--- trunk/xvidcore/src/motion/motion_est.h 2003/02/21 14:44:57 884 +++ branches/dev-api-4/xvidcore/src/motion/motion_est.h 2003/08/03 10:10:54 1108 @@ -1,34 +1,29 @@ -/************************************************************************** +/***************************************************************************** * * XVID MPEG-4 VIDEO CODEC - * - Motion estimation header - + * - Motion Estimation related header - * - * This program is an implementation of a part of one or more MPEG-4 - * 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. + * Copyright(C) 2002 Christoph Lampert + * 2002 Michael Militzer + * 2002-2003 Radoslaw Czyz * - * This program is free software; you can redistribute it and/or modify + * This program is free software ; you can redistribute it and/or modify * it under the terms of the GNU General Public License as published by - * the Free Software Foundation; either version 2 of the License, or + * the Free Software Foundation ; either version 2 of the License, or * (at your option) any later version. * * This program is distributed in the hope that it will be useful, - * but WITHOUT ANY WARRANTY; without even the implied warranty of + * but WITHOUT ANY WARRANTY ; without even the implied warranty of * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the * GNU General Public License for more details. * * You should have received a copy of the GNU General Public License - * along with this program; if not, write to the Free Software + * along with this program ; if not, write to the Free Software * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA * - * $Id: motion_est.h,v 1.3 2003-02-21 14:40:11 syskin Exp $ + * $Id: motion_est.h,v 1.3.2.15 2003-08-03 10:10:54 syskin Exp $ * - ***************************************************************************/ + ****************************************************************************/ #ifndef _MOTION_EST_H_ #define _MOTION_EST_H_ @@ -39,7 +34,7 @@ /* hard coded motion search parameters for motion_est and smp_motion_est */ -// very large value +/* very large value */ #define MV_MAX_ERROR (4096 * 256) /* INTER bias for INTER/INTRA decision; mpeg4 spec suggests 2*nb */ @@ -50,6 +45,8 @@ #define NEIGH_TEND_8X8 40.0 #define NEIGH_8X8_BIAS 30 +#define BITS_MULT 16 + /* Parameters which control inter/inter4v decision */ #define IMV16X16 2 @@ -89,13 +86,15 @@ (int)(21.2656*NEIGH_TEND_8X8 + 0.5), (int)(24.8580*NEIGH_TEND_8X8 + 0.5), (int)(29.6436*NEIGH_TEND_8X8 + 0.5), (int)(36.4949*NEIGH_TEND_8X8 + 0.5) }; -// mv.length table -static const uint32_t mvtab[33] = { - 1, 2, 3, 4, 6, 7, 7, 7, - 9, 9, 9, 10, 10, 10, 10, 10, - 10, 10, 10, 10, 10, 10, 10, 10, - 10, 11, 11, 11, 11, 11, 11, 12, 12 -}; +/* mv.length table */ +static const int mvtab[64] = { + 1, 2, 3, 4, 6, 7, 7, 7, + 9, 9, 9, 10, 10, 10, 10, 10, + 10, 10, 10, 10, 10, 10, 10, 10, + 10, 11, 11, 11, 11, 11, 11, 12, + 12, 12, 12, 12, 12, 12, 12, 12, + 12, 12, 12, 12, 12, 12, 12, 12, + 12, 12, 12, 12, 12, 12, 12, 12, 12 }; static const int DQtab[4] = { -1, -2, 1, 2 @@ -105,19 +104,14 @@ typedef struct { -// general fields + /* general fields */ int max_dx, min_dx, max_dy, min_dy; uint32_t rounding; VECTOR predMV; VECTOR * currentMV; VECTOR * currentQMV; int32_t * iMinSAD; - const uint8_t * Ref; - const uint8_t * RefH; - const uint8_t * RefV; - const uint8_t * RefHV; - const uint8_t * RefCU; - const uint8_t * RefCV; + const uint8_t * RefP[6]; /* N, V, H, HV, cU, cV */ const uint8_t * CurU; const uint8_t * CurV; uint8_t * RefQ; @@ -130,21 +124,23 @@ int qpel, qpel_precision; int chroma; 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; + /* fields for interpolate and direct modes */ + const uint8_t * b_RefP[6]; /* N, V, H, HV, cU, cV */ VECTOR bpredMV; uint32_t bFcode; -// fields for direct mode + + /* fields for direct mode */ VECTOR directmvF[4]; VECTOR directmvB[4]; const VECTOR * referencemv; + /* BITS/R-D stuff */ + int16_t * dctSpace; + uint32_t iQuant; + uint32_t quant_type; + int * cbp; + } SearchData; @@ -213,6 +209,7 @@ const IMAGE * const pRefH, const IMAGE * const pRefV, const IMAGE * const pRefHV, + const IMAGE * const pGMC, const uint32_t iLimit); static void @@ -224,18 +221,38 @@ const int x, const int y, const uint32_t MotionFlags, - const uint32_t GlobalFlags, - const uint32_t iQuant, + const uint32_t VopFlags, + const uint32_t VolFlags, SearchData * const Data, const MBParam * const pParam, const MACROBLOCK * const pMBs, const MACROBLOCK * const prevMBs, - int inter4v, MACROBLOCK * const pMB); +static __inline void +GMEanalyzeMB (const uint8_t * const pCur, + const uint8_t * const pRef, + const uint8_t * const pRefH, + const uint8_t * const pRefV, + const uint8_t * const pRefHV, + const int x, + const int y, + const MBParam * const pParam, + MACROBLOCK * const pMBs, + SearchData * const Data); + +void +GMEanalysis(const MBParam * const pParam, + const FRAMEINFO * const current, + const FRAMEINFO * const reference, + const IMAGE * const pRefH, + const IMAGE * const pRefV, + const IMAGE * const pRefHV); -static WARPPOINTS -GlobalMotionEst(const MACROBLOCK * const pMBs, + + +WARPPOINTS +GlobalMotionEst(MACROBLOCK * const pMBs, const MBParam * const pParam, const FRAMEINFO * const current, const FRAMEINFO * const reference, @@ -243,69 +260,180 @@ const IMAGE * const pRefV, const IMAGE * const pRefHV ); + +int +GlobalMotionEstRefine(WARPPOINTS *const startwp, + MACROBLOCK * const pMBs, + const MBParam * const pParam, + const FRAMEINFO * const current, + const FRAMEINFO * const reference, + const IMAGE * const pCurr, + const IMAGE * const pRef, + const IMAGE * const pRefH, + const IMAGE * const pRefV, + const IMAGE * const pRefHV); + + + + +int +globalSAD(const WARPPOINTS *const wp, + const MBParam * const pParam, + const MACROBLOCK * const pMBs, + const FRAMEINFO * const current, + const IMAGE * const pRef, + const IMAGE * const pCurr, + uint8_t *const GMCblock); + #define iDiamondSize 2 static __inline uint32_t -MakeGoodMotionFlags(const uint32_t MotionFlags, const uint32_t GlobalFlags) +MakeGoodMotionFlags(const uint32_t MotionFlags, const uint32_t VopFlags, const uint32_t VolFlags) { uint32_t Flags = MotionFlags; - if (!(GlobalFlags & XVID_MODEDECISION_BITS)) - Flags &= ~(QUARTERPELREFINE16_BITS+QUARTERPELREFINE8_BITS+HALFPELREFINE16_BITS+HALFPELREFINE8_BITS+EXTSEARCH_BITS); + if (!(VopFlags & XVID_VOP_MODEDECISION_RD)) + Flags &= ~(XVID_ME_QUARTERPELREFINE16_RD+XVID_ME_QUARTERPELREFINE8_RD+XVID_ME_HALFPELREFINE16_RD+XVID_ME_HALFPELREFINE8_RD+XVID_ME_EXTSEARCH_RD); - if (Flags & EXTSEARCH_BITS) - Flags |= HALFPELREFINE16_BITS; + if (Flags & XVID_ME_EXTSEARCH_RD) + Flags |= XVID_ME_HALFPELREFINE16_RD; - if (Flags & EXTSEARCH_BITS && MotionFlags & PMV_EXTSEARCH8) - Flags |= HALFPELREFINE8_BITS; + if (Flags & XVID_ME_EXTSEARCH_RD && MotionFlags & XVID_ME_EXTSEARCH8) + Flags |= XVID_ME_HALFPELREFINE8_RD; - if (Flags & HALFPELREFINE16_BITS) - Flags |= QUARTERPELREFINE16_BITS; + if (Flags & XVID_ME_HALFPELREFINE16_RD) + Flags |= XVID_ME_QUARTERPELREFINE16_RD; - if (Flags & HALFPELREFINE8_BITS) { - Flags |= QUARTERPELREFINE8_BITS; - Flags &= ~PMV_HALFPELREFINE8; + if (Flags & XVID_ME_HALFPELREFINE8_RD) { + Flags |= XVID_ME_QUARTERPELREFINE8_RD; + Flags &= ~XVID_ME_HALFPELREFINE8; } - if (Flags & QUARTERPELREFINE8_BITS) - Flags &= ~PMV_QUARTERPELREFINE8; + if (Flags & XVID_ME_QUARTERPELREFINE8_RD) + Flags &= ~XVID_ME_QUARTERPELREFINE8; - if (!(GlobalFlags & XVID_QUARTERPEL)) - Flags &= ~(PMV_QUARTERPELREFINE16+PMV_QUARTERPELREFINE8+QUARTERPELREFINE16_BITS+QUARTERPELREFINE8_BITS); + if (!(VolFlags & XVID_VOL_QUARTERPEL)) + Flags &= ~(XVID_ME_QUARTERPELREFINE16+XVID_ME_QUARTERPELREFINE8+XVID_ME_QUARTERPELREFINE16_RD+XVID_ME_QUARTERPELREFINE8_RD); - if (!(GlobalFlags & XVID_HALFPEL)) - Flags &= ~(PMV_EXTSEARCH16+PMV_HALFPELREFINE16+PMV_HALFPELREFINE8+HALFPELREFINE16_BITS+HALFPELREFINE8_BITS); + if (!(VopFlags & XVID_VOP_HALFPEL)) + Flags &= ~(XVID_ME_EXTSEARCH16+XVID_ME_HALFPELREFINE16+XVID_ME_HALFPELREFINE8+XVID_ME_HALFPELREFINE16_RD+XVID_ME_HALFPELREFINE8_RD); - if (GlobalFlags & (XVID_GREYSCALE + XVID_REDUCED)) - Flags &= ~(PMV_CHROMA16 + PMV_CHROMA8); + if ((VopFlags & XVID_VOP_GREYSCALE) || (VopFlags & XVID_VOP_REDUCED)) + Flags &= ~(XVID_ME_CHROMA_PVOP + XVID_ME_CHROMA_BVOP); return Flags; } -/* BITS mode decision and search */ +/* RD mode decision and search */ #include "../bitstream/zigzag.h" #include "../quant/quant_mpeg4.h" #include "../quant/quant_h263.h" #include "../bitstream/vlc_codes.h" +#include "../dct/fdct.h" static int -CountMBBitsInter(SearchData * const Data, - const MACROBLOCK * const pMBs, const int x, const int y, - const MBParam * const pParam, - const uint32_t MotionFlags); +findRDinter(SearchData * const Data, + const MACROBLOCK * const pMBs, const int x, const int y, + const MBParam * const pParam, + const uint32_t MotionFlags); + +static int +findRDinter4v(const SearchData * const Data, + MACROBLOCK * const pMB, const MACROBLOCK * const pMBs, + const int x, const int y, + const MBParam * const pParam, const uint32_t MotionFlags, + const VECTOR * const backup); static int -CountMBBitsInter4v(const SearchData * const Data, - MACROBLOCK * const pMB, const MACROBLOCK * const pMBs, - const int x, const int y, - const MBParam * const pParam, const uint32_t MotionFlags, - const VECTOR * const backup); +findRDintra(const SearchData * const Data); static int -CountMBBitsIntra(const SearchData * const Data); +findRDgmc(const SearchData * const Data, const IMAGE * const vGMC, const int x, const int y); int CodeCoeffIntra_CalcBits(const int16_t qcoeff[64], const uint16_t * zigzag); int CodeCoeffInter_CalcBits(const int16_t qcoeff[64], const uint16_t * zigzag); +/* one over lambda for R-D mode decision and motion search */ +#define LAMBDA ( (int)(BITS_MULT/1.0) ) + +static __inline unsigned int +Block_CalcBits( int16_t * const coeff, + int16_t * const data, + int16_t * const dqcoeff, + const uint32_t quant, const int quant_type, + uint32_t * cbp, + const int block) +{ + int sum; + int bits; + int distortion = 0; + int i; + + fdct(data); + + if (quant_type) sum = quant_inter(coeff, data, quant); + else sum = quant4_inter(coeff, data, quant); + + if (sum > 0) { + *cbp |= 1 << (5 - block); + bits = BITS_MULT * CodeCoeffInter_CalcBits(coeff, scan_tables[0]); + + if (quant_type) dequant_inter(dqcoeff, coeff, quant); + else dequant4_inter(dqcoeff, coeff, quant); + + for (i = 0; i < 64; i++) + distortion += (data[i] - dqcoeff[i])*(data[i] - dqcoeff[i]); + + } else { + bits = 0; + for (i = 0; i < 64; i++) + distortion += data[i]*data[i]; + } + + return bits + (LAMBDA*distortion)/(quant*quant); +} + +static __inline unsigned int +Block_CalcBitsIntra(int16_t * const coeff, + int16_t * const data, + int16_t * const dqcoeff, + const uint32_t quant, const int quant_type, + uint32_t * cbp, + const int block, + int * dcpred) +{ + int bits, i; + int distortion = 0; + uint32_t iDcScaler = get_dc_scaler(quant, block < 4); + int b_dc; + + fdct(data); + data[0] -= 1024; + + if (quant_type) quant_intra(coeff, data, quant, iDcScaler); + else quant4_intra(coeff, data, quant, iDcScaler); + + b_dc = coeff[0]; + if (block < 4) { + coeff[0] -= *dcpred; + *dcpred = b_dc; + } + + bits = BITS_MULT*CodeCoeffIntra_CalcBits(coeff, scan_tables[0]); + if (bits != 0) *cbp |= 1 << (5 - block); + + if (block < 4) bits += BITS_MULT*dcy_tab[coeff[0] + 255].len; + else bits += BITS_MULT*dcc_tab[coeff[0] + 255].len; + + coeff[0] = b_dc; + if (quant_type) dequant_intra(dqcoeff, coeff, quant, iDcScaler); + else dequant4_intra(dqcoeff, coeff, quant, iDcScaler); + + for (i = 0; i < 64; i++) + distortion += (data[i] - dqcoeff[i])*(data[i] - dqcoeff[i]); + + return bits + (LAMBDA*distortion)/(quant*quant); +} + #endif /* _MOTION_EST_H_ */