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 */ |
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 |
|
|
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); |
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) { |
599 |
interlacing?rgbai_to_yv12_c:rgba_to_yv12_c, 4); |
interlacing?rgbai_to_yv12_c:rgba_to_yv12_c, 4); |
600 |
break; |
break; |
601 |
|
|
602 |
|
case XVID_CSP_ARGB: |
603 |
|
safe_packed_conv( |
604 |
|
src[0], src_stride[0], image->y, image->u, image->v, |
605 |
|
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
606 |
|
interlacing?argbi_to_yv12 : argb_to_yv12, |
607 |
|
interlacing?argbi_to_yv12_c: argb_to_yv12_c, 4); |
608 |
|
break; |
609 |
|
|
610 |
case XVID_CSP_YUY2: |
case XVID_CSP_YUY2: |
611 |
safe_packed_conv( |
safe_packed_conv( |
612 |
src[0], src_stride[0], image->y, image->u, image->v, |
src[0], src_stride[0], image->y, image->u, image->v, |
631 |
interlacing?uyvyi_to_yv12_c:uyvy_to_yv12_c, 2); |
interlacing?uyvyi_to_yv12_c:uyvy_to_yv12_c, 2); |
632 |
break; |
break; |
633 |
|
|
634 |
case XVID_CSP_I420: |
case XVID_CSP_YV12: /* YCrCb == internal colorspace for MPEG */ |
635 |
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, |
636 |
src[0], src[0] + src_stride[0]*height, src[0] + src_stride[0]*height + (src_stride[0]/2)*height2, |
src[0], src[0] + src_stride[0]*height, src[0] + src_stride[0]*height + (src_stride[0]/2)*height2, |
637 |
src_stride[0], src_stride[0]/2, width, height, (csp & XVID_CSP_VFLIP)); |
src_stride[0], src_stride[0]/2, width, height, (csp & XVID_CSP_VFLIP)); |
638 |
break |
break; |
639 |
; |
|
640 |
case XVID_CSP_YV12: /* u/v swapped */ |
case XVID_CSP_I420: /* YCbCr == U and V plane swapped */ |
641 |
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, |
642 |
src[0], src[0] + src_stride[0]*height, src[0] + src_stride[0]*height + (src_stride[0]/2)*height2, |
src[0], src[0] + src_stride[0]*height, src[0] + src_stride[0]*height + (src_stride[0]/2)*height2, |
643 |
src_stride[0], src_stride[0]/2, width, height, (csp & XVID_CSP_VFLIP)); |
src_stride[0], src_stride[0]/2, width, height, (csp & XVID_CSP_VFLIP)); |
644 |
break; |
break; |
645 |
|
|
646 |
case XVID_CSP_USER: |
case XVID_CSP_PLANAR: /* YCbCr with arbitrary pointers and different strides for Y and UV */ |
|
/*XXX: support for different u & v strides */ |
|
647 |
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, |
648 |
src[0], src[1], src[2], src_stride[0], src_stride[1], |
src[0], src[1], src[2], src_stride[0], src_stride[1], /* v: dst_stride[2] not yet supported */ |
649 |
width, height, (csp & XVID_CSP_VFLIP)); |
width, height, (csp & XVID_CSP_VFLIP)); |
650 |
break; |
break; |
651 |
|
|
775 |
interlacing?yv12_to_rgbai_c:yv12_to_rgba_c, 4); |
interlacing?yv12_to_rgbai_c:yv12_to_rgba_c, 4); |
776 |
return 0; |
return 0; |
777 |
|
|
778 |
|
case XVID_CSP_ARGB: |
779 |
|
safe_packed_conv( |
780 |
|
dst[0], dst_stride[0], image->y, image->u, image->v, |
781 |
|
edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), |
782 |
|
interlacing?yv12_to_argbi :yv12_to_argb, |
783 |
|
interlacing?yv12_to_argbi_c:yv12_to_argb_c, 4); |
784 |
|
return 0; |
785 |
|
|
786 |
case XVID_CSP_YUY2: |
case XVID_CSP_YUY2: |
787 |
safe_packed_conv( |
safe_packed_conv( |
788 |
dst[0], dst_stride[0], image->y, image->u, image->v, |
dst[0], dst_stride[0], image->y, image->u, image->v, |
807 |
interlacing?yv12_to_uyvyi_c:yv12_to_uyvy_c, 2); |
interlacing?yv12_to_uyvyi_c:yv12_to_uyvy_c, 2); |
808 |
return 0; |
return 0; |
809 |
|
|
810 |
case XVID_CSP_I420: |
case XVID_CSP_YV12: /* YCbCr == internal colorspace for MPEG */ |
811 |
yv12_to_yv12(dst[0], dst[0] + dst_stride[0]*height, dst[0] + dst_stride[0]*height + (dst_stride[0]/2)*height2, |
yv12_to_yv12(dst[0], dst[0] + dst_stride[0]*height, dst[0] + dst_stride[0]*height + (dst_stride[0]/2)*height2, |
812 |
dst_stride[0], dst_stride[0]/2, |
dst_stride[0], dst_stride[0]/2, |
813 |
image->y, image->u, image->v, edged_width, edged_width2, |
image->y, image->u, image->v, edged_width, edged_width2, |
814 |
width, height, (csp & XVID_CSP_VFLIP)); |
width, height, (csp & XVID_CSP_VFLIP)); |
815 |
return 0; |
return 0; |
816 |
|
|
817 |
case XVID_CSP_YV12: /* u,v swapped */ |
case XVID_CSP_I420: /* YCrCb == U and V plane swapped */ |
818 |
yv12_to_yv12(dst[0], dst[0] + dst_stride[0]*height, dst[0] + dst_stride[0]*height + (dst_stride[0]/2)*height2, |
yv12_to_yv12(dst[0], dst[0] + dst_stride[0]*height, dst[0] + dst_stride[0]*height + (dst_stride[0]/2)*height2, |
819 |
dst_stride[0], dst_stride[0]/2, |
dst_stride[0], dst_stride[0]/2, |
820 |
image->y, image->v, image->u, edged_width, edged_width2, |
image->y, image->v, image->u, edged_width, edged_width2, |
821 |
width, height, (csp & XVID_CSP_VFLIP)); |
width, height, (csp & XVID_CSP_VFLIP)); |
822 |
return 0; |
return 0; |
823 |
|
|
824 |
case XVID_CSP_USER : /* u,v swapped */ |
case XVID_CSP_PLANAR: /* YCbCr with arbitrary pointers and different strides for Y and UV */ |
825 |
yv12_to_yv12(dst[0], dst[1], dst[2], |
yv12_to_yv12(dst[0], dst[1], dst[2], |
826 |
dst_stride[0], dst_stride[1], /* v: dst_stride[2] */ |
dst_stride[0], dst_stride[1], /* v: dst_stride[2] not yet supported */ |
827 |
image->y, image->v, image->u, edged_width, edged_width2, |
image->y, image->u, image->v, edged_width, edged_width2, |
828 |
width, height, (csp & XVID_CSP_VFLIP)); |
width, height, (csp & XVID_CSP_VFLIP)); |
829 |
return 0; |
return 0; |
830 |
|
|