3 |
* XVID MPEG-4 VIDEO CODEC |
* XVID MPEG-4 VIDEO CODEC |
4 |
* - Motion Compensation module - |
* - Motion Compensation module - |
5 |
* |
* |
6 |
* Copyright(C) 2002 Michael Militzer <michael@xvid.org> |
* Copyright(C) 2002 Peter Ross <pross@xvid.org> |
|
* Copyright(C) 2002 Edouard Gomez <ed.gomez@wanadoo.fr> |
|
|
* Copyright(C) 2002 Christoph Lampert <gruel@web.de> |
|
|
* |
|
|
* This program is an implementation of a part of one or more MPEG-4 |
|
|
* 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 file is part of XviD, a free MPEG-4 video encoder/decoder |
9 |
* it under the terms of the GNU General Public License as published by |
* |
10 |
|
* XviD is free software; you can redistribute it and/or modify it |
11 |
|
* under the terms of the GNU General Public License as published by |
12 |
* the Free Software Foundation; either version 2 of the License, or |
* the Free Software Foundation; either version 2 of the License, or |
13 |
* (at your option) any later version. |
* (at your option) any later version. |
14 |
* |
* |
21 |
* along with this program; if not, write to the Free Software |
* along with this program; if not, write to the Free Software |
22 |
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA |
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA |
23 |
* |
* |
24 |
|
* Under section 8 of the GNU General Public License, the copyright |
25 |
|
* holders of XVID explicitly forbid distribution in the following |
26 |
|
* countries: |
27 |
|
* |
28 |
|
* - Japan |
29 |
|
* - United States of America |
30 |
|
* |
31 |
|
* Linking XviD statically or dynamically with other modules is making a |
32 |
|
* combined work based on XviD. Thus, the terms and conditions of the |
33 |
|
* GNU General Public License cover the whole combination. |
34 |
|
* |
35 |
|
* As a special exception, the copyright holders of XviD give you |
36 |
|
* permission to link XviD with independent modules that communicate with |
37 |
|
* XviD solely through the VFW1.1 and DShow interfaces, regardless of the |
38 |
|
* license terms of these independent modules, and to copy and distribute |
39 |
|
* the resulting combined work under terms of your choice, provided that |
40 |
|
* every copy of the combined work is accompanied by a complete copy of |
41 |
|
* the source code of XviD (the version of XviD used to produce the |
42 |
|
* combined work), being distributed under the terms of the GNU General |
43 |
|
* Public License plus this exception. An independent module is a module |
44 |
|
* which is not derived from or based on XviD. |
45 |
|
* |
46 |
|
* Note that people who make modified versions of XviD are not obligated |
47 |
|
* to grant this special exception for their modified versions; it is |
48 |
|
* their choice whether to do so. The GNU General Public License gives |
49 |
|
* permission to release a modified version without this exception; this |
50 |
|
* exception also makes it possible to release a modified version which |
51 |
|
* carries forward this exception. |
52 |
|
* |
53 |
|
* $Id$ |
54 |
|
* |
55 |
*************************************************************************/ |
*************************************************************************/ |
56 |
|
|
57 |
#include "../encoder.h" |
#include "../encoder.h" |
152 |
dy = (dy & 3) ? (dy >> 1) | 1 : dy / 2; |
dy = (dy & 3) ? (dy >> 1) | 1 : dy / 2; |
153 |
|
|
154 |
/* uv-image-based compensation */ |
/* uv-image-based compensation */ |
155 |
/* Always do block-based compensation, until check for HALFPEL is possible |
|
|
#ifdef BFRAMES |
|
|
compensate8x8_halfpel(&dct_codes[4 * 64], cur->u, ref->u, refh->u, |
|
|
refv->u, refhv->u, 8 * i, 8 * j, dx, dy, |
|
|
edged_width / 2); |
|
|
compensate8x8_halfpel(&dct_codes[5 * 64], cur->v, ref->v, refh->v, |
|
|
refv->v, refhv->v, 8 * i, 8 * j, dx, dy, |
|
|
edged_width / 2); |
|
|
#else |
|
|
*/ |
|
|
/* uv-block-based compensation */ |
|
156 |
interpolate8x8_switch(refv->u, ref->u, 8 * i, 8 * j, dx, dy, |
interpolate8x8_switch(refv->u, ref->u, 8 * i, 8 * j, dx, dy, |
157 |
edged_width / 2, rounding); |
edged_width / 2, rounding); |
158 |
transfer_8to16sub(&dct_codes[4 * 64], |
transfer_8to16sub(&dct_codes[4 * 64], |
166 |
cur->v + 8 * j * edged_width / 2 + 8 * i, |
cur->v + 8 * j * edged_width / 2 + 8 * i, |
167 |
refv->v + 8 * j * edged_width / 2 + 8 * i, |
refv->v + 8 * j * edged_width / 2 + 8 * i, |
168 |
edged_width / 2); |
edged_width / 2); |
169 |
/* |
|
|
#endif |
|
|
*/ |
|
170 |
} else // mode == MODE_INTER4V |
} else // mode == MODE_INTER4V |
171 |
{ |
{ |
172 |
int32_t sum, dx, dy; |
int32_t sum, dx, dy; |
192 |
dy = (sum ? SIGN(sum) * |
dy = (sum ? SIGN(sum) * |
193 |
(roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2) : 0); |
(roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2) : 0); |
194 |
|
|
|
/* uv-image-based compensation */ |
|
|
#ifdef BFRAMES |
|
|
compensate8x8_halfpel(&dct_codes[4 * 64], cur->u, ref->u, refh->u, |
|
|
refv->u, refhv->u, 8 * i, 8 * j, dx, dy, |
|
|
edged_width / 2); |
|
|
compensate8x8_halfpel(&dct_codes[5 * 64], cur->v, ref->v, refh->v, |
|
|
refv->v, refhv->v, 8 * i, 8 * j, dx, dy, |
|
|
edged_width / 2); |
|
|
#else |
|
195 |
/* uv-block-based compensation */ |
/* uv-block-based compensation */ |
196 |
interpolate8x8_switch(refv->u, ref->u, 8 * i, 8 * j, dx, dy, |
interpolate8x8_switch(refv->u, ref->u, 8 * i, 8 * j, dx, dy, |
197 |
edged_width / 2, rounding); |
edged_width / 2, rounding); |
206 |
cur->v + 8 * j * edged_width / 2 + 8 * i, |
cur->v + 8 * j * edged_width / 2 + 8 * i, |
207 |
refv->v + 8 * j * edged_width / 2 + 8 * i, |
refv->v + 8 * j * edged_width / 2 + 8 * i, |
208 |
edged_width / 2); |
edged_width / 2); |
|
#endif |
|
|
} |
|
|
} |
|
|
|
|
|
|
|
|
void |
|
|
MBMotionCompensationBVOP(MBParam * pParam, |
|
|
MACROBLOCK * const mb, |
|
|
const uint32_t i, |
|
|
const uint32_t j, |
|
|
IMAGE * const cur, |
|
|
const IMAGE * const f_ref, |
|
|
const IMAGE * const f_refh, |
|
|
const IMAGE * const f_refv, |
|
|
const IMAGE * const f_refhv, |
|
|
const IMAGE * const b_ref, |
|
|
const IMAGE * const b_refh, |
|
|
const IMAGE * const b_refv, |
|
|
const IMAGE * const b_refhv, |
|
|
int16_t * dct_codes) |
|
|
{ |
|
|
static const uint32_t roundtab[16] = |
|
|
{ 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 2, 2 }; |
|
|
|
|
|
const int32_t edged_width = pParam->edged_width; |
|
|
int32_t dx, dy; |
|
|
int32_t b_dx, b_dy; |
|
|
int k,sum; |
|
|
int x = i; |
|
|
int y = j; |
|
|
|
|
|
|
|
|
switch (mb->mode) { |
|
|
case MODE_FORWARD: |
|
|
dx = mb->mvs[0].x; |
|
|
dy = mb->mvs[0].y; |
|
|
|
|
|
transfer_8to16sub_c(&dct_codes[0 * 64], |
|
|
cur->y + (j * 16) * edged_width + (i * 16), |
|
|
get_ref(f_ref->y, f_refh->y, f_refv->y, f_refhv->y, |
|
|
i * 16, j * 16, 1, dx, dy, edged_width), |
|
|
edged_width); |
|
|
|
|
|
transfer_8to16sub(&dct_codes[1 * 64], |
|
|
cur->y + (j * 16) * edged_width + (i * 16 + 8), |
|
|
get_ref(f_ref->y, f_refh->y, f_refv->y, f_refhv->y, |
|
|
i * 16 + 8, j * 16, 1, dx, dy, edged_width), |
|
|
edged_width); |
|
|
|
|
|
transfer_8to16sub_c(&dct_codes[2 * 64], |
|
|
cur->y + (j * 16 + 8) * edged_width + (i * 16), |
|
|
get_ref(f_ref->y, f_refh->y, f_refv->y, f_refhv->y, |
|
|
i * 16, j * 16 + 8, 1, dx, dy, |
|
|
edged_width), edged_width); |
|
|
|
|
|
transfer_8to16sub(&dct_codes[3 * 64], |
|
|
cur->y + (j * 16 + 8) * edged_width + (i * 16 + 8), |
|
|
get_ref(f_ref->y, f_refh->y, f_refv->y, f_refhv->y, |
|
|
i * 16 + 8, j * 16 + 8, 1, dx, dy, |
|
|
edged_width), edged_width); |
|
|
|
|
|
|
|
|
dx = (dx & 3) ? (dx >> 1) | 1 : dx / 2; |
|
|
dy = (dy & 3) ? (dy >> 1) | 1 : dy / 2; |
|
|
|
|
|
/* uv-image-based compensation */ |
|
|
compensate8x8_halfpel(&dct_codes[4 * 64], cur->u, f_ref->u, f_refh->u, |
|
|
f_refv->u, f_refhv->u, 8 * i, 8 * j, dx, dy, |
|
|
edged_width / 2); |
|
|
compensate8x8_halfpel(&dct_codes[5 * 64], cur->v, f_ref->v, f_refh->v, |
|
|
f_refv->v, f_refhv->v, 8 * i, 8 * j, dx, dy, |
|
|
edged_width / 2); |
|
|
|
|
|
break; |
|
|
|
|
|
case MODE_BACKWARD: |
|
|
b_dx = mb->b_mvs[0].x; |
|
|
b_dy = mb->b_mvs[0].y; |
|
|
|
|
|
transfer_8to16sub_c(&dct_codes[0 * 64], |
|
|
cur->y + (j * 16) * edged_width + (i * 16), |
|
|
get_ref(b_ref->y, b_refh->y, b_refv->y, b_refhv->y, |
|
|
i * 16, j * 16, 1, b_dx, b_dy, |
|
|
edged_width), edged_width); |
|
|
|
|
|
transfer_8to16sub(&dct_codes[1 * 64], |
|
|
cur->y + (j * 16) * edged_width + (i * 16 + 8), |
|
|
get_ref(b_ref->y, b_refh->y, b_refv->y, b_refhv->y, |
|
|
i * 16 + 8, j * 16, 1, b_dx, b_dy, |
|
|
edged_width), edged_width); |
|
|
|
|
|
transfer_8to16sub_c(&dct_codes[2 * 64], |
|
|
cur->y + (j * 16 + 8) * edged_width + (i * 16), |
|
|
get_ref(b_ref->y, b_refh->y, b_refv->y, b_refhv->y, |
|
|
i * 16, j * 16 + 8, 1, b_dx, b_dy, |
|
|
edged_width), edged_width); |
|
|
|
|
|
transfer_8to16sub(&dct_codes[3 * 64], |
|
|
cur->y + (j * 16 + 8) * edged_width + (i * 16 + 8), |
|
|
get_ref(b_ref->y, b_refh->y, b_refv->y, b_refhv->y, |
|
|
i * 16 + 8, j * 16 + 8, 1, b_dx, b_dy, |
|
|
edged_width), edged_width); |
|
|
|
|
|
b_dx = (b_dx & 3) ? (b_dx >> 1) | 1 : b_dx / 2; |
|
|
b_dy = (b_dy & 3) ? (b_dy >> 1) | 1 : b_dy / 2; |
|
|
|
|
|
/* uv-image-based compensation */ |
|
|
compensate8x8_halfpel(&dct_codes[4 * 64], cur->u, b_ref->u, b_refh->u, |
|
|
b_refv->u, b_refhv->u, 8 * i, 8 * j, b_dx, b_dy, |
|
|
edged_width / 2); |
|
|
compensate8x8_halfpel(&dct_codes[5 * 64], cur->v, b_ref->v, b_refh->v, |
|
|
b_refv->v, b_refhv->v, 8 * i, 8 * j, b_dx, b_dy, |
|
|
edged_width / 2); |
|
|
|
|
|
break; |
|
|
|
|
|
|
|
|
case MODE_INTERPOLATE: /* _could_ use DIRECT, but would be overkill (no 4MV there) */ |
|
|
|
|
|
dx = mb->mvs[0].x; |
|
|
dy = mb->mvs[0].y; |
|
|
|
|
|
b_dx = mb->b_mvs[0].x; |
|
|
b_dy = mb->b_mvs[0].y; |
|
|
|
|
|
for (k=0;k<4;k++) |
|
|
{ |
|
|
transfer_8to16sub2_c(&dct_codes[k * 64], |
|
|
cur->y + (i * 16+(k&1)*8) + (j * 16+((k>>1)*8)) * edged_width, |
|
|
get_ref(f_ref->y, f_refh->y, f_refv->y, |
|
|
f_refhv->y, 2*i + (k&1), 2*j + (k>>1), 8, dx, dy, |
|
|
edged_width), |
|
|
get_ref(b_ref->y, b_refh->y, b_refv->y, |
|
|
b_refhv->y, 2*i + (k&1), 2 * j+(k>>1), 8, b_dx, b_dy, |
|
|
edged_width), |
|
|
edged_width); |
|
|
} |
|
|
|
|
|
dx = (dx & 3) ? (dx >> 1) | 1 : dx / 2; |
|
|
dy = (dy & 3) ? (dy >> 1) | 1 : dy / 2; |
|
|
|
|
|
b_dx = (b_dx & 3) ? (b_dx >> 1) | 1 : b_dx / 2; |
|
|
b_dy = (b_dy & 3) ? (b_dy >> 1) | 1 : b_dy / 2; |
|
|
|
|
|
transfer_8to16sub2_c(&dct_codes[4 * 64], |
|
|
cur->u + (y * 8) * edged_width / 2 + (x * 8), |
|
|
get_ref(f_ref->u, f_refh->u, f_refv->u, |
|
|
f_refhv->u, i, j, 8, dx, dy, |
|
|
edged_width / 2), |
|
|
get_ref(b_ref->u, b_refh->u, b_refv->u, |
|
|
b_refhv->u, i, j, 8, b_dx, b_dy, |
|
|
edged_width / 2), |
|
|
edged_width / 2); |
|
|
|
|
|
transfer_8to16sub2_c(&dct_codes[5 * 64], |
|
|
cur->v + (y * 8) * edged_width / 2 + (x * 8), |
|
|
get_ref(f_ref->v, f_refh->v, f_refv->v, |
|
|
f_refhv->v, 8 * i, 8 * j, 1, dx, dy, |
|
|
edged_width / 2), |
|
|
get_ref(b_ref->v, b_refh->v, b_refv->v, |
|
|
b_refhv->v, 8 * i, 8 * j, 1, b_dx, b_dy, |
|
|
edged_width / 2), |
|
|
edged_width / 2); |
|
|
|
|
|
break; |
|
|
|
|
|
case MODE_DIRECT: |
|
|
|
|
|
for (k=0;k<4;k++) |
|
|
{ |
|
|
dx = mb->mvs[k].x; |
|
|
dy = mb->mvs[k].y; |
|
|
|
|
|
b_dx = mb->b_mvs[k].x; |
|
|
b_dy = mb->b_mvs[k].y; |
|
|
|
|
|
// fprintf(stderr,"Direct Vector %d -- %d:%d %d:%d\n",k,dx,dy,b_dx,b_dy); |
|
|
|
|
|
transfer_8to16sub2_c(&dct_codes[k * 64], |
|
|
cur->y + (i*16 + (k&1)*8) + (j*16 + (k>>1)*8 ) * edged_width, |
|
|
get_ref(f_ref->y, f_refh->y, f_refv->y, f_refhv->y, |
|
|
2*i + (k&1), 2*j + (k>>1), 8, dx, dy, |
|
|
edged_width), |
|
|
get_ref(b_ref->y, b_refh->y, b_refv->y, b_refhv->y, |
|
|
2*i + (k&1), 2*j + (k>>1), 8, b_dx, b_dy, |
|
|
edged_width), |
|
|
edged_width); |
|
|
} |
|
|
|
|
|
sum = mb->mvs[0].x + mb->mvs[1].x + mb->mvs[2].x + mb->mvs[3].x; |
|
|
dx = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2)); |
|
|
|
|
|
sum = mb->mvs[0].y + mb->mvs[1].y + mb->mvs[2].y + mb->mvs[3].y; |
|
|
dy = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2)); |
|
|
|
|
|
|
|
|
sum = mb->b_mvs[0].x + mb->b_mvs[1].x + mb->b_mvs[2].x + mb->b_mvs[3].x; |
|
|
b_dx = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2)); |
|
|
|
|
|
sum = mb->b_mvs[0].y + mb->b_mvs[1].y + mb->b_mvs[2].y + mb->b_mvs[3].y; |
|
|
b_dy = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2)); |
|
|
|
|
|
/* // for QPel don't forget to always do |
|
|
|
|
|
if (quarterpel) |
|
|
sum /= 2; |
|
|
*/ |
|
|
|
|
|
transfer_8to16sub2_c(&dct_codes[4 * 64], |
|
|
cur->u + (y * 8) * edged_width / 2 + (x * 8), |
|
|
get_ref(f_ref->u, f_refh->u, f_refv->u, |
|
|
f_refhv->u, i, j, 8, dx, dy, |
|
|
edged_width / 2), |
|
|
get_ref(b_ref->u, b_refh->u, b_refv->u, |
|
|
b_refhv->u, i, j, 8, b_dx, b_dy, |
|
|
edged_width / 2), |
|
|
edged_width / 2); |
|
|
|
|
|
transfer_8to16sub2_c(&dct_codes[5 * 64], |
|
|
cur->v + (y * 8) * edged_width / 2 + (x * 8), |
|
|
get_ref(f_ref->v, f_refh->v, f_refv->v, |
|
|
f_refhv->v, i, j, 8, dx, dy, |
|
|
edged_width / 2), |
|
|
get_ref(b_ref->v, b_refh->v, b_refv->v, |
|
|
b_refhv->v, i, j, 8, b_dx, b_dy, |
|
|
edged_width / 2), |
|
|
edged_width / 2); |
|
|
|
|
|
|
|
|
break; |
|
209 |
} |
} |
210 |
} |
} |