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) 2001-2003 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 |
*************************************************************************/ |
* $Id$ |
|
|
|
|
/************************************************************************** |
|
|
* |
|
|
* 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 |
|
|
30 |
#include "../portab.h" |
#include "../portab.h" |
31 |
#include "../global.h" // XVID_CSP_XXX's |
#include "../global.h" /* XVID_CSP_XXX's */ |
32 |
#include "../xvid.h" // XVID_CSP_XXX's |
#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 "reduced.h" |
#include "reduced.h" |
|
#include "../divx4.h" |
|
37 |
#include "../utils/mem_align.h" |
#include "../utils/mem_align.h" |
38 |
|
|
39 |
#include "font.h" // XXX: remove later |
#include "font.h" /* XXX: remove later */ |
40 |
|
|
41 |
#define SAFETY 64 |
#define SAFETY 64 |
42 |
#define EDGE_SIZE2 (EDGE_SIZE/2) |
#define EDGE_SIZE2 (EDGE_SIZE/2) |
49 |
{ |
{ |
50 |
const uint32_t edged_width2 = edged_width / 2; |
const uint32_t edged_width2 = edged_width / 2; |
51 |
const uint32_t edged_height2 = edged_height / 2; |
const uint32_t edged_height2 = edged_height / 2; |
|
uint32_t i; |
|
52 |
|
|
53 |
image->y = |
image->y = |
54 |
xvid_malloc(edged_width * (edged_height + 1) + SAFETY, CACHE_LINE); |
xvid_malloc(edged_width * (edged_height + 1) + SAFETY, CACHE_LINE); |
55 |
if (image->y == NULL) { |
if (image->y == NULL) { |
56 |
return -1; |
return -1; |
57 |
} |
} |
58 |
|
memset(image->y, 0, edged_width * (edged_height + 1) + SAFETY); |
|
for (i = 0; i < edged_width * edged_height + SAFETY; i++) { |
|
|
image->y[i] = 0; |
|
|
} |
|
59 |
|
|
60 |
image->u = xvid_malloc(edged_width2 * edged_height2 + SAFETY, CACHE_LINE); |
image->u = xvid_malloc(edged_width2 * edged_height2 + SAFETY, CACHE_LINE); |
61 |
if (image->u == NULL) { |
if (image->u == NULL) { |
62 |
xvid_free(image->y); |
xvid_free(image->y); |
63 |
|
image->y = NULL; |
64 |
return -1; |
return -1; |
65 |
} |
} |
66 |
|
memset(image->u, 0, edged_width2 * edged_height2 + SAFETY); |
67 |
|
|
68 |
image->v = xvid_malloc(edged_width2 * edged_height2 + SAFETY, CACHE_LINE); |
image->v = xvid_malloc(edged_width2 * edged_height2 + SAFETY, CACHE_LINE); |
69 |
if (image->v == NULL) { |
if (image->v == NULL) { |
70 |
xvid_free(image->u); |
xvid_free(image->u); |
71 |
|
image->u = NULL; |
72 |
xvid_free(image->y); |
xvid_free(image->y); |
73 |
|
image->y = NULL; |
74 |
return -1; |
return -1; |
75 |
} |
} |
76 |
|
memset(image->v, 0, edged_width2 * edged_height2 + SAFETY); |
77 |
|
|
78 |
image->y += EDGE_SIZE * edged_width + EDGE_SIZE; |
image->y += EDGE_SIZE * edged_width + EDGE_SIZE; |
79 |
image->u += EDGE_SIZE2 * edged_width2 + EDGE_SIZE2; |
image->u += EDGE_SIZE2 * edged_width2 + EDGE_SIZE2; |
93 |
|
|
94 |
if (image->y) { |
if (image->y) { |
95 |
xvid_free(image->y - (EDGE_SIZE * edged_width + EDGE_SIZE)); |
xvid_free(image->y - (EDGE_SIZE * edged_width + EDGE_SIZE)); |
96 |
|
image->y = NULL; |
97 |
} |
} |
98 |
if (image->u) { |
if (image->u) { |
99 |
xvid_free(image->u - (EDGE_SIZE2 * edged_width2 + EDGE_SIZE2)); |
xvid_free(image->u - (EDGE_SIZE2 * edged_width2 + EDGE_SIZE2)); |
100 |
|
image->u = NULL; |
101 |
} |
} |
102 |
if (image->v) { |
if (image->v) { |
103 |
xvid_free(image->v - (EDGE_SIZE2 * edged_width2 + EDGE_SIZE2)); |
xvid_free(image->v - (EDGE_SIZE2 * edged_width2 + EDGE_SIZE2)); |
104 |
|
image->v = NULL; |
105 |
} |
} |
106 |
} |
} |
107 |
|
|
110 |
image_swap(IMAGE * image1, |
image_swap(IMAGE * image1, |
111 |
IMAGE * image2) |
IMAGE * image2) |
112 |
{ |
{ |
113 |
uint8_t *tmp; |
SWAP(uint8_t*, image1->y, image2->y); |
114 |
|
SWAP(uint8_t*, image1->u, image2->u); |
115 |
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; |
|
116 |
} |
} |
117 |
|
|
118 |
|
|
136 |
uint32_t height) |
uint32_t height) |
137 |
{ |
{ |
138 |
const uint32_t edged_width2 = edged_width / 2; |
const uint32_t edged_width2 = edged_width / 2; |
139 |
const uint32_t width2 = width / 2; |
uint32_t width2; |
140 |
uint32_t i; |
uint32_t i; |
141 |
uint8_t *dst; |
uint8_t *dst; |
142 |
uint8_t *src; |
uint8_t *src; |
145 |
dst = image->y - (EDGE_SIZE + EDGE_SIZE * edged_width); |
dst = image->y - (EDGE_SIZE + EDGE_SIZE * edged_width); |
146 |
src = image->y; |
src = image->y; |
147 |
|
|
148 |
|
/* According to the Standard Clause 7.6.4, padding is done starting at 16 |
149 |
|
* pixel width and height multiples */ |
150 |
|
width = (width+15)&~15; |
151 |
|
height = (height+15)&~15; |
152 |
|
width2 = width/2; |
153 |
|
|
154 |
for (i = 0; i < EDGE_SIZE; i++) { |
for (i = 0; i < EDGE_SIZE; i++) { |
155 |
memset(dst, *src, EDGE_SIZE); |
memset(dst, *src, EDGE_SIZE); |
156 |
memcpy(dst + EDGE_SIZE, src, width); |
memcpy(dst + EDGE_SIZE, src, width); |
176 |
} |
} |
177 |
|
|
178 |
|
|
179 |
//U |
/* U */ |
180 |
dst = image->u - (EDGE_SIZE2 + EDGE_SIZE2 * edged_width2); |
dst = image->u - (EDGE_SIZE2 + EDGE_SIZE2 * edged_width2); |
181 |
src = image->u; |
src = image->u; |
182 |
|
|
204 |
} |
} |
205 |
|
|
206 |
|
|
207 |
// V |
/* V */ |
208 |
dst = image->v - (EDGE_SIZE2 + EDGE_SIZE2 * edged_width2); |
dst = image->v - (EDGE_SIZE2 + EDGE_SIZE2 * edged_width2); |
209 |
src = image->v; |
src = image->v; |
210 |
|
|
232 |
} |
} |
233 |
} |
} |
234 |
|
|
235 |
// bframe encoding requires image-based u,v interpolation |
/* bframe encoding requires image-based u,v interpolation */ |
236 |
void |
void |
237 |
image_interpolate(const IMAGE * refn, |
image_interpolate(const IMAGE * refn, |
238 |
IMAGE * refh, |
IMAGE * refh, |
243 |
uint32_t quarterpel, |
uint32_t quarterpel, |
244 |
uint32_t rounding) |
uint32_t rounding) |
245 |
{ |
{ |
246 |
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 */ |
247 |
const uint32_t stride_add = 7 * edged_width; |
const uint32_t stride_add = 7 * edged_width; |
248 |
/* |
#if 0 |
|
#ifdef BFRAMES |
|
249 |
const uint32_t edged_width2 = edged_width / 2; |
const uint32_t edged_width2 = edged_width / 2; |
250 |
const uint32_t edged_height2 = edged_height / 2; |
const uint32_t edged_height2 = edged_height / 2; |
251 |
const uint32_t offset2 = EDGE_SIZE2 * (edged_width2 + 1); |
const uint32_t offset2 = EDGE_SIZE2 * (edged_width2 + 1); |
252 |
const uint32_t stride_add2 = 7 * edged_width2; |
const uint32_t stride_add2 = 7 * edged_width2; |
253 |
#endif |
#endif |
|
*/ |
|
254 |
uint8_t *n_ptr, *h_ptr, *v_ptr, *hv_ptr; |
uint8_t *n_ptr, *h_ptr, *v_ptr, *hv_ptr; |
255 |
uint32_t x, y; |
uint32_t x, y; |
256 |
|
|
258 |
n_ptr = refn->y; |
n_ptr = refn->y; |
259 |
h_ptr = refh->y; |
h_ptr = refh->y; |
260 |
v_ptr = refv->y; |
v_ptr = refv->y; |
|
hv_ptr = refhv->y; |
|
261 |
|
|
262 |
n_ptr -= offset; |
n_ptr -= offset; |
263 |
h_ptr -= offset; |
h_ptr -= offset; |
264 |
v_ptr -= offset; |
v_ptr -= offset; |
|
hv_ptr -= offset; |
|
265 |
|
|
266 |
|
/* Note we initialize the hv pointer later, as we can optimize code a bit |
267 |
|
* doing it down to up in quarterpel and up to down in halfpel */ |
268 |
if(quarterpel) { |
if(quarterpel) { |
269 |
|
|
270 |
for (y = 0; y < (edged_height - EDGE_SIZE); y += 8) { |
for (y = 0; y < (edged_height - EDGE_SIZE); y += 8) { |
286 |
n_ptr += stride_add; |
n_ptr += stride_add; |
287 |
} |
} |
288 |
|
|
289 |
h_ptr = refh->y; |
h_ptr = refh->y + (edged_height - EDGE_SIZE - EDGE_SIZE2)*edged_width - EDGE_SIZE2; |
290 |
h_ptr -= offset; |
hv_ptr = refhv->y + (edged_height - EDGE_SIZE - EDGE_SIZE2)*edged_width - EDGE_SIZE2; |
291 |
|
|
292 |
for (y = 0; y < (edged_height - EDGE_SIZE); y = y + 8) { |
for (y = 0; y < (edged_height - EDGE_SIZE); y = y + 8) { |
293 |
|
hv_ptr -= stride_add; |
294 |
|
h_ptr -= stride_add; |
295 |
|
hv_ptr -= EDGE_SIZE; |
296 |
|
h_ptr -= EDGE_SIZE; |
297 |
|
|
298 |
for (x = 0; x < (edged_width - EDGE_SIZE); x = x + 8) { |
for (x = 0; x < (edged_width - EDGE_SIZE); x = x + 8) { |
299 |
|
hv_ptr -= 8; |
300 |
|
h_ptr -= 8; |
301 |
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; |
|
|
} |
|
|
|
|
|
hv_ptr += EDGE_SIZE; |
|
|
h_ptr += EDGE_SIZE; |
|
|
|
|
|
hv_ptr += stride_add; |
|
|
h_ptr += stride_add; |
|
302 |
} |
} |
303 |
} |
} |
304 |
else { |
} else { |
305 |
|
|
306 |
|
hv_ptr = refhv->y; |
307 |
|
hv_ptr -= offset; |
308 |
|
|
309 |
for (y = 0; y < (edged_height - EDGE_SIZE); y += 8) { |
for (y = 0; y < (edged_height - EDGE_SIZE); y += 8) { |
310 |
for (x = 0; x < (edged_width - EDGE_SIZE); x += 8) { |
for (x = 0; x < (edged_width - EDGE_SIZE); x += 8) { |
485 |
#undef IMG_V |
#undef IMG_V |
486 |
} |
} |
487 |
|
|
488 |
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); |
489 |
} |
} |
490 |
|
|
491 |
|
|
537 |
uint32_t width, |
uint32_t width, |
538 |
int height, |
int height, |
539 |
uint32_t edged_width, |
uint32_t edged_width, |
540 |
uint8_t * src, |
uint8_t * src[4], |
541 |
int src_stride, |
int src_stride[4], |
542 |
int csp, |
int csp, |
543 |
int interlacing) |
int interlacing) |
544 |
{ |
{ |
545 |
const int edged_width2 = edged_width/2; |
const int edged_width2 = edged_width/2; |
546 |
const int width2 = width/2; |
const int width2 = width/2; |
547 |
const int height2 = height/2; |
const int height2 = height/2; |
548 |
//const int height_signed = (csp & XVID_CSP_VFLIP) ? -height : height; |
#if 0 |
549 |
|
const int height_signed = (csp & XVID_CSP_VFLIP) ? -height : height; |
550 |
|
#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 ---^ |
|
551 |
|
|
552 |
switch (csp & ~XVID_CSP_VFLIP) { |
switch (csp & ~XVID_CSP_VFLIP) { |
553 |
case XVID_CSP_RGB555: |
case XVID_CSP_RGB555: |
554 |
safe_packed_conv( |
safe_packed_conv( |
555 |
src, src_stride, image->y, image->u, image->v, |
src[0], src_stride[0], image->y, image->u, image->v, |
556 |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
557 |
interlacing?rgb555i_to_yv12 :rgb555_to_yv12, |
interlacing?rgb555i_to_yv12 :rgb555_to_yv12, |
558 |
interlacing?rgb555i_to_yv12_c:rgb555_to_yv12_c, 2); |
interlacing?rgb555i_to_yv12_c:rgb555_to_yv12_c, 2); |
560 |
|
|
561 |
case XVID_CSP_RGB565: |
case XVID_CSP_RGB565: |
562 |
safe_packed_conv( |
safe_packed_conv( |
563 |
src, src_stride, image->y, image->u, image->v, |
src[0], src_stride[0], image->y, image->u, image->v, |
564 |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
565 |
interlacing?rgb565i_to_yv12 :rgb565_to_yv12, |
interlacing?rgb565i_to_yv12 :rgb565_to_yv12, |
566 |
interlacing?rgb565i_to_yv12_c:rgb565_to_yv12_c, 2); |
interlacing?rgb565i_to_yv12_c:rgb565_to_yv12_c, 2); |
567 |
break; |
break; |
568 |
|
|
569 |
|
|
570 |
case XVID_CSP_RGB24: |
case XVID_CSP_BGR: |
571 |
safe_packed_conv( |
safe_packed_conv( |
572 |
src, src_stride, image->y, image->u, image->v, |
src[0], src_stride[0], image->y, image->u, image->v, |
573 |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
574 |
interlacing?bgri_to_yv12 :bgr_to_yv12, |
interlacing?bgri_to_yv12 :bgr_to_yv12, |
575 |
interlacing?bgri_to_yv12_c:bgr_to_yv12_c, 3); |
interlacing?bgri_to_yv12_c:bgr_to_yv12_c, 3); |
576 |
break; |
break; |
577 |
|
|
578 |
case XVID_CSP_RGB32: |
case XVID_CSP_BGRA: |
579 |
safe_packed_conv( |
safe_packed_conv( |
580 |
src, src_stride, image->y, image->u, image->v, |
src[0], src_stride[0], image->y, image->u, image->v, |
581 |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
582 |
interlacing?bgrai_to_yv12 :bgra_to_yv12, |
interlacing?bgrai_to_yv12 :bgra_to_yv12, |
583 |
interlacing?bgrai_to_yv12_c:bgra_to_yv12_c, 4); |
interlacing?bgrai_to_yv12_c:bgra_to_yv12_c, 4); |
585 |
|
|
586 |
case XVID_CSP_ABGR : |
case XVID_CSP_ABGR : |
587 |
safe_packed_conv( |
safe_packed_conv( |
588 |
src, src_stride, image->y, image->u, image->v, |
src[0], src_stride[0], image->y, image->u, image->v, |
589 |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
590 |
interlacing?abgri_to_yv12 :abgr_to_yv12, |
interlacing?abgri_to_yv12 :abgr_to_yv12, |
591 |
interlacing?abgri_to_yv12_c:abgr_to_yv12_c, 4); |
interlacing?abgri_to_yv12_c:abgr_to_yv12_c, 4); |
593 |
|
|
594 |
case XVID_CSP_RGBA : |
case XVID_CSP_RGBA : |
595 |
safe_packed_conv( |
safe_packed_conv( |
596 |
src, src_stride, image->y, image->u, image->v, |
src[0], src_stride[0], image->y, image->u, image->v, |
597 |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
598 |
interlacing?rgbai_to_yv12 :rgba_to_yv12, |
interlacing?rgbai_to_yv12 :rgba_to_yv12, |
599 |
interlacing?rgbai_to_yv12_c:rgba_to_yv12_c, 4); |
interlacing?rgbai_to_yv12_c:rgba_to_yv12_c, 4); |
601 |
|
|
602 |
case XVID_CSP_YUY2: |
case XVID_CSP_YUY2: |
603 |
safe_packed_conv( |
safe_packed_conv( |
604 |
src, src_stride, image->y, image->u, image->v, |
src[0], src_stride[0], image->y, image->u, image->v, |
605 |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
606 |
interlacing?yuyvi_to_yv12 :yuyv_to_yv12, |
interlacing?yuyvi_to_yv12 :yuyv_to_yv12, |
607 |
interlacing?yuyvi_to_yv12_c:yuyv_to_yv12_c, 2); |
interlacing?yuyvi_to_yv12_c:yuyv_to_yv12_c, 2); |
609 |
|
|
610 |
case XVID_CSP_YVYU: /* u/v swapped */ |
case XVID_CSP_YVYU: /* u/v swapped */ |
611 |
safe_packed_conv( |
safe_packed_conv( |
612 |
src, src_stride, image->y, image->v, image->y, |
src[0], src_stride[0], image->y, image->v, image->y, |
613 |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
614 |
interlacing?yuyvi_to_yv12 :yuyv_to_yv12, |
interlacing?yuyvi_to_yv12 :yuyv_to_yv12, |
615 |
interlacing?yuyvi_to_yv12_c:yuyv_to_yv12_c, 2); |
interlacing?yuyvi_to_yv12_c:yuyv_to_yv12_c, 2); |
617 |
|
|
618 |
case XVID_CSP_UYVY: |
case XVID_CSP_UYVY: |
619 |
safe_packed_conv( |
safe_packed_conv( |
620 |
src, src_stride, image->y, image->u, image->v, |
src[0], src_stride[0], image->y, image->u, image->v, |
621 |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
622 |
interlacing?uyvyi_to_yv12 :uyvy_to_yv12, |
interlacing?uyvyi_to_yv12 :uyvy_to_yv12, |
623 |
interlacing?uyvyi_to_yv12_c:uyvy_to_yv12_c, 2); |
interlacing?uyvyi_to_yv12_c:uyvy_to_yv12_c, 2); |
624 |
break; |
break; |
625 |
|
|
626 |
case XVID_CSP_I420: |
case XVID_CSP_I420: /* YCrCb == internal colorspace for MPEG */ |
627 |
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, |
628 |
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, |
629 |
src_stride, src_stride/2, width, height, (csp & XVID_CSP_VFLIP)); |
src_stride[0], src_stride[0]/2, width, height, (csp & XVID_CSP_VFLIP)); |
630 |
break |
break; |
631 |
; |
|
632 |
case XVID_CSP_YV12: /* u/v swapped */ |
case XVID_CSP_YV12: /* YCbCr == U and V plane swapped */ |
633 |
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, |
634 |
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, |
635 |
src_stride, src_stride/2, width, height, (csp & XVID_CSP_VFLIP)); |
src_stride[0], src_stride[0]/2, width, height, (csp & XVID_CSP_VFLIP)); |
636 |
break; |
break; |
637 |
|
|
638 |
case XVID_CSP_USER: |
case XVID_CSP_USER : /* YCrCb with arbitrary pointers and different strides for Y and UV */ |
|
{ |
|
|
DEC_PICTURE * pic = (DEC_PICTURE*)src; |
|
639 |
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, |
640 |
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 */ |
641 |
width, height, (csp & XVID_CSP_VFLIP)); |
width, height, (csp & XVID_CSP_VFLIP)); |
|
} |
|
642 |
break; |
break; |
643 |
|
|
644 |
case XVID_CSP_NULL: |
case XVID_CSP_NULL: |
704 |
uint32_t width, |
uint32_t width, |
705 |
int height, |
int height, |
706 |
uint32_t edged_width, |
uint32_t edged_width, |
707 |
uint8_t * dst, |
uint8_t * dst[4], |
708 |
uint32_t dst_stride, |
uint32_t dst_stride[4], |
709 |
int csp, |
int csp, |
710 |
int interlacing) |
int interlacing) |
711 |
{ |
{ |
718 |
image_dump_yuvpgm(image, edged_width, width, height, "\\decode.pgm"); |
image_dump_yuvpgm(image, edged_width, width, height, "\\decode.pgm"); |
719 |
*/ |
*/ |
720 |
|
|
|
|
|
|
// --- 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 ---^ |
|
|
|
|
|
|
|
721 |
switch (csp & ~XVID_CSP_VFLIP) { |
switch (csp & ~XVID_CSP_VFLIP) { |
722 |
case XVID_CSP_RGB555: |
case XVID_CSP_RGB555: |
723 |
safe_packed_conv( |
safe_packed_conv( |
724 |
dst, dst_stride, image->y, image->u, image->v, |
dst[0], dst_stride[0], image->y, image->u, image->v, |
725 |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
726 |
interlacing?yv12_to_rgb555i :yv12_to_rgb555, |
interlacing?yv12_to_rgb555i :yv12_to_rgb555, |
727 |
interlacing?yv12_to_rgb555i_c:yv12_to_rgb555_c, 2); |
interlacing?yv12_to_rgb555i_c:yv12_to_rgb555_c, 2); |
729 |
|
|
730 |
case XVID_CSP_RGB565: |
case XVID_CSP_RGB565: |
731 |
safe_packed_conv( |
safe_packed_conv( |
732 |
dst, dst_stride, image->y, image->u, image->v, |
dst[0], dst_stride[0], image->y, image->u, image->v, |
733 |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
734 |
interlacing?yv12_to_rgb565i :yv12_to_rgb565, |
interlacing?yv12_to_rgb565i :yv12_to_rgb565, |
735 |
interlacing?yv12_to_rgb565i_c:yv12_to_rgb565_c, 2); |
interlacing?yv12_to_rgb565i_c:yv12_to_rgb565_c, 2); |
736 |
return 0; |
return 0; |
737 |
|
|
738 |
case XVID_CSP_RGB24: |
case XVID_CSP_BGR: |
739 |
safe_packed_conv( |
safe_packed_conv( |
740 |
dst, dst_stride, image->y, image->u, image->v, |
dst[0], dst_stride[0], image->y, image->u, image->v, |
741 |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
742 |
interlacing?yv12_to_bgri :yv12_to_bgr, |
interlacing?yv12_to_bgri :yv12_to_bgr, |
743 |
interlacing?yv12_to_bgri_c:yv12_to_bgr_c, 3); |
interlacing?yv12_to_bgri_c:yv12_to_bgr_c, 3); |
744 |
return 0; |
return 0; |
745 |
|
|
746 |
case XVID_CSP_RGB32: |
case XVID_CSP_BGRA: |
747 |
safe_packed_conv( |
safe_packed_conv( |
748 |
dst, dst_stride, image->y, image->u, image->v, |
dst[0], dst_stride[0], image->y, image->u, image->v, |
749 |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
750 |
interlacing?yv12_to_bgrai :yv12_to_bgra, |
interlacing?yv12_to_bgrai :yv12_to_bgra, |
751 |
interlacing?yv12_to_bgrai_c:yv12_to_bgra_c, 4); |
interlacing?yv12_to_bgrai_c:yv12_to_bgra_c, 4); |
753 |
|
|
754 |
case XVID_CSP_ABGR: |
case XVID_CSP_ABGR: |
755 |
safe_packed_conv( |
safe_packed_conv( |
756 |
dst, dst_stride, image->y, image->u, image->v, |
dst[0], dst_stride[0], image->y, image->u, image->v, |
757 |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
758 |
interlacing?yv12_to_abgri :yv12_to_abgr, |
interlacing?yv12_to_abgri :yv12_to_abgr, |
759 |
interlacing?yv12_to_abgri_c:yv12_to_abgr_c, 4); |
interlacing?yv12_to_abgri_c:yv12_to_abgr_c, 4); |
761 |
|
|
762 |
case XVID_CSP_RGBA: |
case XVID_CSP_RGBA: |
763 |
safe_packed_conv( |
safe_packed_conv( |
764 |
dst, dst_stride, image->y, image->u, image->v, |
dst[0], dst_stride[0], image->y, image->u, image->v, |
765 |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
766 |
interlacing?yv12_to_rgbai :yv12_to_rgba, |
interlacing?yv12_to_rgbai :yv12_to_rgba, |
767 |
interlacing?yv12_to_rgbai_c:yv12_to_rgba_c, 4); |
interlacing?yv12_to_rgbai_c:yv12_to_rgba_c, 4); |
769 |
|
|
770 |
case XVID_CSP_YUY2: |
case XVID_CSP_YUY2: |
771 |
safe_packed_conv( |
safe_packed_conv( |
772 |
dst, dst_stride, image->y, image->u, image->v, |
dst[0], dst_stride[0], image->y, image->u, image->v, |
773 |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
774 |
interlacing?yv12_to_yuyvi :yv12_to_yuyv, |
interlacing?yv12_to_yuyvi :yv12_to_yuyv, |
775 |
interlacing?yv12_to_yuyvi_c:yv12_to_yuyv_c, 2); |
interlacing?yv12_to_yuyvi_c:yv12_to_yuyv_c, 2); |
776 |
return 0; |
return 0; |
777 |
|
|
778 |
case XVID_CSP_YVYU: // u,v swapped |
case XVID_CSP_YVYU: /* u,v swapped */ |
779 |
safe_packed_conv( |
safe_packed_conv( |
780 |
dst, dst_stride, image->y, image->v, image->u, |
dst[0], dst_stride[0], image->y, image->v, image->u, |
781 |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
782 |
interlacing?yv12_to_yuyvi :yv12_to_yuyv, |
interlacing?yv12_to_yuyvi :yv12_to_yuyv, |
783 |
interlacing?yv12_to_yuyvi_c:yv12_to_yuyv_c, 2); |
interlacing?yv12_to_yuyvi_c:yv12_to_yuyv_c, 2); |
785 |
|
|
786 |
case XVID_CSP_UYVY: |
case XVID_CSP_UYVY: |
787 |
safe_packed_conv( |
safe_packed_conv( |
788 |
dst, dst_stride, image->y, image->u, image->v, |
dst[0], dst_stride[0], image->y, image->u, image->v, |
789 |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
790 |
interlacing?yv12_to_uyvyi :yv12_to_uyvy, |
interlacing?yv12_to_uyvyi :yv12_to_uyvy, |
791 |
interlacing?yv12_to_uyvyi_c:yv12_to_uyvy_c, 2); |
interlacing?yv12_to_uyvyi_c:yv12_to_uyvy_c, 2); |
792 |
return 0; |
return 0; |
793 |
|
|
794 |
case XVID_CSP_I420: |
case XVID_CSP_I420: /* YCrCb == internal colorspace for MPEG */ |
795 |
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, |
796 |
dst_stride, dst_stride/2, |
dst_stride[0], dst_stride[0]/2, |
797 |
image->y, image->u, image->v, edged_width, edged_width2, |
image->y, image->u, image->v, edged_width, edged_width2, |
798 |
width, height, (csp & XVID_CSP_VFLIP)); |
width, height, (csp & XVID_CSP_VFLIP)); |
799 |
return 0; |
return 0; |
800 |
|
|
801 |
case XVID_CSP_YV12: // u,v swapped |
case XVID_CSP_YV12: /* YCbCr == U and V plane swapped */ |
802 |
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, |
803 |
dst_stride, dst_stride/2, |
dst_stride[0], dst_stride[0]/2, |
804 |
image->y, image->v, image->u, edged_width, edged_width2, |
image->y, image->v, image->u, edged_width, edged_width2, |
805 |
width, height, (csp & XVID_CSP_VFLIP)); |
width, height, (csp & XVID_CSP_VFLIP)); |
806 |
return 0; |
return 0; |
807 |
|
|
808 |
case XVID_CSP_USER: |
case XVID_CSP_USER : /* YCrCb with arbitrary pointers and different strides for Y and UV */ |
809 |
{ |
yv12_to_yv12(dst[0], dst[1], dst[2], |
810 |
DEC_PICTURE * pic = (DEC_PICTURE*)dst; |
dst_stride[0], dst_stride[1], /* v: dst_stride[2] not yet supported */ |
811 |
pic->y = image->y; |
image->y, image->u, image->v, edged_width, edged_width2, |
812 |
pic->u = image->u; |
width, height, (csp & XVID_CSP_VFLIP)); |
813 |
pic->v = image->v; |
return 0; |
814 |
pic->stride_y = edged_width; |
|
815 |
pic->stride_uv = edged_width / 2; |
case XVID_CSP_INTERNAL : |
816 |
} |
dst[0] = image->y; |
817 |
|
dst[1] = image->u; |
818 |
|
dst[2] = image->v; |
819 |
|
dst_stride[0] = edged_width; |
820 |
|
dst_stride[1] = edged_width/2; |
821 |
|
dst_stride[2] = edged_width/2; |
822 |
return 0; |
return 0; |
823 |
|
|
824 |
case XVID_CSP_NULL: |
case XVID_CSP_NULL: |
825 |
case XVID_CSP_EXTERN: |
case XVID_CSP_SLICE: |
826 |
return 0; |
return 0; |
827 |
|
|
828 |
} |
} |
866 |
float sse_to_PSNR(long sse, int pixels) |
float sse_to_PSNR(long sse, int pixels) |
867 |
{ |
{ |
868 |
if (sse==0) |
if (sse==0) |
869 |
return 99.99; |
return 99.99F; |
870 |
|
|
871 |
return 48.131 - 10*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 */ |
872 |
|
|
873 |
} |
} |
874 |
|
|
892 |
return sse; |
return sse; |
893 |
} |
} |
894 |
|
|
895 |
/* |
#if 0 |
896 |
|
|
897 |
#include <stdio.h> |
#include <stdio.h> |
898 |
#include <string.h> |
#include <string.h> |
916 |
} |
} |
917 |
|
|
918 |
|
|
919 |
// dump image+edges to yuv pgm files |
/* dump image+edges to yuv pgm files */ |
920 |
|
|
921 |
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) |
922 |
{ |
{ |
939 |
|
|
940 |
return 0; |
return 0; |
941 |
} |
} |
942 |
*/ |
#endif |
943 |
|
|
944 |
|
|
945 |
|
|
1004 |
|
|
1005 |
for (y = 0; y < height; y++) |
for (y = 0; y < height; y++) |
1006 |
for (x = 0; x < width; x++) |
for (x = 0; x < width; x++) |
1007 |
sum += ABS(img1->y[x + y * stride] - img2->y[x + y * stride]); |
sum += abs(img1->y[x + y * stride] - img2->y[x + y * stride]); |
1008 |
|
|
1009 |
for (y = 0; y < height2; y++) |
for (y = 0; y < height2; y++) |
1010 |
for (x = 0; x < width2; x++) |
for (x = 0; x < width2; x++) |
1011 |
sum += ABS(img1->u[x + y * stride2] - img2->u[x + y * stride2]); |
sum += abs(img1->u[x + y * stride2] - img2->u[x + y * stride2]); |
1012 |
|
|
1013 |
for (y = 0; y < height2; y++) |
for (y = 0; y < height2; y++) |
1014 |
for (x = 0; x < width2; x++) |
for (x = 0; x < width2; x++) |
1015 |
sum += ABS(img1->v[x + y * stride2] - img2->v[x + y * stride2]); |
sum += abs(img1->v[x + y * stride2] - img2->v[x + y * stride2]); |
1016 |
|
|
1017 |
return (float) sum / (width * height * 3 / 2); |
return (float) sum / (width * height * 3 / 2); |
1018 |
} |
} |
1019 |
|
|
1020 |
void |
void |
1021 |
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) { |
1022 |
uint8_t *dY,*dU,*dV,*sY,*sU,*sV; |
uint8_t *dY,*dU,*dV,*sY,*sU,*sV; |
1023 |
int std2 = std >> 1; |
int std2 = std >> 1; |
1024 |
int w = mbl << 4, w2,i; |
int w = mbl << 4, w2,i; |
1027 |
w = width; |
w = width; |
1028 |
w2 = w >> 1; |
w2 = w >> 1; |
1029 |
|
|
1030 |
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); |
1031 |
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); |
1032 |
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); |
1033 |
sY = cur->y + (mby << 4) * std + (mbx << 4); |
sY = cur->y + (mby << 4) * std + (mbx << 4); |
1034 |
sU = cur->u + (mby << 3) * std2 + (mbx << 3); |
sU = cur->u + (mby << 3) * std2 + (mbx << 3); |
1035 |
sV = cur->v + (mby << 3) * std2 + (mbx << 3); |
sV = cur->v + (mby << 3) * std2 + (mbx << 3); |
1036 |
|
|
1037 |
for(i = 0 ; i < 16 ; i++) { |
for(i = 0 ; i < 16 ; i++) { |
1038 |
memcpy(dY,sY,w); |
memcpy(dY,sY,w); |
1039 |
dY += out_frm->stride_y; |
dY += out_frm->stride[0]; |
1040 |
sY += std; |
sY += std; |
1041 |
} |
} |
1042 |
for(i = 0 ; i < 8 ; i++) { |
for(i = 0 ; i < 8 ; i++) { |
1043 |
memcpy(dU,sU,w2); |
memcpy(dU,sU,w2); |
1044 |
dU += out_frm->stride_u; |
dU += out_frm->stride[1]; |
1045 |
sU += std2; |
sU += std2; |
1046 |
} |
} |
1047 |
for(i = 0 ; i < 8 ; i++) { |
for(i = 0 ; i < 8 ; i++) { |
1048 |
memcpy(dV,sV,w2); |
memcpy(dV,sV,w2); |
1049 |
dV += out_frm->stride_v; |
dV += out_frm->stride[2]; |
1050 |
sV += std2; |
sV += std2; |
1051 |
} |
} |
1052 |
} |
} |
1093 |
int i,j; |
int i,j; |
1094 |
|
|
1095 |
/* luma: j,i in block units */ |
/* luma: j,i in block units */ |
1096 |
if ((flags & XVID_DEC_DEBLOCKY)) |
|
|
{ |
|
1097 |
for (j = 1; j < mb_height*2; j++) /* horizontal deblocking */ |
for (j = 1; j < mb_height*2; j++) /* horizontal deblocking */ |
1098 |
for (i = 0; i < mb_width*2; i++) |
for (i = 0; i < mb_width*2; i++) |
1099 |
{ |
{ |
1116 |
edged_width, nblocks); |
edged_width, nblocks); |
1117 |
} |
} |
1118 |
} |
} |
1119 |
} |
|
1120 |
|
|
1121 |
|
|
1122 |
/* chroma */ |
/* chroma */ |
1123 |
if ((flags & XVID_DEC_DEBLOCKUV)) |
|
|
{ |
|
1124 |
for (j = 1; j < mb_height; j++) /* horizontal deblocking */ |
for (j = 1; j < mb_height; j++) /* horizontal deblocking */ |
1125 |
for (i = 0; i < mb_width; i++) |
for (i = 0; i < mb_width; i++) |
1126 |
{ |
{ |
1148 |
edged_width2, nblocks); |
edged_width2, nblocks); |
1149 |
} |
} |
1150 |
} |
} |
1151 |
} |
|
1152 |
|
|
1153 |
} |
} |
1154 |
|
|