1 |
/************************************************************************** |
/***************************************************************************** |
2 |
* |
* |
3 |
* XVID MPEG-4 VIDEO CODEC |
* XVID MPEG-4 VIDEO CODEC |
4 |
* - Motion estimation header - |
* - Motion Estimation related 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> |
7 |
* Video tools as specified in ISO/IEC 14496-2 standard. Those intending |
* 2002 Michael Militzer <michael@xvid.org> |
8 |
* to use this software module in hardware or software products are |
* 2002-2003 Radoslaw Czyz <xvid@syskin.cjb.net> |
|
* 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. |
|
9 |
* |
* |
10 |
* This program is free software; you can redistribute it and/or modify |
* This program is free software; you can redistribute it and/or modify |
11 |
* it under the terms of the GNU General Public License as published by |
* it under the terms of the GNU General Public License as published by |
23 |
* |
* |
24 |
* $Id$ |
* $Id$ |
25 |
* |
* |
26 |
***************************************************************************/ |
****************************************************************************/ |
27 |
|
|
28 |
#ifndef _MOTION_EST_H_ |
#ifndef _MOTION_EST_H_ |
29 |
#define _MOTION_EST_H_ |
#define _MOTION_EST_H_ |
34 |
|
|
35 |
/* hard coded motion search parameters for motion_est and smp_motion_est */ |
/* hard coded motion search parameters for motion_est and smp_motion_est */ |
36 |
|
|
37 |
// very large value |
/* very large value */ |
38 |
#define MV_MAX_ERROR (4096 * 256) |
#define MV_MAX_ERROR (4096 * 256) |
39 |
|
|
40 |
/* INTER bias for INTER/INTRA decision; mpeg4 spec suggests 2*nb */ |
/* INTER bias for INTER/INTRA decision; mpeg4 spec suggests 2*nb */ |
45 |
#define NEIGH_TEND_8X8 40.0 |
#define NEIGH_TEND_8X8 40.0 |
46 |
#define NEIGH_8X8_BIAS 30 |
#define NEIGH_8X8_BIAS 30 |
47 |
|
|
48 |
|
#define BITS_MULT 16 |
49 |
|
|
50 |
/* Parameters which control inter/inter4v decision */ |
/* Parameters which control inter/inter4v decision */ |
51 |
#define IMV16X16 2 |
#define IMV16X16 2 |
52 |
|
|
86 |
(int)(21.2656*NEIGH_TEND_8X8 + 0.5), (int)(24.8580*NEIGH_TEND_8X8 + 0.5), |
(int)(21.2656*NEIGH_TEND_8X8 + 0.5), (int)(24.8580*NEIGH_TEND_8X8 + 0.5), |
87 |
(int)(29.6436*NEIGH_TEND_8X8 + 0.5), (int)(36.4949*NEIGH_TEND_8X8 + 0.5) }; |
(int)(29.6436*NEIGH_TEND_8X8 + 0.5), (int)(36.4949*NEIGH_TEND_8X8 + 0.5) }; |
88 |
|
|
89 |
// mv.length table |
/* mv.length table */ |
90 |
static const uint32_t mvtab[33] = { |
static const int mvtab[64] = { |
91 |
1, 2, 3, 4, 6, 7, 7, 7, |
1, 2, 3, 4, 6, 7, 7, 7, |
92 |
9, 9, 9, 10, 10, 10, 10, 10, |
9, 9, 9, 10, 10, 10, 10, 10, |
93 |
10, 10, 10, 10, 10, 10, 10, 10, |
10, 10, 10, 10, 10, 10, 10, 10, |
94 |
10, 11, 11, 11, 11, 11, 11, 12, 12 |
10, 11, 11, 11, 11, 11, 11, 12, |
95 |
}; |
12, 12, 12, 12, 12, 12, 12, 12, |
96 |
|
12, 12, 12, 12, 12, 12, 12, 12, |
97 |
|
12, 12, 12, 12, 12, 12, 12, 12, 12 }; |
98 |
|
|
99 |
static const int DQtab[4] = { |
static const int DQtab[4] = { |
100 |
-1, -2, 1, 2 |
-1, -2, 1, 2 |
104 |
|
|
105 |
typedef struct |
typedef struct |
106 |
{ |
{ |
107 |
// general fields |
/* general fields */ |
108 |
int max_dx, min_dx, max_dy, min_dy; |
int max_dx, min_dx, max_dy, min_dy; |
109 |
uint32_t rounding; |
uint32_t rounding; |
110 |
VECTOR predMV; |
VECTOR predMV; |
111 |
VECTOR * currentMV; |
VECTOR * currentMV; |
112 |
VECTOR * currentQMV; |
VECTOR * currentQMV; |
113 |
int32_t * iMinSAD; |
int32_t * iMinSAD; |
114 |
const uint8_t * Ref; |
const uint8_t * RefP[6]; /* N, V, H, HV, cU, cV */ |
|
const uint8_t * RefH; |
|
|
const uint8_t * RefV; |
|
|
const uint8_t * RefHV; |
|
|
const uint8_t * RefCU; |
|
|
const uint8_t * RefCV; |
|
115 |
const uint8_t * CurU; |
const uint8_t * CurU; |
116 |
const uint8_t * CurV; |
const uint8_t * CurV; |
117 |
uint8_t * RefQ; |
uint8_t * RefQ; |
124 |
int qpel, qpel_precision; |
int qpel, qpel_precision; |
125 |
int chroma; |
int chroma; |
126 |
int rrv; |
int rrv; |
|
//fields for interpolate and direct modes |
|
|
const uint8_t * bRef; |
|
|
const uint8_t * bRefH; |
|
|
const uint8_t * bRefV; |
|
|
const uint8_t * bRefHV; |
|
|
const uint8_t * b_RefCU; |
|
|
const uint8_t * b_RefCV; |
|
127 |
|
|
128 |
|
/* fields for interpolate and direct modes */ |
129 |
|
const uint8_t * b_RefP[6]; /* N, V, H, HV, cU, cV */ |
130 |
VECTOR bpredMV; |
VECTOR bpredMV; |
131 |
uint32_t bFcode; |
uint32_t bFcode; |
132 |
// fields for direct mode |
|
133 |
|
/* fields for direct mode */ |
134 |
VECTOR directmvF[4]; |
VECTOR directmvF[4]; |
135 |
VECTOR directmvB[4]; |
VECTOR directmvB[4]; |
136 |
const VECTOR * referencemv; |
const VECTOR * referencemv; |
137 |
|
|
138 |
|
/* BITS/R-D stuff */ |
139 |
|
int16_t * dctSpace; |
140 |
|
uint32_t iQuant; |
141 |
|
uint32_t quant_type; |
142 |
|
|
143 |
} SearchData; |
} SearchData; |
144 |
|
|
145 |
|
|
208 |
const IMAGE * const pRefH, |
const IMAGE * const pRefH, |
209 |
const IMAGE * const pRefV, |
const IMAGE * const pRefV, |
210 |
const IMAGE * const pRefHV, |
const IMAGE * const pRefHV, |
211 |
|
const IMAGE * const pGMC, |
212 |
const uint32_t iLimit); |
const uint32_t iLimit); |
213 |
|
|
214 |
static void |
static void |
220 |
const int x, |
const int x, |
221 |
const int y, |
const int y, |
222 |
const uint32_t MotionFlags, |
const uint32_t MotionFlags, |
223 |
const uint32_t GlobalFlags, |
const uint32_t VopFlags, |
224 |
const uint32_t iQuant, |
const uint32_t VolFlags, |
225 |
SearchData * const Data, |
SearchData * const Data, |
226 |
const MBParam * const pParam, |
const MBParam * const pParam, |
227 |
const MACROBLOCK * const pMBs, |
const MACROBLOCK * const pMBs, |
228 |
const MACROBLOCK * const prevMBs, |
const MACROBLOCK * const prevMBs, |
|
int inter4v, |
|
229 |
MACROBLOCK * const pMB); |
MACROBLOCK * const pMB); |
230 |
|
|
231 |
|
static __inline void |
232 |
|
GMEanalyzeMB (const uint8_t * const pCur, |
233 |
|
const uint8_t * const pRef, |
234 |
|
const uint8_t * const pRefH, |
235 |
|
const uint8_t * const pRefV, |
236 |
|
const uint8_t * const pRefHV, |
237 |
|
const int x, |
238 |
|
const int y, |
239 |
|
const MBParam * const pParam, |
240 |
|
MACROBLOCK * const pMBs, |
241 |
|
SearchData * const Data); |
242 |
|
|
243 |
|
void |
244 |
|
GMEanalysis(const MBParam * const pParam, |
245 |
|
const FRAMEINFO * const current, |
246 |
|
const FRAMEINFO * const reference, |
247 |
|
const IMAGE * const pRefH, |
248 |
|
const IMAGE * const pRefV, |
249 |
|
const IMAGE * const pRefHV); |
250 |
|
|
251 |
|
|
252 |
|
|
253 |
|
WARPPOINTS |
254 |
|
GlobalMotionEst(MACROBLOCK * const pMBs, |
255 |
|
const MBParam * const pParam, |
256 |
|
const FRAMEINFO * const current, |
257 |
|
const FRAMEINFO * const reference, |
258 |
|
const IMAGE * const pRefH, |
259 |
|
const IMAGE * const pRefV, |
260 |
|
const IMAGE * const pRefHV ); |
261 |
|
|
262 |
|
|
263 |
static WARPPOINTS |
int |
264 |
GlobalMotionEst(const MACROBLOCK * const pMBs, |
GlobalMotionEstRefine(WARPPOINTS *const startwp, |
265 |
|
MACROBLOCK * const pMBs, |
266 |
const MBParam * const pParam, |
const MBParam * const pParam, |
267 |
const FRAMEINFO * const current, |
const FRAMEINFO * const current, |
268 |
const FRAMEINFO * const reference, |
const FRAMEINFO * const reference, |
269 |
|
const IMAGE * const pCurr, |
270 |
|
const IMAGE * const pRef, |
271 |
const IMAGE * const pRefH, |
const IMAGE * const pRefH, |
272 |
const IMAGE * const pRefV, |
const IMAGE * const pRefV, |
273 |
const IMAGE * const pRefHV ); |
const IMAGE * const pRefHV ); |
274 |
|
|
275 |
|
|
276 |
|
|
277 |
|
|
278 |
|
int |
279 |
|
globalSAD(const WARPPOINTS *const wp, |
280 |
|
const MBParam * const pParam, |
281 |
|
const MACROBLOCK * const pMBs, |
282 |
|
const FRAMEINFO * const current, |
283 |
|
const IMAGE * const pRef, |
284 |
|
const IMAGE * const pCurr, |
285 |
|
uint8_t *const GMCblock); |
286 |
|
|
287 |
#define iDiamondSize 2 |
#define iDiamondSize 2 |
288 |
|
|
289 |
static __inline uint32_t |
static __inline uint32_t |
290 |
MakeGoodMotionFlags(const uint32_t MotionFlags, const uint32_t GlobalFlags) |
MakeGoodMotionFlags(const uint32_t MotionFlags, const uint32_t VopFlags, const uint32_t VolFlags) |
291 |
{ |
{ |
292 |
uint32_t Flags = MotionFlags; |
uint32_t Flags = MotionFlags; |
293 |
|
|
294 |
if (!(GlobalFlags & XVID_MODEDECISION_BITS)) |
if (!(VopFlags & XVID_VOP_MODEDECISION_BITS)) |
295 |
Flags &= ~(QUARTERPELREFINE16_BITS+QUARTERPELREFINE8_BITS+HALFPELREFINE16_BITS+HALFPELREFINE8_BITS+EXTSEARCH_BITS); |
Flags &= ~(XVID_ME_QUARTERPELREFINE16_BITS+XVID_ME_QUARTERPELREFINE8_BITS+XVID_ME_HALFPELREFINE16_BITS+XVID_ME_HALFPELREFINE8_BITS+XVID_ME_EXTSEARCH_BITS); |
296 |
|
|
297 |
if (Flags & EXTSEARCH_BITS) |
if (Flags & XVID_ME_EXTSEARCH_BITS) |
298 |
Flags |= HALFPELREFINE16_BITS; |
Flags |= XVID_ME_HALFPELREFINE16_BITS; |
299 |
|
|
300 |
if (Flags & EXTSEARCH_BITS && MotionFlags & PMV_EXTSEARCH8) |
if (Flags & XVID_ME_EXTSEARCH_BITS && MotionFlags & XVID_ME_EXTSEARCH8) |
301 |
Flags |= HALFPELREFINE8_BITS; |
Flags |= XVID_ME_HALFPELREFINE8_BITS; |
302 |
|
|
303 |
if (Flags & HALFPELREFINE16_BITS) |
if (Flags & XVID_ME_HALFPELREFINE16_BITS) |
304 |
Flags |= QUARTERPELREFINE16_BITS; |
Flags |= XVID_ME_QUARTERPELREFINE16_BITS; |
305 |
|
|
306 |
if (Flags & HALFPELREFINE8_BITS) { |
if (Flags & XVID_ME_HALFPELREFINE8_BITS) { |
307 |
Flags |= QUARTERPELREFINE8_BITS; |
Flags |= XVID_ME_QUARTERPELREFINE8_BITS; |
308 |
Flags &= ~PMV_HALFPELREFINE8; |
Flags &= ~XVID_ME_HALFPELREFINE8; |
309 |
} |
} |
310 |
|
|
311 |
if (Flags & QUARTERPELREFINE8_BITS) |
if (Flags & XVID_ME_QUARTERPELREFINE8_BITS) |
312 |
Flags &= ~PMV_QUARTERPELREFINE8; |
Flags &= ~XVID_ME_QUARTERPELREFINE8; |
313 |
|
|
314 |
if (!(GlobalFlags & XVID_QUARTERPEL)) |
if (!(VolFlags & XVID_VOL_QUARTERPEL)) |
315 |
Flags &= ~(PMV_QUARTERPELREFINE16+PMV_QUARTERPELREFINE8+QUARTERPELREFINE16_BITS+QUARTERPELREFINE8_BITS); |
Flags &= ~(XVID_ME_QUARTERPELREFINE16+XVID_ME_QUARTERPELREFINE8+XVID_ME_QUARTERPELREFINE16_BITS+XVID_ME_QUARTERPELREFINE8_BITS); |
316 |
|
|
317 |
if (!(GlobalFlags & XVID_HALFPEL)) |
if (!(VopFlags & XVID_VOP_HALFPEL)) |
318 |
Flags &= ~(PMV_EXTSEARCH16+PMV_HALFPELREFINE16+PMV_HALFPELREFINE8+HALFPELREFINE16_BITS+HALFPELREFINE8_BITS); |
Flags &= ~(XVID_ME_EXTSEARCH16+XVID_ME_HALFPELREFINE16+XVID_ME_HALFPELREFINE8+XVID_ME_HALFPELREFINE16_BITS+XVID_ME_HALFPELREFINE8_BITS); |
319 |
|
|
320 |
if (GlobalFlags & (XVID_GREYSCALE + XVID_REDUCED)) |
if ((VopFlags & XVID_VOP_GREYSCALE) || (VopFlags & XVID_VOP_REDUCED)) |
321 |
Flags &= ~(PMV_CHROMA16 + PMV_CHROMA8); |
Flags &= ~(XVID_ME_CHROMA16 + XVID_ME_CHROMA8); |
322 |
|
|
323 |
return Flags; |
return Flags; |
324 |
} |
} |
329 |
#include "../quant/quant_mpeg4.h" |
#include "../quant/quant_mpeg4.h" |
330 |
#include "../quant/quant_h263.h" |
#include "../quant/quant_h263.h" |
331 |
#include "../bitstream/vlc_codes.h" |
#include "../bitstream/vlc_codes.h" |
332 |
|
#include "../dct/fdct.h" |
333 |
|
|
334 |
static int |
static int |
335 |
CountMBBitsInter(SearchData * const Data, |
CountMBBitsInter(SearchData * const Data, |
350 |
int CodeCoeffIntra_CalcBits(const int16_t qcoeff[64], const uint16_t * zigzag); |
int CodeCoeffIntra_CalcBits(const int16_t qcoeff[64], const uint16_t * zigzag); |
351 |
int CodeCoeffInter_CalcBits(const int16_t qcoeff[64], const uint16_t * zigzag); |
int CodeCoeffInter_CalcBits(const int16_t qcoeff[64], const uint16_t * zigzag); |
352 |
|
|
353 |
|
/* one over lambda for R-D mode decision and motion search */ |
354 |
|
#define LAMBDA ( (int)(BITS_MULT/1.0) ) |
355 |
|
|
356 |
|
static __inline unsigned int |
357 |
|
Block_CalcBits( int16_t * const coeff, |
358 |
|
int16_t * const data, |
359 |
|
int16_t * const dqcoeff, |
360 |
|
const uint32_t quant, const int quant_type, |
361 |
|
uint32_t * cbp, |
362 |
|
const int block) |
363 |
|
{ |
364 |
|
int sum; |
365 |
|
int bits; |
366 |
|
int distortion = 0; |
367 |
|
int i; |
368 |
|
|
369 |
|
fdct(data); |
370 |
|
|
371 |
|
if (quant_type == 0) sum = quant_inter(coeff, data, quant); |
372 |
|
else sum = quant4_inter(coeff, data, quant); |
373 |
|
|
374 |
|
if (sum > 0) { |
375 |
|
*cbp |= 1 << (5 - block); |
376 |
|
bits = BITS_MULT * CodeCoeffInter_CalcBits(coeff, scan_tables[0]); |
377 |
|
} else bits = 0; |
378 |
|
|
379 |
|
if (quant_type == 0) dequant_inter(dqcoeff, coeff, quant); |
380 |
|
else dequant4_inter(dqcoeff, coeff, quant); |
381 |
|
|
382 |
|
for (i = 0; i < 64; i++) { |
383 |
|
distortion += (data[i] - dqcoeff[i])*(data[i] - dqcoeff[i]); |
384 |
|
} |
385 |
|
|
386 |
|
bits += (LAMBDA*distortion)/(quant*quant); |
387 |
|
|
388 |
|
return bits; |
389 |
|
} |
390 |
|
|
391 |
|
static __inline unsigned int |
392 |
|
Block_CalcBitsIntra(int16_t * const coeff, |
393 |
|
int16_t * const data, |
394 |
|
int16_t * const dqcoeff, |
395 |
|
const uint32_t quant, const int quant_type, |
396 |
|
uint32_t * cbp, |
397 |
|
const int block, |
398 |
|
int * dcpred) |
399 |
|
{ |
400 |
|
int bits, i; |
401 |
|
int distortion = 0; |
402 |
|
uint32_t iDcScaler = get_dc_scaler(quant, block < 4); |
403 |
|
int b_dc; |
404 |
|
|
405 |
|
fdct(data); |
406 |
|
data[0] -= 1024; |
407 |
|
|
408 |
|
if (quant_type == 0) quant_intra(coeff, data, quant, iDcScaler); |
409 |
|
else quant4_intra(coeff, data, quant, iDcScaler); |
410 |
|
|
411 |
|
b_dc = coeff[0]; |
412 |
|
if (block < 4) { |
413 |
|
coeff[0] -= *dcpred; |
414 |
|
*dcpred = b_dc; |
415 |
|
} |
416 |
|
|
417 |
|
bits = BITS_MULT*CodeCoeffIntra_CalcBits(coeff, scan_tables[0]); |
418 |
|
if (bits != 0) *cbp |= 1 << (5 - block); |
419 |
|
|
420 |
|
if (block < 4) bits += BITS_MULT*dcy_tab[coeff[0] + 255].len; |
421 |
|
else bits += BITS_MULT*dcc_tab[coeff[0] + 255].len; |
422 |
|
|
423 |
|
coeff[0] = b_dc; |
424 |
|
if (quant_type == 0) dequant_intra(dqcoeff, coeff, quant, iDcScaler); |
425 |
|
else dequant4_intra(dqcoeff, coeff, quant, iDcScaler); |
426 |
|
|
427 |
|
for (i = 0; i < 64; i++) { |
428 |
|
distortion += (data[i] - dqcoeff[i])*(data[i] - dqcoeff[i]); |
429 |
|
} |
430 |
|
|
431 |
|
bits += (LAMBDA*distortion)/(quant*quant); |
432 |
|
|
433 |
|
return bits; |
434 |
|
} |
435 |
|
|
436 |
#endif /* _MOTION_EST_H_ */ |
#endif /* _MOTION_EST_H_ */ |