1 |
|
/***************************************************************************** |
2 |
|
* |
3 |
|
* XVID MPEG-4 VIDEO CODEC |
4 |
|
* - Motion Compensation module - |
5 |
|
* |
6 |
|
* Copyright(C) 2002 Peter Ross <pross@xvid.org> |
7 |
|
* |
8 |
|
* This file is part of XviD, a free MPEG-4 video encoder/decoder |
9 |
|
* |
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 |
13 |
|
* (at your option) any later version. |
14 |
|
* |
15 |
|
* This program is distributed in the hope that it will be useful, |
16 |
|
* but WITHOUT ANY WARRANTY; without even the implied warranty of |
17 |
|
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
18 |
|
* GNU General Public License for more details. |
19 |
|
* |
20 |
|
* You should have received a copy of the GNU General Public License |
21 |
|
* along with this program; if not, write to the Free Software |
22 |
|
* 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: motion_comp.c,v 1.15 2002-11-17 00:32:06 edgomez Exp $ |
54 |
|
* |
55 |
|
*************************************************************************/ |
56 |
|
|
57 |
#include "../encoder.h" |
#include "../encoder.h" |
58 |
#include "../utils/mbfunctions.h" |
#include "../utils/mbfunctions.h" |
59 |
#include "../image/interpolate8x8.h" |
#include "../image/interpolate8x8.h" |
60 |
#include "../utils/timer.h" |
#include "../utils/timer.h" |
61 |
|
#include "motion.h" |
62 |
|
|
63 |
#define ABS(X) (((X)>0)?(X):-(X)) |
#define ABS(X) (((X)>0)?(X):-(X)) |
64 |
#define SIGN(X) (((X)>0)?1:-1) |
#define SIGN(X) (((X)>0)?1:-1) |
65 |
|
|
66 |
static __inline void compensate8x8_halfpel( |
static __inline void |
67 |
int16_t * const dct_codes, |
compensate8x8_halfpel(int16_t * const dct_codes, |
68 |
uint8_t * const cur, |
uint8_t * const cur, |
69 |
const uint8_t * const ref, |
const uint8_t * const ref, |
70 |
const uint8_t * const refh, |
const uint8_t * const refh, |
71 |
const uint8_t * const refv, |
const uint8_t * const refv, |
72 |
const uint8_t * const refhv, |
const uint8_t * const refhv, |
73 |
const uint32_t x, const uint32_t y, |
const uint32_t x, |
74 |
const int32_t dx, const int dy, |
const uint32_t y, |
75 |
|
const int32_t dx, |
76 |
|
const int dy, |
77 |
const uint32_t stride) |
const uint32_t stride) |
78 |
{ |
{ |
79 |
int32_t ddx,ddy; |
int32_t ddx,ddy; |
84 |
ddx = dx/2; |
ddx = dx/2; |
85 |
ddy = dy/2; |
ddy = dy/2; |
86 |
transfer_8to16sub(dct_codes, cur + y*stride + x, |
transfer_8to16sub(dct_codes, cur + y*stride + x, |
87 |
ref + (y+ddy)*stride + x+ddx, stride); |
ref + (int) ((y + ddy) * stride + x + ddx), stride); |
88 |
break; |
break; |
89 |
|
|
90 |
case 1 : |
case 1 : |
91 |
ddx = dx/2; |
ddx = dx/2; |
92 |
ddy = (dy-1)/2; |
ddy = (dy-1)/2; |
93 |
transfer_8to16sub(dct_codes, cur + y*stride + x, |
transfer_8to16sub(dct_codes, cur + y*stride + x, |
94 |
refv + (y+ddy)*stride + x+ddx, stride); |
refv + (int) ((y + ddy) * stride + x + ddx), stride); |
95 |
break; |
break; |
96 |
|
|
97 |
case 2 : |
case 2 : |
98 |
ddx = (dx-1)/2; |
ddx = (dx-1)/2; |
99 |
ddy = dy/2; |
ddy = dy/2; |
100 |
transfer_8to16sub(dct_codes, cur + y*stride + x, |
transfer_8to16sub(dct_codes, cur + y*stride + x, |
101 |
refh + (y+ddy)*stride + x+ddx, stride); |
refh + (int) ((y + ddy) * stride + x + ddx), stride); |
102 |
break; |
break; |
103 |
|
|
104 |
default : // case 3: |
default : // case 3: |
105 |
ddx = (dx-1)/2; |
ddx = (dx-1)/2; |
106 |
ddy = (dy-1)/2; |
ddy = (dy-1)/2; |
107 |
transfer_8to16sub(dct_codes, cur + y*stride + x, |
transfer_8to16sub(dct_codes, cur + y*stride + x, |
108 |
refhv + (y+ddy)*stride + x+ddx, stride); |
refhv + (int) ((y + ddy) * stride + x + ddx), stride); |
109 |
break; |
break; |
110 |
} |
} |
111 |
} |
} |
112 |
|
|
113 |
|
|
114 |
|
|
115 |
void MBMotionCompensation( |
void |
116 |
MACROBLOCK * const mb, |
MBMotionCompensation(MACROBLOCK * const mb, |
117 |
const uint32_t i, |
const uint32_t i, |
118 |
const uint32_t j, |
const uint32_t j, |
119 |
const IMAGE * const ref, |
const IMAGE * const ref, |
121 |
const IMAGE * const refv, |
const IMAGE * const refv, |
122 |
const IMAGE * const refhv, |
const IMAGE * const refhv, |
123 |
IMAGE * const cur, |
IMAGE * const cur, |
124 |
int16_t dct_codes[][64], |
int16_t * dct_codes, |
125 |
const uint32_t width, |
const uint32_t width, |
126 |
const uint32_t height, |
const uint32_t height, |
127 |
const uint32_t edged_width, |
const uint32_t edged_width, |
131 |
{ 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 2, 2 }; |
{ 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 2, 2 }; |
132 |
|
|
133 |
|
|
134 |
if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) |
if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { |
|
{ |
|
135 |
int32_t dx = mb->mvs[0].x; |
int32_t dx = mb->mvs[0].x; |
136 |
int32_t dy = mb->mvs[0].y; |
int32_t dy = mb->mvs[0].y; |
137 |
|
|
138 |
compensate8x8_halfpel(dct_codes[0], cur->y, ref->y, refh->y, refv->y, refhv->y, |
compensate8x8_halfpel(&dct_codes[0 * 64], cur->y, ref->y, refh->y, |
139 |
16*i, 16*j, dx, dy, edged_width); |
refv->y, refhv->y, 16 * i, 16 * j, dx, dy, |
140 |
compensate8x8_halfpel(dct_codes[1], cur->y, ref->y, refh->y, refv->y, refhv->y, |
edged_width); |
141 |
16*i + 8, 16*j, dx, dy, edged_width); |
compensate8x8_halfpel(&dct_codes[1 * 64], cur->y, ref->y, refh->y, |
142 |
compensate8x8_halfpel(dct_codes[2], cur->y, ref->y, refh->y, refv->y, refhv->y, |
refv->y, refhv->y, 16 * i + 8, 16 * j, dx, dy, |
143 |
16*i, 16*j + 8, dx, dy, edged_width); |
edged_width); |
144 |
compensate8x8_halfpel(dct_codes[3], cur->y, ref->y, refh->y, refv->y, refhv->y, |
compensate8x8_halfpel(&dct_codes[2 * 64], cur->y, ref->y, refh->y, |
145 |
16*i + 8, 16*j + 8, dx, dy, edged_width); |
refv->y, refhv->y, 16 * i, 16 * j + 8, dx, dy, |
146 |
|
edged_width); |
147 |
|
compensate8x8_halfpel(&dct_codes[3 * 64], cur->y, ref->y, refh->y, |
148 |
|
refv->y, refhv->y, 16 * i + 8, 16 * j + 8, dx, |
149 |
|
dy, edged_width); |
150 |
|
|
151 |
dx = (dx & 3) ? (dx >> 1) | 1 : dx / 2; |
dx = (dx & 3) ? (dx >> 1) | 1 : dx / 2; |
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 */ |
|
compensate8x8_halfpel(dct_codes[4], cur->u, ref->u, refh->u, refv->u, refhv->u, |
|
|
8*i, 8*j, dx, dy, edged_width/2); |
|
|
compensate8x8_halfpel(dct_codes[5], cur->v, ref->v, refh->v, refv->v, refhv->v, |
|
|
8*i, 8*j, dx, dy, edged_width/2); */ |
|
155 |
|
|
156 |
/* uv-block-based compensation */ |
interpolate8x8_switch(refv->u, ref->u, 8 * i, 8 * j, dx, dy, |
157 |
interpolate8x8_switch(refv->u, ref->u, 8*i, 8*j, dx, dy, edged_width/2, rounding); |
edged_width / 2, rounding); |
158 |
transfer_8to16sub(dct_codes[4], |
transfer_8to16sub(&dct_codes[4 * 64], |
159 |
cur->u + 8*j*edged_width/2 + 8*i, |
cur->u + 8*j*edged_width/2 + 8*i, |
160 |
refv->u + 8*j*edged_width/2 + 8*i, edged_width/2); |
refv->u + 8 * j * edged_width / 2 + 8 * i, |
161 |
|
edged_width / 2); |
162 |
|
|
163 |
interpolate8x8_switch(refv->v, ref->v, 8*i, 8*j, dx, dy, edged_width/2, rounding); |
interpolate8x8_switch(refv->v, ref->v, 8 * i, 8 * j, dx, dy, |
164 |
transfer_8to16sub(dct_codes[5], |
edged_width / 2, rounding); |
165 |
|
transfer_8to16sub(&dct_codes[5 * 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, edged_width/2); |
refv->v + 8 * j * edged_width / 2 + 8 * i, |
168 |
|
edged_width / 2); |
169 |
|
|
170 |
} |
} else // mode == MODE_INTER4V |
|
else // mode == MODE_INTER4V |
|
171 |
{ |
{ |
172 |
int32_t sum, dx, dy; |
int32_t sum, dx, dy; |
173 |
|
|
174 |
compensate8x8_halfpel(dct_codes[0], cur->y, ref->y, refh->y, refv->y, refhv->y, |
compensate8x8_halfpel(&dct_codes[0 * 64], cur->y, ref->y, refh->y, |
175 |
16*i, 16*j, mb->mvs[0].x, mb->mvs[0].y, edged_width); |
refv->y, refhv->y, 16 * i, 16 * j, mb->mvs[0].x, |
176 |
compensate8x8_halfpel(dct_codes[1], cur->y, ref->y, refh->y, refv->y, refhv->y, |
mb->mvs[0].y, edged_width); |
177 |
16*i + 8, 16*j, mb->mvs[1].x, mb->mvs[1].y, edged_width); |
compensate8x8_halfpel(&dct_codes[1 * 64], cur->y, ref->y, refh->y, |
178 |
compensate8x8_halfpel(dct_codes[2], cur->y, ref->y, refh->y, refv->y, refhv->y, |
refv->y, refhv->y, 16 * i + 8, 16 * j, |
179 |
16*i, 16*j + 8, mb->mvs[2].x, mb->mvs[2].y, edged_width); |
mb->mvs[1].x, mb->mvs[1].y, edged_width); |
180 |
compensate8x8_halfpel(dct_codes[3], cur->y, ref->y, refh->y, refv->y, refhv->y, |
compensate8x8_halfpel(&dct_codes[2 * 64], cur->y, ref->y, refh->y, |
181 |
16*i + 8, 16*j + 8, mb->mvs[3].x, mb->mvs[3].y, edged_width); |
refv->y, refhv->y, 16 * i, 16 * j + 8, |
182 |
|
mb->mvs[2].x, mb->mvs[2].y, edged_width); |
183 |
|
compensate8x8_halfpel(&dct_codes[3 * 64], cur->y, ref->y, refh->y, |
184 |
|
refv->y, refhv->y, 16 * i + 8, 16 * j + 8, |
185 |
|
mb->mvs[3].x, mb->mvs[3].y, edged_width); |
186 |
|
|
187 |
sum = mb->mvs[0].x + mb->mvs[1].x + mb->mvs[2].x + mb->mvs[3].x; |
sum = mb->mvs[0].x + mb->mvs[1].x + mb->mvs[2].x + mb->mvs[3].x; |
188 |
dx = (sum ? SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2) : 0); |
dx = (sum ? SIGN(sum) * |
189 |
|
(roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2) : 0); |
190 |
|
|
191 |
sum = mb->mvs[0].y + mb->mvs[1].y + mb->mvs[2].y + mb->mvs[3].y; |
sum = mb->mvs[0].y + mb->mvs[1].y + mb->mvs[2].y + mb->mvs[3].y; |
192 |
dy = (sum ? SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2) : 0); |
dy = (sum ? SIGN(sum) * |
193 |
|
(roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2) : 0); |
|
/* uv-image-based compensation |
|
|
compensate8x8_halfpel(dct_codes[4], cur->u, ref->u, refh->u, refv->u, refhv->u, |
|
|
8*i, 8*j, dx, dy, edged_width/2); |
|
|
compensate8x8_halfpel(dct_codes[5], cur->v, ref->v, refh->v, refv->v, refhv->v, |
|
|
8*i, 8*j, dx, dy, edged_width/2); */ |
|
194 |
|
|
195 |
/* uv-block-based compensation */ |
/* uv-block-based compensation */ |
196 |
interpolate8x8_switch(refv->u, ref->u, 8*i, 8*j, dx, dy, edged_width/2, rounding); |
interpolate8x8_switch(refv->u, ref->u, 8 * i, 8 * j, dx, dy, |
197 |
transfer_8to16sub(dct_codes[4], |
edged_width / 2, rounding); |
198 |
|
transfer_8to16sub(&dct_codes[4 * 64], |
199 |
cur->u + 8*j*edged_width/2 + 8*i, |
cur->u + 8*j*edged_width/2 + 8*i, |
200 |
refv->u + 8*j*edged_width/2 + 8*i, edged_width/2); |
refv->u + 8 * j * edged_width / 2 + 8 * i, |
201 |
|
edged_width / 2); |
202 |
|
|
203 |
interpolate8x8_switch(refv->v, ref->v, 8*i, 8*j, dx, dy, edged_width/2, rounding); |
interpolate8x8_switch(refv->v, ref->v, 8 * i, 8 * j, dx, dy, |
204 |
transfer_8to16sub(dct_codes[5], |
edged_width / 2, rounding); |
205 |
|
transfer_8to16sub(&dct_codes[5 * 64], |
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, edged_width/2); |
refv->v + 8 * j * edged_width / 2 + 8 * i, |
208 |
|
edged_width / 2); |
209 |
} |
} |
210 |
} |
} |