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" |
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 |
} |
} |
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, |
compensate8x8_halfpel(&dct_codes[4 * 64], cur->u, ref->u, refh->u, |
136 |
refv->u, refhv->u, 8 * i, 8 * j, dx, dy, |
refv->u, refhv->u, 8 * i, 8 * j, dx, dy, |
139 |
refv->v, refhv->v, 8 * i, 8 * j, dx, dy, |
refv->v, refhv->v, 8 * i, 8 * j, dx, dy, |
140 |
edged_width / 2); |
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, |
interpolate8x8_switch(refv->u, ref->u, 8 * i, 8 * j, dx, dy, |
145 |
edged_width / 2, rounding); |
edged_width / 2, rounding); |
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, |
refv->v + 8 * j * edged_width / 2 + 8 * i, |
156 |
edged_width / 2); |
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; |
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; |
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(&dct_codes[0 * 64], |
for (k=0;k<4;k++) |
334 |
cur->y + (i * 16) + (j * 16) * edged_width, |
{ |
335 |
get_ref(f_ref->y, f_refh->y, f_refv->y, |
transfer_8to16sub2_c(&dct_codes[k * 64], |
336 |
f_refhv->y, 16 * i, 16 * j, 1, dx, dy, |
cur->y + (i * 16+(k&1)*8) + (j * 16+((k>>1)*8)) * edged_width, |
|
edged_width), get_ref(b_ref->y, b_refh->y, |
|
|
b_refv->y, |
|
|
b_refhv->y, 16 * i, |
|
|
16 * j, 1, b_dx, |
|
|
b_dy, edged_width), |
|
|
edged_width); |
|
|
|
|
|
transfer_8to16sub2_c(&dct_codes[1 * 64], |
|
|
cur->y + (i * 16 + 8) + (j * 16) * edged_width, |
|
337 |
get_ref(f_ref->y, f_refh->y, f_refv->y, |
get_ref(f_ref->y, f_refh->y, f_refv->y, |
338 |
f_refhv->y, 16 * i + 8, 16 * j, 1, dx, dy, |
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, |
|
|
16 * i + 8, 16 * j, |
|
|
1, b_dx, b_dy, |
|
339 |
edged_width), |
edged_width), |
340 |
edged_width); |
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, |
|
transfer_8to16sub2_c(&dct_codes[2 * 64], |
|
|
cur->y + (i * 16) + (j * 16 + 8) * edged_width, |
|
|
get_ref(f_ref->y, f_refh->y, f_refv->y, |
|
|
f_refhv->y, 16 * i, 16 * j + 8, 1, dx, dy, |
|
|
edged_width), get_ref(b_ref->y, b_refh->y, |
|
|
b_refv->y, |
|
|
b_refhv->y, 16 * i, |
|
|
16 * j + 8, 1, b_dx, |
|
|
b_dy, edged_width), |
|
|
edged_width); |
|
|
|
|
|
transfer_8to16sub2_c(&dct_codes[3 * 64], |
|
|
cur->y + (i * 16 + 8) + (j * 16 + |
|
|
8) * edged_width, |
|
|
get_ref(f_ref->y, f_refh->y, f_refv->y, |
|
|
f_refhv->y, 16 * i + 8, 16 * j + 8, 1, dx, |
|
|
dy, edged_width), get_ref(b_ref->y, |
|
|
b_refh->y, |
|
|
b_refv->y, |
|
|
b_refhv->y, |
|
|
16 * i + 8, |
|
|
16 * j + 8, 1, |
|
|
b_dx, b_dy, |
|
342 |
edged_width), |
edged_width), |
343 |
edged_width); |
edged_width); |
344 |
|
} |
345 |
|
|
346 |
dx = (dx & 3) ? (dx >> 1) | 1 : dx / 2; |
dx = (dx & 3) ? (dx >> 1) | 1 : dx / 2; |
347 |
dy = (dy & 3) ? (dy >> 1) | 1 : dy / 2; |
dy = (dy & 3) ? (dy >> 1) | 1 : dy / 2; |
352 |
transfer_8to16sub2_c(&dct_codes[4 * 64], |
transfer_8to16sub2_c(&dct_codes[4 * 64], |
353 |
cur->u + (y * 8) * edged_width / 2 + (x * 8), |
cur->u + (y * 8) * edged_width / 2 + (x * 8), |
354 |
get_ref(f_ref->u, f_refh->u, f_refv->u, |
get_ref(f_ref->u, f_refh->u, f_refv->u, |
355 |
f_refhv->u, 8 * i, 8 * j, 1, dx, dy, |
f_refhv->u, i, j, 8, dx, dy, |
356 |
edged_width / 2), get_ref(b_ref->u, |
edged_width / 2), |
357 |
b_refh->u, |
get_ref(b_ref->u, b_refh->u, b_refv->u, |
358 |
b_refv->u, |
b_refhv->u, i, j, 8, b_dx, b_dy, |
359 |
b_refhv->u, |
edged_width / 2), |
|
8 * i, 8 * j, 1, |
|
|
b_dx, b_dy, |
|
|
edged_width / |
|
|
2), |
|
360 |
edged_width / 2); |
edged_width / 2); |
361 |
|
|
362 |
transfer_8to16sub2_c(&dct_codes[5 * 64], |
transfer_8to16sub2_c(&dct_codes[5 * 64], |
363 |
cur->v + (y * 8) * edged_width / 2 + (x * 8), |
cur->v + (y * 8) * edged_width / 2 + (x * 8), |
364 |
get_ref(f_ref->v, f_refh->v, f_refv->v, |
get_ref(f_ref->v, f_refh->v, f_refv->v, |
365 |
f_refhv->v, 8 * i, 8 * j, 1, dx, dy, |
f_refhv->v, 8 * i, 8 * j, 1, dx, dy, |
366 |
edged_width / 2), get_ref(b_ref->v, |
edged_width / 2), |
367 |
b_refh->v, |
get_ref(b_ref->v, b_refh->v, b_refv->v, |
368 |
b_refv->v, |
b_refhv->v, 8 * i, 8 * j, 1, b_dx, b_dy, |
369 |
b_refhv->v, |
edged_width / 2), |
|
8 * i, 8 * j, 1, |
|
|
b_dx, b_dy, |
|
|
edged_width / |
|
|
2), |
|
370 |
edged_width / 2); |
edged_width / 2); |
371 |
|
|
372 |
break; |
break; |
373 |
|
|
374 |
case MODE_DIRECT: |
case MODE_DIRECT: |
375 |
// todo |
|
376 |
break; |
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, |
389 |
|
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, |
392 |
|
2*i + (k&1), 2*j + (k>>1), 8, b_dx, b_dy, |
393 |
|
edged_width), |
394 |
|
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 |
|
sum = mb->mvs[0].y + mb->mvs[1].y + mb->mvs[2].y + mb->mvs[3].y; |
401 |
|
dy = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2)); |
402 |
|
|
403 |
|
|
404 |
|
sum = mb->b_mvs[0].x + mb->b_mvs[1].x + mb->b_mvs[2].x + mb->b_mvs[3].x; |
405 |
|
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), |
418 |
|
get_ref(f_ref->u, f_refh->u, f_refv->u, |
419 |
|
f_refhv->u, i, j, 8, dx, dy, |
420 |
|
edged_width / 2), |
421 |
|
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); |
425 |
|
|
426 |
|
transfer_8to16sub2_c(&dct_codes[5 * 64], |
427 |
|
cur->v + (y * 8) * edged_width / 2 + (x * 8), |
428 |
|
get_ref(f_ref->v, f_refh->v, f_refv->v, |
429 |
|
f_refhv->v, i, j, 8, dx, dy, |
430 |
|
edged_width / 2), |
431 |
|
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); |
435 |
|
|
436 |
|
|
437 |
|
break; |
438 |
|
} |
439 |
} |
} |