1 |
/************************************************************************** |
/***************************************************************************** |
2 |
* |
* |
3 |
* XVID MPEG-4 VIDEO CODEC |
* XVID MPEG-4 VIDEO CODEC |
4 |
* - Motion sad header - |
* - Motion Estimation header - |
5 |
* |
* |
6 |
* This program is an implementation of a part of one or more MPEG-4 |
* Copyright(C) 2002 Christoph Lampert <gruel@web.de> |
|
* 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 |
* $Id$ |
* $Id$ |
54 |
* |
* |
55 |
***************************************************************************/ |
***************************************************************************/ |
60 |
#include "../portab.h" |
#include "../portab.h" |
61 |
#include "../global.h" |
#include "../global.h" |
62 |
|
|
63 |
|
/* hard coded motion search parameters for motion_est */ |
64 |
|
|
65 |
|
// very large value |
66 |
|
#define MV_MAX_ERROR (4096 * 256) |
67 |
|
|
68 |
|
// stop search if sdelta < THRESHOLD |
69 |
|
#define MV16_THRESHOLD 192 |
70 |
|
#define MV8_THRESHOLD 56 |
71 |
|
|
72 |
|
#define NEIGH_MOVE_THRESH 0 |
73 |
|
// how much a block's MV must differ from his neighbour |
74 |
|
// to be search for INTER4V. The more, the faster... |
75 |
|
|
76 |
|
/* sad16(0,0) bias; mpeg4 spec suggests nb/2+1 */ |
77 |
|
/* nb = vop pixels * 2^(bpp-8) */ |
78 |
|
#define MV16_00_BIAS (128+1) |
79 |
|
#define MV8_00_BIAS (0) |
80 |
|
|
81 |
|
/* INTER bias for INTER/INTRA decision; mpeg4 spec suggests 2*nb */ |
82 |
|
#define MV16_INTER_BIAS 512 |
83 |
|
|
84 |
|
/* Parameters which control inter/inter4v decision */ |
85 |
|
#define IMV16X16 5 |
86 |
|
|
87 |
|
/* vector map (vlc delta size) smoother parameters */ |
88 |
|
#define NEIGH_TEND_16X16 2 |
89 |
|
#define NEIGH_TEND_8X8 2 |
90 |
|
|
91 |
// fast ((A)/2)*2 |
// fast ((A)/2)*2 |
92 |
#define EVEN(A) (((A)<0?(A)+1:(A)) & ~1) |
#define EVEN(A) (((A)<0?(A)+1:(A)) & ~1) |
93 |
|
|
94 |
#define MVzero(A) ( ((A).x)==(0) && ((A).y)==(0) ) |
#define MVzero(A) ( ((A).x)==(0) && ((A).y)==(0) ) |
95 |
#define MVequal(A,B) ( ((A).x)==((B).x) && ((A).y)==((B).y) ) |
#define MVequal(A,B) ( ((A).x)==((B).x) && ((A).y)==((B).y) ) |
96 |
|
|
97 |
static const uint32_t roundtab[16] = |
/* default methods of Search, will be changed to function variable */ |
98 |
{ 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 2, 2 }; |
|
99 |
|
#ifndef SEARCH16 |
100 |
|
#define SEARCH16 PMVfastSearch16 |
101 |
|
//#define SEARCH16 FullSearch16 |
102 |
|
//#define SEARCH16 EPZSSearch16 |
103 |
|
#endif |
104 |
|
|
105 |
|
#ifndef SEARCH8 |
106 |
|
#define SEARCH8 PMVfastSearch8 |
107 |
|
//#define SEARCH8 EPZSSearch8 |
108 |
|
#endif |
109 |
|
|
110 |
|
|
111 |
/* |
/* |
112 |
* modified rounding tables |
* Calculate the min/max range (in halfpixels) |
113 |
* original tables see ISO spec tables 7-6 -> 7-9 |
* relative to the _MACROBLOCK_ position |
114 |
*/ |
*/ |
115 |
|
|
116 |
/* K = 4 */ |
static void __inline |
117 |
static const uint32_t roundtab_76[16] = { 0, 0, 0, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 1, 1 }; |
get_range(int32_t * const min_dx, |
118 |
/* K = 2 */ |
int32_t * const max_dx, |
119 |
static const uint32_t roundtab_78[8] = { 0, 0, 1, 1, 0, 0, 0, 1 }; |
int32_t * const min_dy, |
120 |
/* K = 1 */ |
int32_t * const max_dy, |
121 |
static const uint32_t roundtab_79[4] = { 0, 1, 0, 0 }; |
const uint32_t x, |
122 |
|
const uint32_t y, |
123 |
|
const uint32_t block_sz, /* block dimension, 8 or 16 */ |
124 |
|
|
125 |
|
const uint32_t width, |
126 |
|
const uint32_t height, |
127 |
|
const uint32_t fcode) |
128 |
|
{ |
129 |
|
|
130 |
|
const int search_range = 32 << (fcode - 1); |
131 |
|
const int high = search_range - 1; |
132 |
|
const int low = -search_range; |
133 |
|
|
134 |
|
/* convert full-pixel measurements to half pixel */ |
135 |
|
const int hp_width = 2 * width; |
136 |
|
const int hp_height = 2 * height; |
137 |
|
const int hp_edge = 2 * block_sz; |
138 |
|
|
139 |
|
/* we need _right end_ of block, not x-coordinate */ |
140 |
|
const int hp_x = 2 * (x) * block_sz; |
141 |
|
|
142 |
|
/* same for _bottom end_ */ |
143 |
|
const int hp_y = 2 * (y) * block_sz; |
144 |
|
|
145 |
|
*max_dx = MIN(high, hp_width - hp_x); |
146 |
|
*max_dy = MIN(high, hp_height - hp_y); |
147 |
|
*min_dx = MAX(low, -(hp_edge + hp_x)); |
148 |
|
*min_dy = MAX(low, -(hp_edge + hp_y)); |
149 |
|
|
150 |
|
} |
151 |
|
|
152 |
|
|
153 |
/* |
/* |
181 |
return refh + (int) ((x * block + (dx - 1) / 2) + (y * block + |
return refh + (int) ((x * block + (dx - 1) / 2) + (y * block + |
182 |
dy / 2) * stride); |
dy / 2) * stride); |
183 |
default: |
default: |
184 |
|
case 3: |
185 |
return refhv + (int) ((x * block + (dx - 1) / 2) + (y * block + |
return refhv + (int) ((x * block + (dx - 1) / 2) + (y * block + |
186 |
(dy - 1) / 2) * stride); |
(dy - 1) / 2) * stride); |
187 |
} |
} |
216 |
return refh + (int) ((x * block + ((mv->x) - 1) / 2) + (y * block + |
return refh + (int) ((x * block + ((mv->x) - 1) / 2) + (y * block + |
217 |
(mv->y) / 2) * stride); |
(mv->y) / 2) * stride); |
218 |
default: |
default: |
219 |
|
case 3: |
220 |
return refhv + (int) ((x * block + ((mv->x) - 1) / 2) + (y * block + |
return refhv + (int) ((x * block + ((mv->x) - 1) / 2) + (y * block + |
221 |
((mv->y) - |
((mv->y) - |
222 |
1) / 2) * stride); |
1) / 2) * stride); |
224 |
|
|
225 |
} |
} |
226 |
|
|
|
void MotionEstimationBVOP(MBParam * const pParam, |
|
|
FRAMEINFO * const frame, |
|
|
// forward (past) reference |
|
|
const int32_t time_bp, |
|
|
const int32_t time_pp, |
|
|
const MACROBLOCK * const f_mbs, |
|
|
const IMAGE * const f_ref, |
|
|
const IMAGE * const f_refH, |
|
|
const IMAGE * const f_refV, |
|
|
const IMAGE * const f_refHV, |
|
|
// backward (future) reference |
|
|
const MACROBLOCK * const b_mbs, |
|
|
const IMAGE * const b_ref, |
|
|
const IMAGE * const b_refH, |
|
|
const IMAGE * const b_refV, |
|
|
const IMAGE * const b_refHV); |
|
|
|
|
|
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); |
|
227 |
|
|
228 |
void |
static __inline const uint8_t * |
229 |
MotionEstimationHinted( MBParam * const pParam, |
get_iref(const uint8_t * const ref, |
230 |
|
const uint32_t x, |
231 |
|
const uint32_t y, |
232 |
|
const uint32_t block, /* block dimension, 8 or 16 */ |
233 |
|
|
234 |
|
const int32_t dx, |
235 |
|
const int32_t dy, |
236 |
|
const uint32_t stride) |
237 |
|
{ |
238 |
|
return ref + (int) ((x * block + dx / 2) + (y * block + dy / 2) * stride); |
239 |
|
} |
240 |
|
|
241 |
|
static __inline const uint8_t * |
242 |
|
get_iref_mv(const uint8_t * const ref, |
243 |
|
const uint32_t x, |
244 |
|
const uint32_t y, |
245 |
|
const uint32_t block, /* block dimension, 8 or 16 */ |
246 |
|
|
247 |
|
const VECTOR * mv, /* as usual measured in half-pel */ |
248 |
|
|
249 |
|
const uint32_t stride) |
250 |
|
{ |
251 |
|
return ref + (int) ((x * block + (mv->x) / 2) + (y * block + (mv->y) / 2) * stride); |
252 |
|
} |
253 |
|
|
254 |
|
|
255 |
|
/* prototypes for MainSearch functions, i.e. Diamondsearch, FullSearch or whatever */ |
256 |
|
|
257 |
|
typedef int32_t(MainSearch16Func) (const uint8_t * const pRef, |
258 |
|
const uint8_t * const pRefH, |
259 |
|
const uint8_t * const pRefV, |
260 |
|
const uint8_t * const pRefHV, |
261 |
|
const uint8_t * const cur, |
262 |
|
const int x, |
263 |
|
const int y, |
264 |
|
const int start_x, |
265 |
|
const int start_y, |
266 |
|
int iMinSAD, |
267 |
|
VECTOR * const currMV, |
268 |
|
const int center_x, |
269 |
|
const int center_y, |
270 |
|
const int32_t min_dx, |
271 |
|
const int32_t max_dx, |
272 |
|
const int32_t min_dy, |
273 |
|
const int32_t max_dy, |
274 |
|
const int32_t iEdgedWidth, |
275 |
|
const int32_t iDiamondSize, |
276 |
|
const int32_t iFcode, |
277 |
|
const int32_t iQuant, |
278 |
|
int iFound); |
279 |
|
|
280 |
|
typedef MainSearch16Func *MainSearch16FuncPtr; |
281 |
|
|
282 |
|
|
283 |
|
typedef int32_t(MainSearch8Func) (const uint8_t * const pRef, |
284 |
|
const uint8_t * const pRefH, |
285 |
|
const uint8_t * const pRefV, |
286 |
|
const uint8_t * const pRefHV, |
287 |
|
const uint8_t * const cur, |
288 |
|
const int x, |
289 |
|
const int y, |
290 |
|
const int start_x, |
291 |
|
const int start_y, |
292 |
|
int iMinSAD, |
293 |
|
VECTOR * const currMV, |
294 |
|
const int center_x, |
295 |
|
const int center_y, |
296 |
|
const int32_t min_dx, |
297 |
|
const int32_t max_dx, |
298 |
|
const int32_t min_dy, |
299 |
|
const int32_t max_dy, |
300 |
|
const int32_t iEdgedWidth, |
301 |
|
const int32_t iDiamondSize, |
302 |
|
const int32_t iFcode, |
303 |
|
const int32_t iQuant, |
304 |
|
int iFound); |
305 |
|
|
306 |
|
typedef MainSearch8Func *MainSearch8FuncPtr; |
307 |
|
|
308 |
|
|
309 |
|
/* prototypes for MotionEstimation functions, i.e. PMVfast, EPZS or whatever */ |
310 |
|
|
311 |
|
typedef int32_t(Search16Func) ( const uint8_t * const pRef, |
312 |
|
const uint8_t * const pRefH, |
313 |
|
const uint8_t * const pRefV, |
314 |
|
const uint8_t * const pRefHV, |
315 |
|
const IMAGE * const pCur, |
316 |
|
const int x, |
317 |
|
const int y, |
318 |
|
const int start_x, |
319 |
|
const int start_y, |
320 |
|
const int center_x, |
321 |
|
const int center_y, |
322 |
|
const uint32_t MotionFlags, |
323 |
|
const uint32_t iQuant, |
324 |
|
const uint32_t iFcode, |
325 |
|
const MBParam * const pParam, |
326 |
|
const MACROBLOCK * const pMBs, |
327 |
|
const MACROBLOCK * const prevMBs, |
328 |
|
VECTOR * const currMV, |
329 |
|
VECTOR * const currPMV); |
330 |
|
|
331 |
|
typedef Search16Func *Search16FuncPtr; |
332 |
|
|
333 |
|
typedef int32_t(Search8Func) ( const uint8_t * const pRef, |
334 |
|
const uint8_t * const pRefH, |
335 |
|
const uint8_t * const pRefV, |
336 |
|
const uint8_t * const pRefHV, |
337 |
|
const IMAGE * const pCur, |
338 |
|
const int x, |
339 |
|
const int y, |
340 |
|
const int start_x, |
341 |
|
const int start_y, |
342 |
|
const int center_x, |
343 |
|
const int center_y, |
344 |
|
const uint32_t MotionFlags, |
345 |
|
const uint32_t iQuant, |
346 |
|
const uint32_t iFcode, |
347 |
|
const MBParam * const pParam, |
348 |
|
const MACROBLOCK * const pMBs, |
349 |
|
const MACROBLOCK * const prevMBs, |
350 |
|
VECTOR * const currMV, |
351 |
|
VECTOR * const currPMV); |
352 |
|
|
353 |
|
typedef Search8Func *Search8FuncPtr; |
354 |
|
|
355 |
|
Search16Func PMVfastSearch16; |
356 |
|
Search16Func EPZSSearch16; |
357 |
|
Search16Func PMVfastIntSearch16; |
358 |
|
|
359 |
|
Search8Func PMVfastSearch8; |
360 |
|
Search8Func EPZSSearch8; |
361 |
|
|
362 |
|
|
363 |
|
bool |
364 |
|
MotionEstimation(MBParam * const pParam, |
365 |
FRAMEINFO * const current, |
FRAMEINFO * const current, |
366 |
FRAMEINFO * const reference, |
FRAMEINFO * const reference, |
367 |
const IMAGE * const pRefH, |
const IMAGE * const pRefH, |
368 |
const IMAGE * const pRefV, |
const IMAGE * const pRefV, |
369 |
const IMAGE * const pRefHV); |
const IMAGE * const pRefHV, |
370 |
|
const uint32_t iLimit); |
371 |
|
|
372 |
int |
typedef int32_t(Halfpel8_RefineFunc) (const uint8_t * const pRef, |
373 |
MEanalysis( const IMAGE * const pRef, |
const uint8_t * const pRefH, |
374 |
const IMAGE * const pCurrent, |
const uint8_t * const pRefV, |
375 |
MBParam * const pParam, |
const uint8_t * const pRefHV, |
376 |
MACROBLOCK * const pMBs, |
const uint8_t * const cur, |
377 |
const uint32_t iFcode); |
const int x, |
378 |
|
const int y, |
379 |
int |
VECTOR * const currMV, |
380 |
FindFcode( const MBParam * const pParam, |
int32_t iMinSAD, |
381 |
const FRAMEINFO * const current); |
const int center_x, |
382 |
|
const int center_y, |
383 |
|
const int32_t min_dx, |
384 |
|
const int32_t max_dx, |
385 |
|
const int32_t min_dy, |
386 |
|
const int32_t max_dy, |
387 |
|
const int32_t iFcode, |
388 |
|
const int32_t iQuant, |
389 |
|
const int32_t iEdgedWidth); |
390 |
|
|
391 |
|
typedef Halfpel8_RefineFunc *Halfpel8_RefineFuncPtr; |
392 |
|
extern Halfpel8_RefineFuncPtr Halfpel8_Refine; |
393 |
|
Halfpel8_RefineFunc Halfpel8_Refine_c; |
394 |
|
Halfpel8_RefineFunc Halfpel8_Refine_ia64; |
395 |
|
|
396 |
|
|
397 |
#endif /* _MOTION_H_ */ |
#endif /* _MOTION_H_ */ |