
244 |
uint32_t edged_width, |
uint32_t edged_width, |
245 |
uint32_t edged_height, |
uint32_t edged_height, |
246 |
uint32_t quarterpel, |
uint32_t quarterpel, |
247 |
uint32_t rounding) |
uint32_t rounding) |
248 |
{ |
{ |
249 |
const uint32_t offset = EDGE_SIZE2 * (edged_width + 1); // we only interpolate half of the edge area |
const uint32_t offset = EDGE_SIZE2 * (edged_width + 1); /* we only interpolate half of the edge area */ |
250 |
const uint32_t stride_add = 7 * edged_width; |
const uint32_t stride_add = 7 * edged_width; |
|
/* |
|
|
#ifdef BFRAMES |
|
|
const uint32_t edged_width2 = edged_width / 2; |
|
|
const uint32_t edged_height2 = edged_height / 2; |
|
|
const uint32_t offset2 = EDGE_SIZE2 * (edged_width2 + 1); |
|
|
const uint32_t stride_add2 = 7 * edged_width2; |
|
|
#endif |
|
|
*/ |
|
|
uint8_t *n_ptr, *h_ptr, *v_ptr, *hv_ptr; |
|
|
uint32_t x, y; |
|
251 |
|
|
252 |
|
uint8_t *n_ptr; |
253 |
|
uint8_t *h_ptr, *v_ptr, *hv_ptr; |
254 |
|
uint32_t x, y; |
255 |
|
|
256 |
n_ptr = refn->y; |
n_ptr = (uint8_t*)refn; |
257 |
h_ptr = refh->y; |
h_ptr = refh; |
258 |
v_ptr = refv->y; |
v_ptr = refhv; |
|
hv_ptr = refhv->y; |
|
259 |
|
|
260 |
n_ptr -= offset; |
n_ptr -= offset; |
261 |
h_ptr -= offset; |
h_ptr -= offset; |
262 |
v_ptr -= offset; |
v_ptr -= offset; |
|
hv_ptr -= offset; |
|
263 |
|
|
264 |
|
/* Note we initialize the hv pointer later, as we can optimize code a bit |
265 |
|
* doing it down to up in quarterpel and up to down in halfpel */ |
266 |
if(quarterpel) { |
if(quarterpel) { |
267 |
|
|
268 |
for (y = 0; y < (edged_height - EDGE_SIZE); y += 8) { |
for (y = 0; y < (edged_height - EDGE_SIZE); y += 8) { |
284 |
n_ptr += stride_add; |
n_ptr += stride_add; |
285 |
} |
} |
286 |
|
|
287 |
h_ptr = refh->y; |
h_ptr = refh + (edged_height - EDGE_SIZE - EDGE_SIZE2)*edged_width - EDGE_SIZE2; |
288 |
h_ptr -= offset; |
hv_ptr = refhv + (edged_height - EDGE_SIZE - EDGE_SIZE2)*edged_width - EDGE_SIZE2; |
289 |
|
|
290 |
for (y = 0; y < (edged_height - EDGE_SIZE); y = y + 8) { |
for (y = 0; y < (edged_height - EDGE_SIZE); y = y + 8) { |
291 |
|
hv_ptr -= stride_add; |
292 |
|
h_ptr -= stride_add; |
293 |
|
hv_ptr -= EDGE_SIZE; |
294 |
|
h_ptr -= EDGE_SIZE; |
295 |
|
|
296 |
for (x = 0; x < (edged_width - EDGE_SIZE); x = x + 8) { |
for (x = 0; x < (edged_width - EDGE_SIZE); x = x + 8) { |
297 |
|
hv_ptr -= 8; |
298 |
|
h_ptr -= 8; |
299 |
interpolate8x8_6tap_lowpass_v(hv_ptr, h_ptr, edged_width, rounding); |
interpolate8x8_6tap_lowpass_v(hv_ptr, h_ptr, edged_width, rounding); |
|
hv_ptr += 8; |
|
|
h_ptr += 8; |
|
300 |
} |
} |
|
|
|
|
hv_ptr += EDGE_SIZE; |
|
|
h_ptr += EDGE_SIZE; |
|
|
|
|
|
hv_ptr += stride_add; |
|
|
h_ptr += stride_add; |
|
301 |
} |
} |
302 |
} |
} else { |
303 |
else { |
|
304 |
|
hv_ptr = refhv; |
305 |
|
hv_ptr -= offset; |
306 |
|
|
307 |
for (y = 0; y < (edged_height - EDGE_SIZE); y += 8) { |
for (y = 0; y < (edged_height - EDGE_SIZE); y += 8) { |
308 |
for (x = 0; x < (edged_width - EDGE_SIZE); x += 8) { |
for (x = 0; x < (edged_width - EDGE_SIZE); x += 8) { |
327 |
n_ptr += stride_add; |
n_ptr += stride_add; |
328 |
} |
} |
329 |
} |
} |
|
/* |
|
|
#ifdef BFRAMES |
|
|
n_ptr = refn->u; |
|
|
h_ptr = refh->u; |
|
|
v_ptr = refv->u; |
|
|
hv_ptr = refhv->u; |
|
|
|
|
|
n_ptr -= offset2; |
|
|
h_ptr -= offset2; |
|
|
v_ptr -= offset2; |
|
|
hv_ptr -= offset2; |
|
|
|
|
|
for (y = 0; y < edged_height2; y += 8) { |
|
|
for (x = 0; x < edged_width2; x += 8) { |
|
|
interpolate8x8_halfpel_h(h_ptr, n_ptr, edged_width2, rounding); |
|
|
interpolate8x8_halfpel_v(v_ptr, n_ptr, edged_width2, rounding); |
|
|
interpolate8x8_halfpel_hv(hv_ptr, n_ptr, edged_width2, rounding); |
|
|
|
|
|
n_ptr += 8; |
|
|
h_ptr += 8; |
|
|
v_ptr += 8; |
|
|
hv_ptr += 8; |
|
|
} |
|
|
h_ptr += stride_add2; |
|
|
v_ptr += stride_add2; |
|
|
hv_ptr += stride_add2; |
|
|
n_ptr += stride_add2; |
|
|
} |
|
|
|
|
|
n_ptr = refn->v; |
|
|
h_ptr = refh->v; |
|
|
v_ptr = refv->v; |
|
|
hv_ptr = refhv->v; |
|
|
|
|
|
n_ptr -= offset2; |
|
|
h_ptr -= offset2; |
|
|
v_ptr -= offset2; |
|
|
hv_ptr -= offset2; |
|
|
|
|
|
for (y = 0; y < edged_height2; y = y + 8) { |
|
|
for (x = 0; x < edged_width2; x = x + 8) { |
|
|
interpolate8x8_halfpel_h(h_ptr, n_ptr, edged_width2, rounding); |
|
|
interpolate8x8_halfpel_v(v_ptr, n_ptr, edged_width2, rounding); |
|
|
interpolate8x8_halfpel_hv(hv_ptr, n_ptr, edged_width2, rounding); |
|
|
|
|
|
n_ptr += 8; |
|
|
h_ptr += 8; |
|
|
v_ptr += 8; |
|
|
hv_ptr += 8; |
|
|
} |
|
|
h_ptr += stride_add2; |
|
|
v_ptr += stride_add2; |
|
|
hv_ptr += stride_add2; |
|
|
n_ptr += stride_add2; |
|
|
} |
|
|
#endif |
|
|
*/ |
|
|
/* |
|
|
interpolate_halfpel_h( |
|
|
refh->y - offset, |
|
|
refn->y - offset, |
|
|
edged_width, edged_height, |
|
|
rounding); |
|
|
|
|
|
interpolate_halfpel_v( |
|
|
refv->y - offset, |
|
|
refn->y - offset, |
|
|
edged_width, edged_height, |
|
|
rounding); |
|
|
|
|
|
interpolate_halfpel_hv( |
|
|
refhv->y - offset, |
|
|
refn->y - offset, |
|
|
edged_width, edged_height, |
|
|
rounding); |
|
|
*/ |
|
|
|
|
|
/* uv-image-based compensation |
|
|
offset = EDGE_SIZE2 * (edged_width / 2 + 1); |
|
|
|
|
|
interpolate_halfpel_h( |
|
|
refh->u - offset, |
|
|
refn->u - offset, |
|
|
edged_width / 2, edged_height / 2, |
|
|
rounding); |
|
|
|
|
|
interpolate_halfpel_v( |
|
|
refv->u - offset, |
|
|
refn->u - offset, |
|
|
edged_width / 2, edged_height / 2, |
|
|
rounding); |
|
|
|
|
|
interpolate_halfpel_hv( |
|
|
refhv->u - offset, |
|
|
refn->u - offset, |
|
|
edged_width / 2, edged_height / 2, |
|
|
rounding); |
|
|
|
|
|
|
|
|
interpolate_halfpel_h( |
|
|
refh->v - offset, |
|
|
refn->v - offset, |
|
|
edged_width / 2, edged_height / 2, |
|
|
rounding); |
|
|
|
|
|
interpolate_halfpel_v( |
|
|
refv->v - offset, |
|
|
refn->v - offset, |
|
|
edged_width / 2, edged_height / 2, |
|
|
rounding); |
|
|
|
|
|
interpolate_halfpel_hv( |
|
|
refhv->v - offset, |
|
|
refn->v - offset, |
|
|
edged_width / 2, edged_height / 2, |
|
|
rounding); |
|
|
*/ |
|
330 |
} |
} |
331 |
|
|
332 |
|
|
366 |
#undef IMG_V |
#undef IMG_V |
367 |
} |
} |
368 |
|
|
369 |
DPRINTF(DPRINTF_DEBUG,"chroma_optimized_pixels = %i/%i", pixels, width*height/4); |
DPRINTF(XVID_DEBUG_DEBUG,"chroma_optimized_pixels = %i/%i\n", pixels, width*height/4); |
370 |
} |
} |
371 |
|
|
372 |
|
|
418 |
uint32_t width, |
uint32_t width, |
419 |
int height, |
int height, |
420 |
uint32_t edged_width, |
uint32_t edged_width, |
421 |
uint8_t * src, |
uint8_t * src[4], |
422 |
int src_stride, |
int src_stride[4], |
423 |
int csp, |
int csp, |
424 |
int interlacing) |
int interlacing) |
425 |
{ |
{ |
426 |
const int edged_width2 = edged_width/2; |
const int edged_width2 = edged_width/2; |
427 |
const int width2 = width/2; |
const int width2 = width/2; |
428 |
const int height2 = height/2; |
const int height2 = height/2; |
429 |
//const int height_signed = (csp & XVID_CSP_VFLIP) ? -height : height; |
#if 0 |
430 |
|
const int height_signed = (csp & XVID_CSP_VFLIP) ? -height : height; |
431 |
|
#endif |
|
// int src_stride = width; |
|
|
|
|
|
// --- xvid 2.1 compatiblity patch --- |
|
|
// --- remove when xvid_dec_frame->stride equals real stride |
|
|
/* |
|
|
if ((csp & ~XVID_CSP_VFLIP) == XVID_CSP_RGB555 || |
|
|
(csp & ~XVID_CSP_VFLIP) == XVID_CSP_RGB565 || |
|
|
(csp & ~XVID_CSP_VFLIP) == XVID_CSP_YUY2 || |
|
|
(csp & ~XVID_CSP_VFLIP) == XVID_CSP_YVYU || |
|
|
(csp & ~XVID_CSP_VFLIP) == XVID_CSP_UYVY) |
|
|
{ |
|
|
src_stride *= 2; |
|
|
} |
|
|
else if ((csp & ~XVID_CSP_VFLIP) == XVID_CSP_RGB24) |
|
|
{ |
|
|
src_stride *= 3; |
|
|
} |
|
|
else if ((csp & ~XVID_CSP_VFLIP) == XVID_CSP_RGB32 || |
|
|
(csp & ~XVID_CSP_VFLIP) == XVID_CSP_ABGR || |
|
|
(csp & ~XVID_CSP_VFLIP) == XVID_CSP_RGBA) |
|
|
{ |
|
|
src_stride *= 4; |
|
|
} |
|
|
*/ |
|
|
// ^--- xvid 2.1 compatiblity fix ---^ |
|
432 |
|
|
433 |
switch (csp & ~XVID_CSP_VFLIP) { |
switch (csp & ~XVID_CSP_VFLIP) { |
434 |
case XVID_CSP_RGB555: |
case XVID_CSP_RGB555: |
435 |
safe_packed_conv( |
safe_packed_conv( |
436 |
src, src_stride, image->y, image->u, image->v, |
src[0], src_stride[0], image->y, image->u, image->v, |
437 |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
438 |
interlacing?rgb555i_to_yv12 :rgb555_to_yv12, |
interlacing?rgb555i_to_yv12 :rgb555_to_yv12, |
439 |
interlacing?rgb555i_to_yv12_c:rgb555_to_yv12_c, 2); |
interlacing?rgb555i_to_yv12_c:rgb555_to_yv12_c, 2); |
441 |
|
|
442 |
case XVID_CSP_RGB565: |
case XVID_CSP_RGB565: |
443 |
safe_packed_conv( |
safe_packed_conv( |
444 |
src, src_stride, image->y, image->u, image->v, |
src[0], src_stride[0], image->y, image->u, image->v, |
445 |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
446 |
interlacing?rgb565i_to_yv12 :rgb565_to_yv12, |
interlacing?rgb565i_to_yv12 :rgb565_to_yv12, |
447 |
interlacing?rgb565i_to_yv12_c:rgb565_to_yv12_c, 2); |
interlacing?rgb565i_to_yv12_c:rgb565_to_yv12_c, 2); |
448 |
break; |
break; |
449 |
|
|
450 |
|
|
451 |
case XVID_CSP_RGB24: |
case XVID_CSP_BGR: |
452 |
safe_packed_conv( |
safe_packed_conv( |
453 |
src, src_stride, image->y, image->u, image->v, |
src[0], src_stride[0], image->y, image->u, image->v, |
454 |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
455 |
interlacing?bgri_to_yv12 :bgr_to_yv12, |
interlacing?bgri_to_yv12 :bgr_to_yv12, |
456 |
interlacing?bgri_to_yv12_c:bgr_to_yv12_c, 3); |
interlacing?bgri_to_yv12_c:bgr_to_yv12_c, 3); |
457 |
break; |
break; |
458 |
|
|
459 |
case XVID_CSP_RGB32: |
case XVID_CSP_BGRA: |
460 |
safe_packed_conv( |
safe_packed_conv( |
461 |
src, src_stride, image->y, image->u, image->v, |
src[0], src_stride[0], image->y, image->u, image->v, |
462 |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
463 |
interlacing?bgrai_to_yv12 :bgra_to_yv12, |
interlacing?bgrai_to_yv12 :bgra_to_yv12, |
464 |
interlacing?bgrai_to_yv12_c:bgra_to_yv12_c, 4); |
interlacing?bgrai_to_yv12_c:bgra_to_yv12_c, 4); |
466 |
|
|
467 |
case XVID_CSP_ABGR : |
case XVID_CSP_ABGR : |
468 |
safe_packed_conv( |
safe_packed_conv( |
469 |
src, src_stride, image->y, image->u, image->v, |
src[0], src_stride[0], image->y, image->u, image->v, |
470 |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
471 |
interlacing?abgri_to_yv12 :abgr_to_yv12, |
interlacing?abgri_to_yv12 :abgr_to_yv12, |
472 |
interlacing?abgri_to_yv12_c:abgr_to_yv12_c, 4); |
interlacing?abgri_to_yv12_c:abgr_to_yv12_c, 4); |
474 |
|
|
475 |
case XVID_CSP_RGBA : |
case XVID_CSP_RGBA : |
476 |
safe_packed_conv( |
safe_packed_conv( |
477 |
src, src_stride, image->y, image->u, image->v, |
src[0], src_stride[0], image->y, image->u, image->v, |
478 |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
479 |
interlacing?rgbai_to_yv12 :rgba_to_yv12, |
interlacing?rgbai_to_yv12 :rgba_to_yv12, |
480 |
interlacing?rgbai_to_yv12_c:rgba_to_yv12_c, 4); |
interlacing?rgbai_to_yv12_c:rgba_to_yv12_c, 4); |
481 |
break; |
break; |
482 |
|
|
483 |
|
case XVID_CSP_ARGB: |
484 |
|
safe_packed_conv( |
485 |
|
src[0], src_stride[0], image->y, image->u, image->v, |
486 |
|
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
487 |
|
interlacing?argbi_to_yv12 : argb_to_yv12, |
488 |
|
interlacing?argbi_to_yv12_c: argb_to_yv12_c, 4); |
489 |
|
break; |
490 |
|
|
491 |
case XVID_CSP_YUY2: |
case XVID_CSP_YUY2: |
492 |
safe_packed_conv( |
safe_packed_conv( |
493 |
src, src_stride, image->y, image->u, image->v, |
src[0], src_stride[0], image->y, image->u, image->v, |
494 |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
495 |
interlacing?yuyvi_to_yv12 :yuyv_to_yv12, |
interlacing?yuyvi_to_yv12 :yuyv_to_yv12, |
496 |
interlacing?yuyvi_to_yv12_c:yuyv_to_yv12_c, 2); |
interlacing?yuyvi_to_yv12_c:yuyv_to_yv12_c, 2); |
498 |
|
|
499 |
case XVID_CSP_YVYU: /* u/v swapped */ |
case XVID_CSP_YVYU: /* u/v swapped */ |
500 |
safe_packed_conv( |
safe_packed_conv( |
501 |
src, src_stride, image->y, image->v, image->y, |
src[0], src_stride[0], image->y, image->v, image->u, |
502 |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
503 |
interlacing?yuyvi_to_yv12 :yuyv_to_yv12, |
interlacing?yuyvi_to_yv12 :yuyv_to_yv12, |
504 |
interlacing?yuyvi_to_yv12_c:yuyv_to_yv12_c, 2); |
interlacing?yuyvi_to_yv12_c:yuyv_to_yv12_c, 2); |
506 |
|
|
507 |
case XVID_CSP_UYVY: |
case XVID_CSP_UYVY: |
508 |
safe_packed_conv( |
safe_packed_conv( |
509 |
src, src_stride, image->y, image->u, image->v, |
src[0], src_stride[0], image->y, image->u, image->v, |
510 |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
511 |
interlacing?uyvyi_to_yv12 :uyvy_to_yv12, |
interlacing?uyvyi_to_yv12 :uyvy_to_yv12, |
512 |
interlacing?uyvyi_to_yv12_c:uyvy_to_yv12_c, 2); |
interlacing?uyvyi_to_yv12_c:uyvy_to_yv12_c, 2); |
513 |
break; |
break; |
514 |
|
|
515 |
case XVID_CSP_I420: |
case XVID_CSP_I420: /* YCbCr == YUV == internal colorspace for MPEG */ |
516 |
yv12_to_yv12(image->y, image->u, image->v, edged_width, edged_width2, |
yv12_to_yv12(image->y, image->u, image->v, edged_width, edged_width2, |
517 |
src, src + src_stride*height, src + src_stride*height + (src_stride/2)*height2, |
src[0], src[0] + src_stride[0]*height, src[0] + src_stride[0]*height + (src_stride[0]/2)*height2, |
518 |
src_stride, src_stride/2, width, height, (csp & XVID_CSP_VFLIP)); |
src_stride[0], src_stride[0]/2, width, height, (csp & XVID_CSP_VFLIP)); |
519 |
break |
break; |
520 |
; |
|
521 |
case XVID_CSP_YV12: /* u/v swapped */ |
case XVID_CSP_YV12: /* YCrCb == YVA == U and V plane swapped */ |
522 |
yv12_to_yv12(image->y, image->v, image->u, edged_width, edged_width2, |
yv12_to_yv12(image->y, image->v, image->u, edged_width, edged_width2, |
523 |
src, src + src_stride*height, src + src_stride*height + (src_stride/2)*height2, |
src[0], src[0] + src_stride[0]*height, src[0] + src_stride[0]*height + (src_stride[0]/2)*height2, |
524 |
src_stride, src_stride/2, width, height, (csp & XVID_CSP_VFLIP)); |
src_stride[0], src_stride[0]/2, width, height, (csp & XVID_CSP_VFLIP)); |
525 |
break; |
break; |
526 |
|
|
527 |
case XVID_CSP_USER: |
case XVID_CSP_PLANAR: /* YCbCr with arbitrary pointers and different strides for Y and UV */ |
|
{ |
|
|
DEC_PICTURE * pic = (DEC_PICTURE*)src; |
|
528 |
yv12_to_yv12(image->y, image->u, image->v, edged_width, edged_width2, |
yv12_to_yv12(image->y, image->u, image->v, edged_width, edged_width2, |
529 |
pic->y, pic->u, pic->v, pic->stride_y, pic->stride_y, |
src[0], src[1], src[2], src_stride[0], src_stride[1], /* v: dst_stride[2] not yet supported */ |
530 |
width, height, (csp & XVID_CSP_VFLIP)); |
width, height, (csp & XVID_CSP_VFLIP)); |
|
} |
|
531 |
break; |
break; |
532 |
|
|
533 |
case XVID_CSP_NULL: |
case XVID_CSP_NULL: |
593 |
uint32_t width, |
uint32_t width, |
594 |
int height, |
int height, |
595 |
uint32_t edged_width, |
uint32_t edged_width, |
596 |
uint8_t * dst, |
uint8_t * dst[4], |
597 |
uint32_t dst_stride, |
int dst_stride[4], |
598 |
int csp, |
int csp, |
599 |
int interlacing) |
int interlacing) |
600 |
{ |
{ |
607 |
image_dump_yuvpgm(image, edged_width, width, height, "\\decode.pgm"); |
image_dump_yuvpgm(image, edged_width, width, height, "\\decode.pgm"); |
608 |
*/ |
*/ |
609 |
|
|
|
|
|
|
// --- xvid 2.1 compatiblity patch --- |
|
|
// --- remove when xvid_dec_frame->stride equals real stride |
|
|
/* |
|
|
if ((csp & ~XVID_CSP_VFLIP) == XVID_CSP_RGB555 || |
|
|
(csp & ~XVID_CSP_VFLIP) == XVID_CSP_RGB565 || |
|
|
(csp & ~XVID_CSP_VFLIP) == XVID_CSP_YUY2 || |
|
|
(csp & ~XVID_CSP_VFLIP) == XVID_CSP_YVYU || |
|
|
(csp & ~XVID_CSP_VFLIP) == XVID_CSP_UYVY) |
|
|
{ |
|
|
dst_stride *= 2; |
|
|
} |
|
|
else if ((csp & ~XVID_CSP_VFLIP) == XVID_CSP_RGB24) |
|
|
{ |
|
|
dst_stride *= 3; |
|
|
} |
|
|
else if ((csp & ~XVID_CSP_VFLIP) == XVID_CSP_RGB32 || |
|
|
(csp & ~XVID_CSP_VFLIP) == XVID_CSP_ABGR || |
|
|
(csp & ~XVID_CSP_VFLIP) == XVID_CSP_RGBA) |
|
|
{ |
|
|
dst_stride *= 4; |
|
|
} |
|
|
*/ |
|
|
// ^--- xvid 2.1 compatiblity fix ---^ |
|
|
|
|
|
|
|
610 |
switch (csp & ~XVID_CSP_VFLIP) { |
switch (csp & ~XVID_CSP_VFLIP) { |
611 |
case XVID_CSP_RGB555: |
case XVID_CSP_RGB555: |
612 |
safe_packed_conv( |
safe_packed_conv( |
613 |
dst, dst_stride, image->y, image->u, image->v, |
dst[0], dst_stride[0], image->y, image->u, image->v, |
614 |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
615 |
interlacing?yv12_to_rgb555i :yv12_to_rgb555, |
interlacing?yv12_to_rgb555i :yv12_to_rgb555, |
616 |
interlacing?yv12_to_rgb555i_c:yv12_to_rgb555_c, 2); |
interlacing?yv12_to_rgb555i_c:yv12_to_rgb555_c, 2); |
618 |
|
|
619 |
case XVID_CSP_RGB565: |
case XVID_CSP_RGB565: |
620 |
safe_packed_conv( |
safe_packed_conv( |
621 |
dst, dst_stride, image->y, image->u, image->v, |
dst[0], dst_stride[0], image->y, image->u, image->v, |
622 |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
623 |
interlacing?yv12_to_rgb565i :yv12_to_rgb565, |
interlacing?yv12_to_rgb565i :yv12_to_rgb565, |
624 |
interlacing?yv12_to_rgb565i_c:yv12_to_rgb565_c, 2); |
interlacing?yv12_to_rgb565i_c:yv12_to_rgb565_c, 2); |
625 |
return 0; |
return 0; |
626 |
|
|
627 |
case XVID_CSP_RGB24: |
case XVID_CSP_BGR: |
628 |
safe_packed_conv( |
safe_packed_conv( |
629 |
dst, dst_stride, image->y, image->u, image->v, |
dst[0], dst_stride[0], image->y, image->u, image->v, |
630 |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
631 |
interlacing?yv12_to_bgri :yv12_to_bgr, |
interlacing?yv12_to_bgri :yv12_to_bgr, |
632 |
interlacing?yv12_to_bgri_c:yv12_to_bgr_c, 3); |
interlacing?yv12_to_bgri_c:yv12_to_bgr_c, 3); |
633 |
return 0; |
return 0; |
634 |
|
|
635 |
case XVID_CSP_RGB32: |
case XVID_CSP_BGRA: |
636 |
safe_packed_conv( |
safe_packed_conv( |
637 |
dst, dst_stride, image->y, image->u, image->v, |
dst[0], dst_stride[0], image->y, image->u, image->v, |
638 |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
639 |
interlacing?yv12_to_bgrai :yv12_to_bgra, |
interlacing?yv12_to_bgrai :yv12_to_bgra, |
640 |
interlacing?yv12_to_bgrai_c:yv12_to_bgra_c, 4); |
interlacing?yv12_to_bgrai_c:yv12_to_bgra_c, 4); |
642 |
|
|
643 |
case XVID_CSP_ABGR: |
case XVID_CSP_ABGR: |
644 |
safe_packed_conv( |
safe_packed_conv( |
645 |
dst, dst_stride, image->y, image->u, image->v, |
dst[0], dst_stride[0], image->y, image->u, image->v, |
646 |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
647 |
interlacing?yv12_to_abgri :yv12_to_abgr, |
interlacing?yv12_to_abgri :yv12_to_abgr, |
648 |
interlacing?yv12_to_abgri_c:yv12_to_abgr_c, 4); |
interlacing?yv12_to_abgri_c:yv12_to_abgr_c, 4); |
650 |
|
|
651 |
case XVID_CSP_RGBA: |
case XVID_CSP_RGBA: |
652 |
safe_packed_conv( |
safe_packed_conv( |
653 |
dst, dst_stride, image->y, image->u, image->v, |
dst[0], dst_stride[0], image->y, image->u, image->v, |
654 |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
655 |
interlacing?yv12_to_rgbai :yv12_to_rgba, |
interlacing?yv12_to_rgbai :yv12_to_rgba, |
656 |
interlacing?yv12_to_rgbai_c:yv12_to_rgba_c, 4); |
interlacing?yv12_to_rgbai_c:yv12_to_rgba_c, 4); |
657 |
return 0; |
return 0; |
658 |
|
|
659 |
|
case XVID_CSP_ARGB: |
660 |
|
safe_packed_conv( |
661 |
|
dst[0], dst_stride[0], image->y, image->u, image->v, |
662 |
|
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
663 |
|
interlacing?yv12_to_argbi :yv12_to_argb, |
664 |
|
interlacing?yv12_to_argbi_c:yv12_to_argb_c, 4); |
665 |
|
return 0; |
666 |
|
|
667 |
case XVID_CSP_YUY2: |
case XVID_CSP_YUY2: |
668 |
safe_packed_conv( |
safe_packed_conv( |
669 |
dst, dst_stride, image->y, image->u, image->v, |
dst[0], dst_stride[0], image->y, image->u, image->v, |
670 |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
671 |
interlacing?yv12_to_yuyvi :yv12_to_yuyv, |
interlacing?yv12_to_yuyvi :yv12_to_yuyv, |
672 |
interlacing?yv12_to_yuyvi_c:yv12_to_yuyv_c, 2); |
interlacing?yv12_to_yuyvi_c:yv12_to_yuyv_c, 2); |
673 |
return 0; |
return 0; |
674 |
|
|
675 |
case XVID_CSP_YVYU: // u,v swapped |
case XVID_CSP_YVYU: /* u,v swapped */ |
676 |
safe_packed_conv( |
safe_packed_conv( |
677 |
dst, dst_stride, image->y, image->v, image->u, |
dst[0], dst_stride[0], image->y, image->v, image->u, |
678 |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
679 |
interlacing?yv12_to_yuyvi :yv12_to_yuyv, |
interlacing?yv12_to_yuyvi :yv12_to_yuyv, |
680 |
interlacing?yv12_to_yuyvi_c:yv12_to_yuyv_c, 2); |
interlacing?yv12_to_yuyvi_c:yv12_to_yuyv_c, 2); |
682 |
|
|
683 |
case XVID_CSP_UYVY: |
case XVID_CSP_UYVY: |
684 |
safe_packed_conv( |
safe_packed_conv( |
685 |
dst, dst_stride, image->y, image->u, image->v, |
dst[0], dst_stride[0], image->y, image->u, image->v, |
686 |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
687 |
interlacing?yv12_to_uyvyi :yv12_to_uyvy, |
interlacing?yv12_to_uyvyi :yv12_to_uyvy, |
688 |
interlacing?yv12_to_uyvyi_c:yv12_to_uyvy_c, 2); |
interlacing?yv12_to_uyvyi_c:yv12_to_uyvy_c, 2); |
689 |
return 0; |
return 0; |
690 |
|
|
691 |
case XVID_CSP_I420: |
case XVID_CSP_I420: /* YCbCr == YUV == internal colorspace for MPEG */ |
692 |
yv12_to_yv12(dst, dst + dst_stride*height, dst + dst_stride*height + (dst_stride/2)*height2, |
yv12_to_yv12(dst[0], dst[0] + dst_stride[0]*height, dst[0] + dst_stride[0]*height + (dst_stride[0]/2)*height2, |
693 |
dst_stride, dst_stride/2, |
dst_stride[0], dst_stride[0]/2, |
694 |
image->y, image->u, image->v, edged_width, edged_width2, |
image->y, image->u, image->v, edged_width, edged_width2, |
695 |
width, height, (csp & XVID_CSP_VFLIP)); |
width, height, (csp & XVID_CSP_VFLIP)); |
696 |
return 0; |
return 0; |
697 |
|
|
698 |
case XVID_CSP_YV12: // u,v swapped |
case XVID_CSP_YV12: /* YCrCb == YVU == U and V plane swapped */ |
699 |
yv12_to_yv12(dst, dst + dst_stride*height, dst + dst_stride*height + (dst_stride/2)*height2, |
yv12_to_yv12(dst[0], dst[0] + dst_stride[0]*height, dst[0] + dst_stride[0]*height + (dst_stride[0]/2)*height2, |
700 |
dst_stride, dst_stride/2, |
dst_stride[0], dst_stride[0]/2, |
701 |
image->y, image->v, image->u, edged_width, edged_width2, |
image->y, image->v, image->u, edged_width, edged_width2, |
702 |
width, height, (csp & XVID_CSP_VFLIP)); |
width, height, (csp & XVID_CSP_VFLIP)); |
703 |
return 0; |
return 0; |
704 |
|
|
705 |
case XVID_CSP_USER: |
case XVID_CSP_PLANAR: /* YCbCr with arbitrary pointers and different strides for Y and UV */ |
706 |
{ |
yv12_to_yv12(dst[0], dst[1], dst[2], |
707 |
DEC_PICTURE * pic = (DEC_PICTURE*)dst; |
dst_stride[0], dst_stride[1], /* v: dst_stride[2] not yet supported */ |
708 |
pic->y = image->y; |
image->y, image->u, image->v, edged_width, edged_width2, |
709 |
pic->u = image->u; |
width, height, (csp & XVID_CSP_VFLIP)); |
710 |
pic->v = image->v; |
return 0; |
711 |
pic->stride_y = edged_width; |
|
712 |
pic->stride_uv = edged_width / 2; |
case XVID_CSP_INTERNAL : |
713 |
} |
dst[0] = image->y; |
714 |
|
dst[1] = image->u; |
715 |
|
dst[2] = image->v; |
716 |
|
dst_stride[0] = edged_width; |
717 |
|
dst_stride[1] = edged_width/2; |
718 |
|
dst_stride[2] = edged_width/2; |
719 |
return 0; |
return 0; |
720 |
|
|
721 |
case XVID_CSP_NULL: |
case XVID_CSP_NULL: |
722 |
case XVID_CSP_EXTERN: |
case XVID_CSP_SLICE: |
723 |
return 0; |
return 0; |
724 |
|
|
725 |
} |
} |
765 |
if (sse==0) |
if (sse==0) |
766 |
return 99.99F; |
return 99.99F; |
767 |
|
|
768 |
return 48.131F - 10*(float)log10((float)sse/(float)(pixels)); // log10(255*255)=4.8131 |
return 48.131F - 10*(float)log10((float)sse/(float)(pixels)); /* log10(255*255)=4.8131 */ |
769 |
|
|
770 |
} |
} |
771 |
|
|
775 |
uint16_t width, |
uint16_t width, |
776 |
uint16_t height) |
uint16_t height) |
777 |
{ |
{ |
778 |
int diff, x, y; |
int y, bwidth, bheight; |
779 |
long sse=0; |
long sse=0; |
780 |
|
|
781 |
for (y = 0; y < height; y++) { |
bwidth = width & (~0x07); |
782 |
|
bheight = height & (~0x07); |
783 |
|
|
784 |
|
/* Compute the 8x8 integer part */ |
785 |
|
for (y = 0; y<bheight; y += 8) { |
786 |
|
int x; |
787 |
|
|
788 |
|
/* Compute sse for the band */ |
789 |
|
for (x = 0; x<bwidth; x += 8) |
790 |
|
sse += sse8_8bit(orig + x, recon + x, stride); |
791 |
|
|
792 |
|
/* remaining pixels of the 8 pixels high band */ |
793 |
|
for (x = bwidth; x < width; x++) { |
794 |
|
int diff; |
795 |
|
diff = *(orig + 0*stride + x) - *(recon + 0*stride + x); |
796 |
|
sse += diff * diff; |
797 |
|
diff = *(orig + 1*stride + x) - *(recon + 1*stride + x); |
798 |
|
sse += diff * diff; |
799 |
|
diff = *(orig + 2*stride + x) - *(recon + 2*stride + x); |
800 |
|
sse += diff * diff; |
801 |
|
diff = *(orig + 3*stride + x) - *(recon + 3*stride + x); |
802 |
|
sse += diff * diff; |
803 |
|
diff = *(orig + 4*stride + x) - *(recon + 4*stride + x); |
804 |
|
sse += diff * diff; |
805 |
|
diff = *(orig + 5*stride + x) - *(recon + 5*stride + x); |
806 |
|
sse += diff * diff; |
807 |
|
diff = *(orig + 6*stride + x) - *(recon + 6*stride + x); |
808 |
|
sse += diff * diff; |
809 |
|
diff = *(orig + 7*stride + x) - *(recon + 7*stride + x); |
810 |
|
sse += diff * diff; |
811 |
|
} |
812 |
|
|
813 |
|
orig += 8*stride; |
814 |
|
recon += 8*stride; |
815 |
|
} |
816 |
|
|
817 |
|
/* Compute the down rectangle sse */ |
818 |
|
for (y = bheight; y < height; y++) { |
819 |
|
int x; |
820 |
for (x = 0; x < width; x++) { |
for (x = 0; x < width; x++) { |
821 |
|
int diff; |
822 |
diff = *(orig + x) - *(recon + x); |
diff = *(orig + x) - *(recon + x); |
823 |
sse += diff * diff; |
sse += diff * diff; |
824 |
} |
} |
825 |
orig += stride; |
orig += stride; |
826 |
recon += stride; |
recon += stride; |
827 |
} |
} |
828 |
return sse; |
|
829 |
|
return (sse); |
830 |
} |
} |
831 |
|
|
832 |
/* |
#if 0 |
833 |
|
|
834 |
#include <stdio.h> |
#include <stdio.h> |
835 |
#include <string.h> |
#include <string.h> |
853 |
} |
} |
854 |
|
|
855 |
|
|
856 |
// dump image+edges to yuv pgm files |
/* dump image+edges to yuv pgm files */ |
857 |
|
|
858 |
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) |
859 |
{ |
{ |
876 |
|
|
877 |
return 0; |
return 0; |
878 |
} |
} |
879 |
*/ |
#endif |
880 |
|
|
881 |
|
|
882 |
|
|
941 |
|
|
942 |
for (y = 0; y < height; y++) |
for (y = 0; y < height; y++) |
943 |
for (x = 0; x < width; x++) |
for (x = 0; x < width; x++) |
944 |
sum += ABS(img1->y[x + y * stride] - img2->y[x + y * stride]); |
sum += abs(img1->y[x + y * stride] - img2->y[x + y * stride]); |
945 |
|
|
946 |
for (y = 0; y < height2; y++) |
for (y = 0; y < height2; y++) |
947 |
for (x = 0; x < width2; x++) |
for (x = 0; x < width2; x++) |
948 |
sum += ABS(img1->u[x + y * stride2] - img2->u[x + y * stride2]); |
sum += abs(img1->u[x + y * stride2] - img2->u[x + y * stride2]); |
949 |
|
|
950 |
for (y = 0; y < height2; y++) |
for (y = 0; y < height2; y++) |
951 |
for (x = 0; x < width2; x++) |
for (x = 0; x < width2; x++) |
952 |
sum += ABS(img1->v[x + y * stride2] - img2->v[x + y * stride2]); |
sum += abs(img1->v[x + y * stride2] - img2->v[x + y * stride2]); |
953 |
|
|
954 |
return (float) sum / (width * height * 3 / 2); |
return (float) sum / (width * height * 3 / 2); |
955 |
} |
} |
956 |
|
|
957 |
void |
void |
958 |
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) { |
959 |
uint8_t *dY,*dU,*dV,*sY,*sU,*sV; |
uint8_t *dY,*dU,*dV,*sY,*sU,*sV; |
960 |
int std2 = std >> 1; |
int stride2 = stride >> 1; |
961 |
int w = mbl << 4, w2,i; |
int w = mbl << 4, w2,i; |
962 |
|
|
963 |
if(w > width) |
if(w > width) |
964 |
w = width; |
w = width; |
965 |
w2 = w >> 1; |
w2 = w >> 1; |
966 |
|
|
967 |
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); |
968 |
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); |
969 |
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); |
970 |
sY = cur->y + (mby << 4) * std + (mbx << 4); |
sY = cur->y + (mby << 4) * stride + (mbx << 4); |
971 |
sU = cur->u + (mby << 3) * std2 + (mbx << 3); |
sU = cur->u + (mby << 3) * stride2 + (mbx << 3); |
972 |
sV = cur->v + (mby << 3) * std2 + (mbx << 3); |
sV = cur->v + (mby << 3) * stride2 + (mbx << 3); |
973 |
|
|
974 |
for(i = 0 ; i < 16 ; i++) { |
for(i = 0 ; i < 16 ; i++) { |
975 |
memcpy(dY,sY,w); |
memcpy(dY,sY,w); |
976 |
dY += out_frm->stride_y; |
dY += out_frm->stride[0]; |
977 |
sY += std; |
sY += stride; |
978 |
} |
} |
979 |
for(i = 0 ; i < 8 ; i++) { |
for(i = 0 ; i < 8 ; i++) { |
980 |
memcpy(dU,sU,w2); |
memcpy(dU,sU,w2); |
981 |
dU += out_frm->stride_u; |
dU += out_frm->stride[1]; |
982 |
sU += std2; |
sU += stride2; |
983 |
} |
} |
984 |
for(i = 0 ; i < 8 ; i++) { |
for(i = 0 ; i < 8 ; i++) { |
985 |
memcpy(dV,sV,w2); |
memcpy(dV,sV,w2); |
986 |
dV += out_frm->stride_v; |
dV += out_frm->stride[2]; |
987 |
sV += std2; |
sV += stride2; |
988 |
} |
} |
989 |
} |
} |
990 |
|
|
1014 |
p += edged_width/2; |
p += edged_width/2; |
1015 |
} |
} |
1016 |
} |
} |
|
|
|
|
|
|
|
/* reduced resolution deblocking filter |
|
|
block = block size (16=rrv, 8=full resolution) |
|
|
flags = XVID_DEC_YDEBLOCK|XVID_DEC_UVDEBLOCK |
|
|
*/ |
|
|
void |
|
|
image_deblock_rrv(IMAGE * img, int edged_width, |
|
|
const MACROBLOCK * mbs, int mb_width, int mb_height, int mb_stride, |
|
|
int block, int flags) |
|
|
{ |
|
|
const int edged_width2 = edged_width /2; |
|
|
const int nblocks = block / 8; /* skals code uses 8pixel block uints */ |
|
|
int i,j; |
|
|
|
|
|
/* luma: j,i in block units */ |
|
|
if ((flags & XVID_DEC_DEBLOCKY)) |
|
|
{ |
|
|
for (j = 1; j < mb_height*2; j++) /* horizontal deblocking */ |
|
|
for (i = 0; i < mb_width*2; i++) |
|
|
{ |
|
|
if (mbs[(j-1)/2*mb_stride + (i/2)].mode != MODE_NOT_CODED || |
|
|
mbs[(j+0)/2*mb_stride + (i/2)].mode != MODE_NOT_CODED) |
|
|
{ |
|
|
hfilter_31(img->y + (j*block - 1)*edged_width + i*block, |
|
|
img->y + (j*block + 0)*edged_width + i*block, nblocks); |
|
|
} |
|
|
} |
|
|
|
|
|
for (j = 0; j < mb_height*2; j++) /* vertical deblocking */ |
|
|
for (i = 1; i < mb_width*2; i++) |
|
|
{ |
|
|
if (mbs[(j/2)*mb_stride + (i-1)/2].mode != MODE_NOT_CODED || |
|
|
mbs[(j/2)*mb_stride + (i+0)/2].mode != MODE_NOT_CODED) |
|
|
{ |
|
|
vfilter_31(img->y + (j*block)*edged_width + i*block - 1, |
|
|
img->y + (j*block)*edged_width + i*block + 0, |
|
|
edged_width, nblocks); |
|
|
} |
|
|
} |
|
|
} |
|
|
|
|
|
|
|
|
/* chroma */ |
|
|
if ((flags & XVID_DEC_DEBLOCKUV)) |
|
|
{ |
|
|
for (j = 1; j < mb_height; j++) /* horizontal deblocking */ |
|
|
for (i = 0; i < mb_width; i++) |
|
|
{ |
|
|
if (mbs[(j-1)*mb_stride + i].mode != MODE_NOT_CODED || |
|
|
mbs[(j+0)*mb_stride + i].mode != MODE_NOT_CODED) |
|
|
{ |
|
|
hfilter_31(img->u + (j*block - 1)*edged_width2 + i*block, |
|
|
img->u + (j*block + 0)*edged_width2 + i*block, nblocks); |
|
|
hfilter_31(img->v + (j*block - 1)*edged_width2 + i*block, |
|
|
img->v + (j*block + 0)*edged_width2 + i*block, nblocks); |
|
|
} |
|
|
} |
|
|
|
|
|
for (j = 0; j < mb_height; j++) /* vertical deblocking */ |
|
|
for (i = 1; i < mb_width; i++) |
|
|
{ |
|
|
if (mbs[j*mb_stride + i - 1].mode != MODE_NOT_CODED || |
|
|
mbs[j*mb_stride + i + 0].mode != MODE_NOT_CODED) |
|
|
{ |
|
|
vfilter_31(img->u + (j*block)*edged_width2 + i*block - 1, |
|
|
img->u + (j*block)*edged_width2 + i*block + 0, |
|
|
edged_width2, nblocks); |
|
|
vfilter_31(img->v + (j*block)*edged_width2 + i*block - 1, |
|
|
img->v + (j*block)*edged_width2 + i*block + 0, |
|
|
edged_width2, nblocks); |
|
|
} |
|
|
} |
|
|
} |
|
|
|
|
|
} |
|
|
|
|