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 |
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 |
* $Id: motion_est.h,v 1.3.2.7 2003-05-14 20:23:02 edgomez Exp $ |
* $Id: motion_est.h,v 1.3.2.13 2003-07-13 09:57:28 syskin Exp $ |
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 */ |
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 int mvtab[64] = { |
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, |
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 * RefP[6]; // N, V, H, HV, cU, cV |
const uint8_t * RefP[6]; /* N, V, H, HV, cU, cV */ |
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; |
127 |
//fields for interpolate and direct modes |
|
128 |
const uint8_t * b_RefP[6]; // N, V, H, HV, cU, cV |
/* 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 |
// BITS/R-D stuff |
|
138 |
|
/* BITS/R-D stuff */ |
139 |
int16_t * dctSpace; |
int16_t * dctSpace; |
140 |
uint32_t iQuant; |
uint32_t iQuant; |
141 |
uint32_t quant_type; |
uint32_t quant_type; |
142 |
|
int * cbp; |
143 |
|
|
144 |
} SearchData; |
} SearchData; |
145 |
|
|
209 |
const IMAGE * const pRefH, |
const IMAGE * const pRefH, |
210 |
const IMAGE * const pRefV, |
const IMAGE * const pRefV, |
211 |
const IMAGE * const pRefHV, |
const IMAGE * const pRefHV, |
212 |
|
const IMAGE * const pGMC, |
213 |
const uint32_t iLimit); |
const uint32_t iLimit); |
214 |
|
|
215 |
static void |
static void |
229 |
const MACROBLOCK * const prevMBs, |
const MACROBLOCK * const prevMBs, |
230 |
MACROBLOCK * const pMB); |
MACROBLOCK * const pMB); |
231 |
|
|
232 |
static WARPPOINTS |
static __inline void |
233 |
GlobalMotionEst(const MACROBLOCK * const pMBs, |
GMEanalyzeMB (const uint8_t * const pCur, |
234 |
|
const uint8_t * const pRef, |
235 |
|
const uint8_t * const pRefH, |
236 |
|
const uint8_t * const pRefV, |
237 |
|
const uint8_t * const pRefHV, |
238 |
|
const int x, |
239 |
|
const int y, |
240 |
const MBParam * const pParam, |
const MBParam * const pParam, |
241 |
|
MACROBLOCK * const pMBs, |
242 |
|
SearchData * const Data); |
243 |
|
|
244 |
|
void |
245 |
|
GMEanalysis(const MBParam * const pParam, |
246 |
const FRAMEINFO * const current, |
const FRAMEINFO * const current, |
247 |
const FRAMEINFO * const reference, |
const FRAMEINFO * const reference, |
248 |
const IMAGE * const pRefH, |
const IMAGE * const pRefH, |
249 |
const IMAGE * const pRefV, |
const IMAGE * const pRefV, |
250 |
const IMAGE * const pRefHV ); |
const IMAGE * const pRefHV ); |
251 |
|
|
252 |
|
|
253 |
|
|
254 |
|
WARPPOINTS |
255 |
|
GlobalMotionEst(MACROBLOCK * const pMBs, |
256 |
|
const MBParam * const pParam, |
257 |
|
const FRAMEINFO * const current, |
258 |
|
const FRAMEINFO * const reference, |
259 |
|
const IMAGE * const pRefH, |
260 |
|
const IMAGE * const pRefV, |
261 |
|
const IMAGE * const pRefHV ); |
262 |
|
|
263 |
|
|
264 |
|
int |
265 |
|
GlobalMotionEstRefine(WARPPOINTS *const startwp, |
266 |
|
MACROBLOCK * const pMBs, |
267 |
|
const MBParam * const pParam, |
268 |
|
const FRAMEINFO * const current, |
269 |
|
const FRAMEINFO * const reference, |
270 |
|
const IMAGE * const pCurr, |
271 |
|
const IMAGE * const pRef, |
272 |
|
const IMAGE * const pRefH, |
273 |
|
const IMAGE * const pRefV, |
274 |
|
const IMAGE * const pRefHV); |
275 |
|
|
276 |
|
|
277 |
|
|
278 |
|
|
279 |
|
int |
280 |
|
globalSAD(const WARPPOINTS *const wp, |
281 |
|
const MBParam * const pParam, |
282 |
|
const MACROBLOCK * const pMBs, |
283 |
|
const FRAMEINFO * const current, |
284 |
|
const IMAGE * const pRef, |
285 |
|
const IMAGE * const pCurr, |
286 |
|
uint8_t *const GMCblock); |
287 |
|
|
288 |
#define iDiamondSize 2 |
#define iDiamondSize 2 |
289 |
|
|
290 |
static __inline uint32_t |
static __inline uint32_t |
348 |
static int |
static int |
349 |
CountMBBitsIntra(const SearchData * const Data); |
CountMBBitsIntra(const SearchData * const Data); |
350 |
|
|
351 |
|
static int |
352 |
|
CountMBBitsGMC(const SearchData * const Data, const IMAGE * const vGMC, const int x, const int y); |
353 |
|
|
354 |
int CodeCoeffIntra_CalcBits(const int16_t qcoeff[64], const uint16_t * zigzag); |
int CodeCoeffIntra_CalcBits(const int16_t qcoeff[64], const uint16_t * zigzag); |
355 |
int CodeCoeffInter_CalcBits(const int16_t qcoeff[64], const uint16_t * zigzag); |
int CodeCoeffInter_CalcBits(const int16_t qcoeff[64], const uint16_t * zigzag); |
356 |
|
|
357 |
#define LAMBDA ( (int)(1.0*BITS_MULT) ) |
/* one over lambda for R-D mode decision and motion search */ |
358 |
|
#define LAMBDA ( (int)(BITS_MULT/1.0) ) |
359 |
|
|
360 |
static __inline unsigned int |
static __inline unsigned int |
361 |
Block_CalcBits( int16_t * const coeff, |
Block_CalcBits( int16_t * const coeff, |
362 |
int16_t * const data, |
int16_t * const data, |
363 |
|
int16_t * const dqcoeff, |
364 |
const uint32_t quant, const int quant_type, |
const uint32_t quant, const int quant_type, |
365 |
uint32_t * cbp, |
uint32_t * cbp, |
366 |
const int block) |
const int block) |
367 |
{ |
{ |
368 |
int sum; |
int sum; |
369 |
int bits; |
int bits; |
|
const int lambda = LAMBDA*quant*quant; |
|
370 |
int distortion = 0; |
int distortion = 0; |
371 |
int i; |
int i; |
372 |
|
|
373 |
fdct(data); |
fdct(data); |
374 |
|
|
375 |
if (quant_type == 0) sum = quant_inter(coeff, data, quant); |
if (quant_type) sum = quant_inter(coeff, data, quant); |
376 |
else sum = quant4_inter(coeff, data, quant); |
else sum = quant4_inter(coeff, data, quant); |
377 |
|
|
378 |
if (sum > 0) { |
if (sum > 0) { |
379 |
*cbp |= 1 << (5 - block); |
*cbp |= 1 << (5 - block); |
380 |
bits = BITS_MULT * CodeCoeffInter_CalcBits(coeff, scan_tables[0]); |
bits = BITS_MULT * CodeCoeffInter_CalcBits(coeff, scan_tables[0]); |
|
} else bits = 0; |
|
381 |
|
|
382 |
if (quant_type == 0) dequant_inter(coeff, coeff, quant); |
if (quant_type) dequant_inter(dqcoeff, coeff, quant); |
383 |
else dequant4_inter(coeff, coeff, quant); |
else dequant4_inter(dqcoeff, coeff, quant); |
384 |
|
|
385 |
for (i = 0; i < 64; i++) { |
for (i = 0; i < 64; i++) |
386 |
distortion += (data[i] - coeff[i])*(data[i] - coeff[i]); |
distortion += (data[i] - dqcoeff[i])*(data[i] - dqcoeff[i]); |
387 |
|
|
388 |
|
} else { |
389 |
|
bits = 0; |
390 |
|
for (i = 0; i < 64; i++) |
391 |
|
distortion += data[i]*data[i]; |
392 |
} |
} |
|
bits += (BITS_MULT*BITS_MULT*distortion)/lambda; |
|
393 |
|
|
394 |
|
|
395 |
|
bits += (LAMBDA*distortion)/(quant*quant); |
396 |
|
|
397 |
return bits; |
return bits; |
398 |
} |
} |
399 |
|
|
400 |
static __inline unsigned int |
static __inline unsigned int |
401 |
Block_CalcBitsIntra(int16_t * const coeff, |
Block_CalcBitsIntra(int16_t * const coeff, |
402 |
int16_t * const data, |
int16_t * const data, |
403 |
|
int16_t * const dqcoeff, |
404 |
const uint32_t quant, const int quant_type, |
const uint32_t quant, const int quant_type, |
405 |
uint32_t * cbp, |
uint32_t * cbp, |
406 |
const int block, |
const int block, |
407 |
int * dcpred) |
int * dcpred) |
408 |
{ |
{ |
409 |
int bits, i; |
int bits, i; |
|
const int lambda = LAMBDA*quant*quant; |
|
410 |
int distortion = 0; |
int distortion = 0; |
411 |
uint32_t iDcScaler = get_dc_scaler(quant, block > 3); |
uint32_t iDcScaler = get_dc_scaler(quant, block < 4); |
412 |
int b_dc; |
int b_dc; |
413 |
|
|
414 |
fdct(data); |
fdct(data); |
415 |
data[0] -= 1024; |
data[0] -= 1024; |
416 |
|
|
417 |
if (quant_type == 0) quant_intra(coeff, data, quant, iDcScaler); |
if (quant_type) quant_intra(coeff, data, quant, iDcScaler); |
418 |
else quant4_intra(coeff, data, quant, iDcScaler); |
else quant4_intra(coeff, data, quant, iDcScaler); |
419 |
|
|
420 |
b_dc = coeff[0]; |
b_dc = coeff[0]; |
423 |
*dcpred = b_dc; |
*dcpred = b_dc; |
424 |
} |
} |
425 |
|
|
|
*cbp |= 1 << (5 - block); |
|
426 |
bits = BITS_MULT*CodeCoeffIntra_CalcBits(coeff, scan_tables[0]); |
bits = BITS_MULT*CodeCoeffIntra_CalcBits(coeff, scan_tables[0]); |
|
bits += BITS_MULT*dcy_tab[coeff[0] + 255].len; |
|
427 |
if (bits != 0) *cbp |= 1 << (5 - block); |
if (bits != 0) *cbp |= 1 << (5 - block); |
428 |
|
|
429 |
|
if (block < 4) bits += BITS_MULT*dcy_tab[coeff[0] + 255].len; |
430 |
|
else bits += BITS_MULT*dcc_tab[coeff[0] + 255].len; |
431 |
|
|
432 |
coeff[0] = b_dc; |
coeff[0] = b_dc; |
433 |
if (quant_type == 0) dequant_intra(coeff, coeff, quant, iDcScaler); |
if (quant_type) dequant_intra(dqcoeff, coeff, quant, iDcScaler); |
434 |
else dequant4_intra(coeff, coeff, quant, iDcScaler); |
else dequant4_intra(dqcoeff, coeff, quant, iDcScaler); |
435 |
|
|
436 |
for (i = 0; i < 64; i++) { |
for (i = 0; i < 64; i++) { |
437 |
distortion += (data[i] - coeff[i])*(data[i] - coeff[i]); |
distortion += (data[i] - dqcoeff[i])*(data[i] - dqcoeff[i]); |
438 |
} |
} |
439 |
|
|
440 |
bits += (BITS_MULT*BITS_MULT*distortion)/lambda; |
bits += (LAMBDA*distortion)/(quant*quant); |
441 |
|
|
442 |
return bits; |
return bits; |
443 |
} |
} |