1 |
// 01.05.2002 updated MBMotionCompensationBVOP |
/***************************************************************************** |
2 |
// 14.04.2002 bframe compensation |
* |
3 |
|
* XVID MPEG-4 VIDEO CODEC |
4 |
|
* - Motion Compensation module - |
5 |
|
* |
6 |
|
* Copyright(C) 2002 Michael Militzer <michael@xvid.org> |
7 |
|
* Copyright(C) 2002 Edouard Gomez <ed.gomez@wanadoo.fr> |
8 |
|
* Copyright(C) 2002 Christoph Lampert <gruel@web.de> |
9 |
|
* |
10 |
|
* This program is an implementation of a part of one or more MPEG-4 |
11 |
|
* Video tools as specified in ISO/IEC 14496-2 standard. Those intending |
12 |
|
* to use this software module in hardware or software products are |
13 |
|
* advised that its use may infringe existing patents or copyrights, and |
14 |
|
* any such use would be at such party's own risk. The original |
15 |
|
* developer of this software module and his/her company, and subsequent |
16 |
|
* editors and their companies, will have no liability for use of this |
17 |
|
* software or modifications or derivatives thereof. |
18 |
|
* |
19 |
|
* This program is free software; you can redistribute it and/or modify |
20 |
|
* it under the terms of the GNU General Public License as published by |
21 |
|
* the Free Software Foundation; either version 2 of the License, or |
22 |
|
* (at your option) any later version. |
23 |
|
* |
24 |
|
* This program is distributed in the hope that it will be useful, |
25 |
|
* but WITHOUT ANY WARRANTY; without even the implied warranty of |
26 |
|
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
27 |
|
* GNU General Public License for more details. |
28 |
|
* |
29 |
|
* You should have received a copy of the GNU General Public License |
30 |
|
* along with this program; if not, write to the Free Software |
31 |
|
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA |
32 |
|
* |
33 |
|
*************************************************************************/ |
34 |
|
|
35 |
#include "../encoder.h" |
#include "../encoder.h" |
36 |
#include "../utils/mbfunctions.h" |
#include "../utils/mbfunctions.h" |
41 |
#define ABS(X) (((X)>0)?(X):-(X)) |
#define ABS(X) (((X)>0)?(X):-(X)) |
42 |
#define SIGN(X) (((X)>0)?1:-1) |
#define SIGN(X) (((X)>0)?1:-1) |
43 |
|
|
44 |
static __inline void compensate8x8_halfpel( |
static __inline void |
45 |
int16_t * const dct_codes, |
compensate8x8_halfpel(int16_t * const dct_codes, |
46 |
uint8_t * const cur, |
uint8_t * const cur, |
47 |
const uint8_t * const ref, |
const uint8_t * const ref, |
48 |
const uint8_t * const refh, |
const uint8_t * const refh, |
49 |
const uint8_t * const refv, |
const uint8_t * const refv, |
50 |
const uint8_t * const refhv, |
const uint8_t * const refhv, |
51 |
const uint32_t x, const uint32_t y, |
const uint32_t x, |
52 |
const int32_t dx, const int dy, |
const uint32_t y, |
53 |
|
const int32_t dx, |
54 |
|
const int dy, |
55 |
const uint32_t stride) |
const uint32_t stride) |
56 |
{ |
{ |
57 |
int32_t ddx,ddy; |
int32_t ddx,ddy; |
62 |
ddx = dx/2; |
ddx = dx/2; |
63 |
ddy = dy/2; |
ddy = dy/2; |
64 |
transfer_8to16sub(dct_codes, cur + y*stride + x, |
transfer_8to16sub(dct_codes, cur + y*stride + x, |
65 |
ref + (y+ddy)*stride + x+ddx, stride); |
ref + (int) ((y + ddy) * stride + x + ddx), stride); |
66 |
break; |
break; |
67 |
|
|
68 |
case 1 : |
case 1 : |
69 |
ddx = dx/2; |
ddx = dx/2; |
70 |
ddy = (dy-1)/2; |
ddy = (dy-1)/2; |
71 |
transfer_8to16sub(dct_codes, cur + y*stride + x, |
transfer_8to16sub(dct_codes, cur + y*stride + x, |
72 |
refv + (y+ddy)*stride + x+ddx, stride); |
refv + (int) ((y + ddy) * stride + x + ddx), stride); |
73 |
break; |
break; |
74 |
|
|
75 |
case 2 : |
case 2 : |
76 |
ddx = (dx-1)/2; |
ddx = (dx-1)/2; |
77 |
ddy = dy/2; |
ddy = dy/2; |
78 |
transfer_8to16sub(dct_codes, cur + y*stride + x, |
transfer_8to16sub(dct_codes, cur + y*stride + x, |
79 |
refh + (y+ddy)*stride + x+ddx, stride); |
refh + (int) ((y + ddy) * stride + x + ddx), stride); |
80 |
break; |
break; |
81 |
|
|
82 |
default : // case 3: |
default : // case 3: |
83 |
ddx = (dx-1)/2; |
ddx = (dx-1)/2; |
84 |
ddy = (dy-1)/2; |
ddy = (dy-1)/2; |
85 |
transfer_8to16sub(dct_codes, cur + y*stride + x, |
transfer_8to16sub(dct_codes, cur + y*stride + x, |
86 |
refhv + (y+ddy)*stride + x+ddx, stride); |
refhv + (int) ((y + ddy) * stride + x + ddx), stride); |
87 |
break; |
break; |
88 |
} |
} |
89 |
} |
} |
90 |
|
|
91 |
|
|
92 |
|
|
93 |
void MBMotionCompensation( |
void |
94 |
MACROBLOCK * const mb, |
MBMotionCompensation(MACROBLOCK * const mb, |
95 |
const uint32_t i, |
const uint32_t i, |
96 |
const uint32_t j, |
const uint32_t j, |
97 |
const IMAGE * const ref, |
const IMAGE * const ref, |
109 |
{ 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 }; |
110 |
|
|
111 |
|
|
112 |
if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) |
if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { |
|
{ |
|
113 |
int32_t dx = mb->mvs[0].x; |
int32_t dx = mb->mvs[0].x; |
114 |
int32_t dy = mb->mvs[0].y; |
int32_t dy = mb->mvs[0].y; |
115 |
|
|
116 |
compensate8x8_halfpel(&dct_codes[0*64], cur->y, ref->y, refh->y, refv->y, refhv->y, |
compensate8x8_halfpel(&dct_codes[0 * 64], cur->y, ref->y, refh->y, |
117 |
16*i, 16*j, dx, dy, edged_width); |
refv->y, refhv->y, 16 * i, 16 * j, dx, dy, |
118 |
compensate8x8_halfpel(&dct_codes[1*64], cur->y, ref->y, refh->y, refv->y, refhv->y, |
edged_width); |
119 |
16*i + 8, 16*j, dx, dy, edged_width); |
compensate8x8_halfpel(&dct_codes[1 * 64], cur->y, ref->y, refh->y, |
120 |
compensate8x8_halfpel(&dct_codes[2*64], cur->y, ref->y, refh->y, refv->y, refhv->y, |
refv->y, refhv->y, 16 * i + 8, 16 * j, dx, dy, |
121 |
16*i, 16*j + 8, dx, dy, edged_width); |
edged_width); |
122 |
compensate8x8_halfpel(&dct_codes[3*64], cur->y, ref->y, refh->y, refv->y, refhv->y, |
compensate8x8_halfpel(&dct_codes[2 * 64], cur->y, ref->y, refh->y, |
123 |
16*i + 8, 16*j + 8, dx, dy, edged_width); |
refv->y, refhv->y, 16 * i, 16 * j + 8, dx, dy, |
124 |
|
edged_width); |
125 |
|
compensate8x8_halfpel(&dct_codes[3 * 64], cur->y, ref->y, refh->y, |
126 |
|
refv->y, refhv->y, 16 * i + 8, 16 * j + 8, dx, |
127 |
|
dy, edged_width); |
128 |
|
|
129 |
dx = (dx & 3) ? (dx >> 1) | 1 : dx / 2; |
dx = (dx & 3) ? (dx >> 1) | 1 : dx / 2; |
130 |
dy = (dy & 3) ? (dy >> 1) | 1 : dy / 2; |
dy = (dy & 3) ? (dy >> 1) | 1 : dy / 2; |
131 |
|
|
132 |
/* uv-image-based compensation */ |
/* uv-image-based compensation */ |
133 |
|
/* Always do block-based compensation, until check for HALFPEL is possible |
134 |
#ifdef BFRAMES |
#ifdef BFRAMES |
135 |
compensate8x8_halfpel(&dct_codes[4*64], cur->u, ref->u, refh->u, refv->u, refhv->u, |
compensate8x8_halfpel(&dct_codes[4 * 64], cur->u, ref->u, refh->u, |
136 |
8*i, 8*j, dx, dy, edged_width/2); |
refv->u, refhv->u, 8 * i, 8 * j, dx, dy, |
137 |
compensate8x8_halfpel(&dct_codes[5*64], cur->v, ref->v, refh->v, refv->v, refhv->v, |
edged_width / 2); |
138 |
8*i, 8*j, dx, dy, edged_width/2); |
compensate8x8_halfpel(&dct_codes[5 * 64], cur->v, ref->v, refh->v, |
139 |
|
refv->v, refhv->v, 8 * i, 8 * j, dx, dy, |
140 |
|
edged_width / 2); |
141 |
#else |
#else |
142 |
|
*/ |
143 |
/* uv-block-based compensation */ |
/* uv-block-based compensation */ |
144 |
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, |
145 |
|
edged_width / 2, rounding); |
146 |
transfer_8to16sub(&dct_codes[4*64], |
transfer_8to16sub(&dct_codes[4*64], |
147 |
cur->u + 8*j*edged_width/2 + 8*i, |
cur->u + 8*j*edged_width/2 + 8*i, |
148 |
refv->u + 8*j*edged_width/2 + 8*i, edged_width/2); |
refv->u + 8 * j * edged_width / 2 + 8 * i, |
149 |
|
edged_width / 2); |
150 |
|
|
151 |
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, |
152 |
|
edged_width / 2, rounding); |
153 |
transfer_8to16sub(&dct_codes[5*64], |
transfer_8to16sub(&dct_codes[5*64], |
154 |
cur->v + 8*j*edged_width/2 + 8*i, |
cur->v + 8*j*edged_width/2 + 8*i, |
155 |
refv->v + 8*j*edged_width/2 + 8*i, edged_width/2); |
refv->v + 8 * j * edged_width / 2 + 8 * i, |
156 |
|
edged_width / 2); |
157 |
|
/* |
158 |
#endif |
#endif |
159 |
} |
*/ |
160 |
else // mode == MODE_INTER4V |
} else // mode == MODE_INTER4V |
161 |
{ |
{ |
162 |
int32_t sum, dx, dy; |
int32_t sum, dx, dy; |
163 |
|
|
164 |
compensate8x8_halfpel(&dct_codes[0*64], cur->y, ref->y, refh->y, refv->y, refhv->y, |
compensate8x8_halfpel(&dct_codes[0 * 64], cur->y, ref->y, refh->y, |
165 |
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, |
166 |
compensate8x8_halfpel(&dct_codes[1*64], cur->y, ref->y, refh->y, refv->y, refhv->y, |
mb->mvs[0].y, edged_width); |
167 |
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, |
168 |
compensate8x8_halfpel(&dct_codes[2*64], cur->y, ref->y, refh->y, refv->y, refhv->y, |
refv->y, refhv->y, 16 * i + 8, 16 * j, |
169 |
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); |
170 |
compensate8x8_halfpel(&dct_codes[3*64], cur->y, ref->y, refh->y, refv->y, refhv->y, |
compensate8x8_halfpel(&dct_codes[2 * 64], cur->y, ref->y, refh->y, |
171 |
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, |
172 |
|
mb->mvs[2].x, mb->mvs[2].y, edged_width); |
173 |
|
compensate8x8_halfpel(&dct_codes[3 * 64], cur->y, ref->y, refh->y, |
174 |
|
refv->y, refhv->y, 16 * i + 8, 16 * j + 8, |
175 |
|
mb->mvs[3].x, mb->mvs[3].y, edged_width); |
176 |
|
|
177 |
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; |
178 |
dx = (sum ? SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2) : 0); |
dx = (sum ? SIGN(sum) * |
179 |
|
(roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2) : 0); |
180 |
|
|
181 |
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; |
182 |
dy = (sum ? SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2) : 0); |
dy = (sum ? SIGN(sum) * |
183 |
|
(roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2) : 0); |
184 |
|
|
185 |
/* uv-image-based compensation */ |
/* uv-image-based compensation */ |
186 |
#ifdef BFRAMES |
#ifdef BFRAMES |
187 |
compensate8x8_halfpel(&dct_codes[4*64], cur->u, ref->u, refh->u, refv->u, refhv->u, |
compensate8x8_halfpel(&dct_codes[4 * 64], cur->u, ref->u, refh->u, |
188 |
8*i, 8*j, dx, dy, edged_width/2); |
refv->u, refhv->u, 8 * i, 8 * j, dx, dy, |
189 |
compensate8x8_halfpel(&dct_codes[5*64], cur->v, ref->v, refh->v, refv->v, refhv->v, |
edged_width / 2); |
190 |
8*i, 8*j, dx, dy, edged_width/2); |
compensate8x8_halfpel(&dct_codes[5 * 64], cur->v, ref->v, refh->v, |
191 |
|
refv->v, refhv->v, 8 * i, 8 * j, dx, dy, |
192 |
|
edged_width / 2); |
193 |
#else |
#else |
194 |
/* uv-block-based compensation */ |
/* uv-block-based compensation */ |
195 |
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, |
196 |
|
edged_width / 2, rounding); |
197 |
transfer_8to16sub(&dct_codes[4*64], |
transfer_8to16sub(&dct_codes[4*64], |
198 |
cur->u + 8*j*edged_width/2 + 8*i, |
cur->u + 8*j*edged_width/2 + 8*i, |
199 |
refv->u + 8*j*edged_width/2 + 8*i, edged_width/2); |
refv->u + 8 * j * edged_width / 2 + 8 * i, |
200 |
|
edged_width / 2); |
201 |
|
|
202 |
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, |
203 |
|
edged_width / 2, rounding); |
204 |
transfer_8to16sub(&dct_codes[5*64], |
transfer_8to16sub(&dct_codes[5*64], |
205 |
cur->v + 8*j*edged_width/2 + 8*i, |
cur->v + 8*j*edged_width/2 + 8*i, |
206 |
refv->v + 8*j*edged_width/2 + 8*i, edged_width/2); |
refv->v + 8 * j * edged_width / 2 + 8 * i, |
207 |
|
edged_width / 2); |
208 |
#endif |
#endif |
209 |
} |
} |
210 |
} |
} |
211 |
|
|
212 |
|
|
213 |
void MBMotionCompensationBVOP( |
void |
214 |
MBParam * pParam, |
MBMotionCompensationBVOP(MBParam * pParam, |
215 |
MACROBLOCK * const mb, |
MACROBLOCK * const mb, |
216 |
const uint32_t i, |
const uint32_t i, |
217 |
const uint32_t j, |
const uint32_t j, |
232 |
const int32_t edged_width = pParam->edged_width; |
const int32_t edged_width = pParam->edged_width; |
233 |
int32_t dx, dy; |
int32_t dx, dy; |
234 |
int32_t b_dx, b_dy; |
int32_t b_dx, b_dy; |
235 |
|
int k,sum; |
236 |
int x = i; |
int x = i; |
237 |
int y = j; |
int y = j; |
238 |
|
|
239 |
|
|
240 |
|
switch (mb->mode) { |
|
switch(mb->mode) |
|
|
{ |
|
241 |
case MODE_FORWARD : |
case MODE_FORWARD : |
242 |
dx = mb->mvs[0].x; |
dx = mb->mvs[0].x; |
243 |
dy = mb->mvs[0].y; |
dy = mb->mvs[0].y; |
244 |
|
|
245 |
transfer_8to16sub_c( |
transfer_8to16sub_c(&dct_codes[0 * 64], |
|
&dct_codes[0*64], |
|
246 |
cur->y + (j*16)*edged_width + (i*16), |
cur->y + (j*16)*edged_width + (i*16), |
247 |
get_ref(f_ref->y, f_refh->y, f_refv->y, f_refhv->y, |
get_ref(f_ref->y, f_refh->y, f_refv->y, f_refhv->y, |
248 |
i*16, j*16, 1, dx, dy, edged_width), |
i*16, j*16, 1, dx, dy, edged_width), |
249 |
edged_width); |
edged_width); |
250 |
|
|
251 |
transfer_8to16sub( |
transfer_8to16sub(&dct_codes[1 * 64], |
|
&dct_codes[1*64], |
|
252 |
cur->y + (j*16)*edged_width + (i*16+8), |
cur->y + (j*16)*edged_width + (i*16+8), |
253 |
get_ref(f_ref->y, f_refh->y, f_refv->y, f_refhv->y, |
get_ref(f_ref->y, f_refh->y, f_refv->y, f_refhv->y, |
254 |
i*16+8, j*16, 1, dx, dy, edged_width), |
i*16+8, j*16, 1, dx, dy, edged_width), |
255 |
edged_width); |
edged_width); |
256 |
|
|
257 |
transfer_8to16sub_c( |
transfer_8to16sub_c(&dct_codes[2 * 64], |
|
&dct_codes[2*64], |
|
258 |
cur->y + (j*16+8)*edged_width + (i*16), |
cur->y + (j*16+8)*edged_width + (i*16), |
259 |
get_ref(f_ref->y, f_refh->y, f_refv->y, f_refhv->y, |
get_ref(f_ref->y, f_refh->y, f_refv->y, f_refhv->y, |
260 |
i*16, j*16+8, 1, dx, dy, edged_width), |
i * 16, j * 16 + 8, 1, dx, dy, |
261 |
edged_width); |
edged_width), edged_width); |
262 |
|
|
263 |
transfer_8to16sub( |
transfer_8to16sub(&dct_codes[3 * 64], |
|
&dct_codes[3*64], |
|
264 |
cur->y + (j*16+8)*edged_width + (i*16+8), |
cur->y + (j*16+8)*edged_width + (i*16+8), |
265 |
get_ref(f_ref->y, f_refh->y, f_refv->y, f_refhv->y, |
get_ref(f_ref->y, f_refh->y, f_refv->y, f_refhv->y, |
266 |
i*16+8, j*16+8, 1, dx, dy, edged_width), |
i * 16 + 8, j * 16 + 8, 1, dx, dy, |
267 |
edged_width); |
edged_width), edged_width); |
268 |
|
|
269 |
|
|
270 |
dx = (dx & 3) ? (dx >> 1) | 1 : dx / 2; |
dx = (dx & 3) ? (dx >> 1) | 1 : dx / 2; |
271 |
dy = (dy & 3) ? (dy >> 1) | 1 : dy / 2; |
dy = (dy & 3) ? (dy >> 1) | 1 : dy / 2; |
272 |
|
|
273 |
/* uv-image-based compensation */ |
/* uv-image-based compensation */ |
274 |
compensate8x8_halfpel(&dct_codes[4*64], cur->u, f_ref->u, f_refh->u, f_refv->u, f_refhv->u, |
compensate8x8_halfpel(&dct_codes[4 * 64], cur->u, f_ref->u, f_refh->u, |
275 |
8*i, 8*j, dx, dy, edged_width/2); |
f_refv->u, f_refhv->u, 8 * i, 8 * j, dx, dy, |
276 |
compensate8x8_halfpel(&dct_codes[5*64], cur->v, f_ref->v, f_refh->v, f_refv->v, f_refhv->v, |
edged_width / 2); |
277 |
8*i, 8*j, dx, dy, edged_width/2); |
compensate8x8_halfpel(&dct_codes[5 * 64], cur->v, f_ref->v, f_refh->v, |
278 |
|
f_refv->v, f_refhv->v, 8 * i, 8 * j, dx, dy, |
279 |
|
edged_width / 2); |
280 |
|
|
281 |
break; |
break; |
282 |
|
|
284 |
b_dx = mb->b_mvs[0].x; |
b_dx = mb->b_mvs[0].x; |
285 |
b_dy = mb->b_mvs[0].y; |
b_dy = mb->b_mvs[0].y; |
286 |
|
|
287 |
transfer_8to16sub_c( |
transfer_8to16sub_c(&dct_codes[0 * 64], |
|
&dct_codes[0*64], |
|
288 |
cur->y + (j*16)*edged_width + (i*16), |
cur->y + (j*16)*edged_width + (i*16), |
289 |
get_ref(b_ref->y, b_refh->y, b_refv->y, b_refhv->y, |
get_ref(b_ref->y, b_refh->y, b_refv->y, b_refhv->y, |
290 |
i*16, j*16, 1, b_dx, b_dy, edged_width), |
i * 16, j * 16, 1, b_dx, b_dy, |
291 |
edged_width); |
edged_width), edged_width); |
292 |
|
|
293 |
transfer_8to16sub( |
transfer_8to16sub(&dct_codes[1 * 64], |
|
&dct_codes[1*64], |
|
294 |
cur->y + (j*16)*edged_width + (i*16+8), |
cur->y + (j*16)*edged_width + (i*16+8), |
295 |
get_ref(b_ref->y, b_refh->y, b_refv->y, b_refhv->y, |
get_ref(b_ref->y, b_refh->y, b_refv->y, b_refhv->y, |
296 |
i*16+8, j*16, 1, b_dx, b_dy, edged_width), |
i * 16 + 8, j * 16, 1, b_dx, b_dy, |
297 |
edged_width); |
edged_width), edged_width); |
298 |
|
|
299 |
transfer_8to16sub_c( |
transfer_8to16sub_c(&dct_codes[2 * 64], |
|
&dct_codes[2*64], |
|
300 |
cur->y + (j*16+8)*edged_width + (i*16), |
cur->y + (j*16+8)*edged_width + (i*16), |
301 |
get_ref(b_ref->y, b_refh->y, b_refv->y, b_refhv->y, |
get_ref(b_ref->y, b_refh->y, b_refv->y, b_refhv->y, |
302 |
i*16, j*16+8, 1, b_dx, b_dy, edged_width), |
i * 16, j * 16 + 8, 1, b_dx, b_dy, |
303 |
edged_width); |
edged_width), edged_width); |
304 |
|
|
305 |
transfer_8to16sub( |
transfer_8to16sub(&dct_codes[3 * 64], |
|
&dct_codes[3*64], |
|
306 |
cur->y + (j*16+8)*edged_width + (i*16+8), |
cur->y + (j*16+8)*edged_width + (i*16+8), |
307 |
get_ref(b_ref->y, b_refh->y, b_refv->y, b_refhv->y, |
get_ref(b_ref->y, b_refh->y, b_refv->y, b_refhv->y, |
308 |
i*16+8, j*16+8, 1, b_dx, b_dy, edged_width), |
i * 16 + 8, j * 16 + 8, 1, b_dx, b_dy, |
309 |
edged_width); |
edged_width), edged_width); |
310 |
|
|
311 |
b_dx = (b_dx & 3) ? (b_dx >> 1) | 1 : b_dx / 2; |
b_dx = (b_dx & 3) ? (b_dx >> 1) | 1 : b_dx / 2; |
312 |
b_dy = (b_dy & 3) ? (b_dy >> 1) | 1 : b_dy / 2; |
b_dy = (b_dy & 3) ? (b_dy >> 1) | 1 : b_dy / 2; |
313 |
|
|
314 |
/* uv-image-based compensation */ |
/* uv-image-based compensation */ |
315 |
compensate8x8_halfpel(&dct_codes[4*64], cur->u, |
compensate8x8_halfpel(&dct_codes[4 * 64], cur->u, b_ref->u, b_refh->u, |
316 |
b_ref->u, b_refh->u, b_refv->u, b_refhv->u, |
b_refv->u, b_refhv->u, 8 * i, 8 * j, b_dx, b_dy, |
317 |
8*i, 8*j, b_dx, b_dy, edged_width/2); |
edged_width / 2); |
318 |
compensate8x8_halfpel(&dct_codes[5*64], cur->v, |
compensate8x8_halfpel(&dct_codes[5 * 64], cur->v, b_ref->v, b_refh->v, |
319 |
b_ref->v, b_refh->v, b_refv->v, b_refhv->v, |
b_refv->v, b_refhv->v, 8 * i, 8 * j, b_dx, b_dy, |
320 |
8*i, 8*j, b_dx, b_dy, edged_width/2); |
edged_width / 2); |
321 |
|
|
322 |
break; |
break; |
323 |
|
|
324 |
|
|
325 |
case MODE_INTERPOLATE : |
case MODE_INTERPOLATE: /* _could_ use DIRECT, but would be overkill (no 4MV there) */ |
326 |
|
|
327 |
dx = mb->mvs[0].x; |
dx = mb->mvs[0].x; |
328 |
dy = mb->mvs[0].y; |
dy = mb->mvs[0].y; |
329 |
|
|
330 |
b_dx = mb->b_mvs[0].x; |
b_dx = mb->b_mvs[0].x; |
331 |
b_dy = mb->b_mvs[0].y; |
b_dy = mb->b_mvs[0].y; |
332 |
|
|
333 |
transfer_8to16sub2_c( |
for (k=0;k<4;k++) |
334 |
&dct_codes[0*64], |
{ |
335 |
cur->y + (i*16) + (j*16)*edged_width, |
transfer_8to16sub2_c(&dct_codes[k * 64], |
336 |
get_ref(f_ref->y, f_refh->y, f_refv->y, f_refhv->y, |
cur->y + (i * 16+(k&1)*8) + (j * 16+((k>>1)*8)) * edged_width, |
337 |
16*i, 16*j, 1, dx, dy, edged_width), |
get_ref(f_ref->y, f_refh->y, f_refv->y, |
338 |
get_ref(b_ref->y, b_refh->y, b_refv->y, b_refhv->y, |
f_refhv->y, 2*i + (k&1), 2*j + (k>>1), 8, dx, dy, |
339 |
16*i, 16*j, 1, b_dx, b_dy, edged_width), |
edged_width), |
340 |
|
get_ref(b_ref->y, b_refh->y, b_refv->y, |
341 |
|
b_refhv->y, 2*i + (k&1), 2 * j+(k>>1), 8, b_dx, b_dy, |
342 |
|
edged_width), |
343 |
edged_width); |
edged_width); |
344 |
|
} |
345 |
|
|
346 |
transfer_8to16sub2_c( |
dx = (dx & 3) ? (dx >> 1) | 1 : dx / 2; |
347 |
&dct_codes[1*64], |
dy = (dy & 3) ? (dy >> 1) | 1 : dy / 2; |
|
cur->y + (i*16+8) + (j*16)*edged_width, |
|
|
get_ref(f_ref->y, f_refh->y, f_refv->y, f_refhv->y, |
|
|
16*i+8, 16*j, 1, dx, dy, edged_width), |
|
|
get_ref(b_ref->y, b_refh->y, b_refv->y, b_refhv->y, |
|
|
16*i+8, 16*j, 1, b_dx, b_dy, edged_width), |
|
|
edged_width); |
|
348 |
|
|
349 |
transfer_8to16sub2_c( |
b_dx = (b_dx & 3) ? (b_dx >> 1) | 1 : b_dx / 2; |
350 |
&dct_codes[2*64], |
b_dy = (b_dy & 3) ? (b_dy >> 1) | 1 : b_dy / 2; |
351 |
cur->y + (i*16) + (j*16+8)*edged_width, |
|
352 |
get_ref(f_ref->y, f_refh->y, f_refv->y, f_refhv->y, |
transfer_8to16sub2_c(&dct_codes[4 * 64], |
353 |
16*i, 16*j+8, 1, dx, dy, edged_width), |
cur->u + (y * 8) * edged_width / 2 + (x * 8), |
354 |
get_ref(b_ref->y, b_refh->y, b_refv->y, b_refhv->y, |
get_ref(f_ref->u, f_refh->u, f_refv->u, |
355 |
16*i, 16*j+8, 1, b_dx, b_dy, edged_width), |
f_refhv->u, i, j, 8, dx, dy, |
356 |
edged_width); |
edged_width / 2), |
357 |
|
get_ref(b_ref->u, b_refh->u, b_refv->u, |
358 |
|
b_refhv->u, i, j, 8, b_dx, b_dy, |
359 |
|
edged_width / 2), |
360 |
|
edged_width / 2); |
361 |
|
|
362 |
transfer_8to16sub2_c( |
transfer_8to16sub2_c(&dct_codes[5 * 64], |
363 |
&dct_codes[3*64], |
cur->v + (y * 8) * edged_width / 2 + (x * 8), |
364 |
cur->y + (i*16+8) + (j*16+8)*edged_width, |
get_ref(f_ref->v, f_refh->v, f_refv->v, |
365 |
|
f_refhv->v, 8 * i, 8 * j, 1, dx, dy, |
366 |
|
edged_width / 2), |
367 |
|
get_ref(b_ref->v, b_refh->v, b_refv->v, |
368 |
|
b_refhv->v, 8 * i, 8 * j, 1, b_dx, b_dy, |
369 |
|
edged_width / 2), |
370 |
|
edged_width / 2); |
371 |
|
|
372 |
|
break; |
373 |
|
|
374 |
|
case MODE_DIRECT: |
375 |
|
|
376 |
|
for (k=0;k<4;k++) |
377 |
|
{ |
378 |
|
dx = mb->mvs[k].x; |
379 |
|
dy = mb->mvs[k].y; |
380 |
|
|
381 |
|
b_dx = mb->b_mvs[k].x; |
382 |
|
b_dy = mb->b_mvs[k].y; |
383 |
|
|
384 |
|
// fprintf(stderr,"Direct Vector %d -- %d:%d %d:%d\n",k,dx,dy,b_dx,b_dy); |
385 |
|
|
386 |
|
transfer_8to16sub2_c(&dct_codes[k * 64], |
387 |
|
cur->y + (i*16 + (k&1)*8) + (j*16 + (k>>1)*8 ) * edged_width, |
388 |
get_ref(f_ref->y, f_refh->y, f_refv->y, f_refhv->y, |
get_ref(f_ref->y, f_refh->y, f_refv->y, f_refhv->y, |
389 |
16*i + 8, 16*j + 8, 1, dx, dy, edged_width), |
2*i + (k&1), 2*j + (k>>1), 8, dx, dy, |
390 |
|
edged_width), |
391 |
get_ref(b_ref->y, b_refh->y, b_refv->y, b_refhv->y, |
get_ref(b_ref->y, b_refh->y, b_refv->y, b_refhv->y, |
392 |
16*i + 8, 16*j + 8, 1, b_dx, b_dy, edged_width), |
2*i + (k&1), 2*j + (k>>1), 8, b_dx, b_dy, |
393 |
|
edged_width), |
394 |
edged_width); |
edged_width); |
395 |
|
} |
396 |
|
|
397 |
|
sum = mb->mvs[0].x + mb->mvs[1].x + mb->mvs[2].x + mb->mvs[3].x; |
398 |
|
dx = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2)); |
399 |
|
|
400 |
dx = (dx & 3) ? (dx >> 1) | 1 : dx / 2; |
sum = mb->mvs[0].y + mb->mvs[1].y + mb->mvs[2].y + mb->mvs[3].y; |
401 |
dy = (dy & 3) ? (dy >> 1) | 1 : dy / 2; |
dy = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2)); |
402 |
|
|
|
b_dx = (b_dx & 3) ? (b_dx >> 1) | 1 : b_dx / 2; |
|
|
b_dy = (b_dy & 3) ? (b_dy >> 1) | 1 : b_dy / 2; |
|
403 |
|
|
404 |
transfer_8to16sub2_c( |
sum = mb->b_mvs[0].x + mb->b_mvs[1].x + mb->b_mvs[2].x + mb->b_mvs[3].x; |
405 |
&dct_codes[4*64], |
b_dx = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2)); |
406 |
|
|
407 |
|
sum = mb->b_mvs[0].y + mb->b_mvs[1].y + mb->b_mvs[2].y + mb->b_mvs[3].y; |
408 |
|
b_dy = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2)); |
409 |
|
|
410 |
|
/* // for QPel don't forget to always do |
411 |
|
|
412 |
|
if (quarterpel) |
413 |
|
sum /= 2; |
414 |
|
*/ |
415 |
|
|
416 |
|
transfer_8to16sub2_c(&dct_codes[4 * 64], |
417 |
cur->u + (y*8)*edged_width/2 + (x*8), |
cur->u + (y*8)*edged_width/2 + (x*8), |
418 |
get_ref(f_ref->u, f_refh->u, f_refv->u, f_refhv->u, |
get_ref(f_ref->u, f_refh->u, f_refv->u, |
419 |
8*i, 8*j, 1, dx, dy, edged_width/2), |
f_refhv->u, i, j, 8, dx, dy, |
420 |
get_ref(b_ref->u, b_refh->u, b_refv->u, b_refhv->u, |
edged_width / 2), |
421 |
8*i, 8*j, 1, b_dx, b_dy, edged_width/2), |
get_ref(b_ref->u, b_refh->u, b_refv->u, |
422 |
|
b_refhv->u, i, j, 8, b_dx, b_dy, |
423 |
|
edged_width / 2), |
424 |
edged_width/2); |
edged_width/2); |
425 |
|
|
426 |
transfer_8to16sub2_c( |
transfer_8to16sub2_c(&dct_codes[5 * 64], |
|
&dct_codes[5*64], |
|
427 |
cur->v + (y*8)*edged_width/2 + (x*8), |
cur->v + (y*8)*edged_width/2 + (x*8), |
428 |
get_ref(f_ref->v, f_refh->v, f_refv->v, f_refhv->v, |
get_ref(f_ref->v, f_refh->v, f_refv->v, |
429 |
8*i, 8*j, 1, dx, dy, edged_width/2), |
f_refhv->v, i, j, 8, dx, dy, |
430 |
get_ref(b_ref->v, b_refh->v, b_refv->v, b_refhv->v, |
edged_width / 2), |
431 |
8*i, 8*j, 1, b_dx, b_dy, edged_width/2), |
get_ref(b_ref->v, b_refh->v, b_refv->v, |
432 |
|
b_refhv->v, i, j, 8, b_dx, b_dy, |
433 |
|
edged_width / 2), |
434 |
edged_width/2); |
edged_width/2); |
435 |
|
|
|
break; |
|
436 |
|
|
|
case MODE_DIRECT : |
|
|
// todo |
|
437 |
break; |
break; |
438 |
} |
} |
|
|
|
439 |
} |
} |