--- estimation_rd_based_bvop.c 2004/12/17 11:57:59 1.5 +++ estimation_rd_based_bvop.c 2004/12/19 14:05:55 1.8 @@ -1,3 +1,27 @@ +/***************************************************************************** + * + * XVID MPEG-4 VIDEO CODEC + * - Rate-Distortion Based Motion Estimation for B- VOPs - + * + * Copyright(C) 2004 Radoslaw Czyz + * + * 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 + * (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 + * 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 + * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA + * + * $Id: estimation_rd_based_bvop.c,v 1.8 2004/12/19 14:05:55 syskin Exp $ + * + ****************************************************************************/ #include #include @@ -31,7 +55,9 @@ const int block, const uint16_t * scan_table, const unsigned int lambda, - const uint16_t * mpeg_quant_matrices) + const uint16_t * mpeg_quant_matrices, + const unsigned int quant_sq, + int * const cbpcost) { int sum; int bits; @@ -45,6 +71,8 @@ if ((sum >= 3) || (coeff[1] != 0) || (coeff[8] != 0) || (coeff[0] != 0)) { *cbp |= 1 << (5 - block); bits = BITS_MULT * CodeCoeffInter_CalcBits(coeff, scan_table); + bits += *cbpcost; + *cbpcost = 0; /* don't add cbp cost second time */ if (quant_type) dequant_h263_inter(dqcoeff, coeff, quant, mpeg_quant_matrices); else dequant_mpeg_inter(dqcoeff, coeff, quant, mpeg_quant_matrices); @@ -66,7 +94,7 @@ distortion = sse8_16bit(data, zero_block, 8*sizeof(int16_t)); } - return bits + (lambda*distortion)/(quant*quant); + return bits + (lambda*distortion)/quant_sq; } @@ -79,7 +107,9 @@ const int block, const uint16_t * scan_table, const unsigned int lambda, - const uint16_t * mpeg_quant_matrices) + const uint16_t * mpeg_quant_matrices, + const unsigned int quant_sq, + int * const cbpcost) { int sum; int bits; @@ -93,6 +123,8 @@ if ((sum >= 3) || (coeff[1] != 0) || (coeff[8] != 0) || (coeff[0] > 0) || (coeff[0] < -1)) { *cbp |= 1 << (5 - block); bits = BITS_MULT * CodeCoeffInter_CalcBits(coeff, scan_table); + bits += *cbpcost; + *cbpcost = 0; if (quant_type) dequant_h263_inter(dqcoeff, coeff, quant, mpeg_quant_matrices); else dequant_mpeg_inter(dqcoeff, coeff, quant, mpeg_quant_matrices); @@ -114,26 +146,7 @@ distortion = sse8_16bit(data, zero_block, 8*sizeof(int16_t)); } - return bits + (lambda*distortion)/(quant*quant); -} - - -static void -transfer_8to16sub2ro(int16_t * const dct, - const uint8_t * const cur, - const uint8_t * ref1, - const uint8_t * ref2, - const uint32_t stride) -{ - uint32_t i, j; - - for (j = 0; j < 8; j++) { - for (i = 0; i < 8; i++) { - uint8_t c = cur[j * stride + i]; - int r = (ref1[j * stride + i] + ref2[j * stride + i] + 1) / 2; - dct[j * 8 + i] = (int16_t) c - (int16_t) r; - } - } + return bits + (lambda*distortion)/quant_sq; } static void @@ -141,11 +154,12 @@ { int16_t *in = data->dctSpace, *coeff = data->dctSpace + 64; - int32_t rd = 3*BITS_MULT; /* note to self: 3 bits minimum, but maybe 4 if it's forward mode */ + int32_t rd = (3+2)*BITS_MULT; /* 3 bits for mode + 2 for vector (minimum) */ VECTOR * current; const uint8_t * ptr; int i, xc, yc; unsigned cbp = 0; + int cbpcost = 7*BITS_MULT; /* how much to add if cbp turns out to be non-zero */ if ( (x > data->max_dx) || (x < data->min_dx) || (y > data->max_dy) || (y < data->min_dy) ) return; @@ -160,12 +174,14 @@ xc = x/2; yc = y/2; } - rd += BITS_MULT*d_mv_bits(x, y, data->predMV, data->iFcode, data->qpel^data->qpel_precision); + rd += BITS_MULT*(d_mv_bits(x, y, data->predMV, data->iFcode, data->qpel^data->qpel_precision)-2); for(i = 0; i < 4; i++) { int s = 8*((i&1) + (i>>1)*data->iEdgedWidth); transfer_8to16subro(in, data->Cur + s, ptr + s, data->iEdgedWidth); - rd += Block_CalcBits_BVOP(coeff, in, data->dctSpace + 128, data->iQuant, data->quant_type, &cbp, i, data->scan_table, data->lambda[i], data->mpeg_quant_matrices); + rd += Block_CalcBits_BVOP(coeff, in, data->dctSpace + 128, data->iQuant, data->quant_type, + &cbp, i, data->scan_table, data->lambda[i], data->mpeg_quant_matrices, + data->quant_sq, &cbpcost); if (rd >= data->iMinSAD[0]) return; } @@ -176,15 +192,17 @@ /* chroma U */ ptr = interpolate8x8_switch2(data->RefQ, data->RefP[4], 0, 0, xc, yc, data->iEdgedWidth/2, data->rounding); transfer_8to16subro(in, data->CurU, ptr, data->iEdgedWidth/2); - rd += Block_CalcBits_BVOP(coeff, in, data->dctSpace + 128, data->iQuant, data->quant_type, &cbp, 4, data->scan_table, data->lambda[4], data->mpeg_quant_matrices); + rd += Block_CalcBits_BVOP(coeff, in, data->dctSpace + 128, data->iQuant, data->quant_type, + &cbp, 4, data->scan_table, data->lambda[4], data->mpeg_quant_matrices, + data->quant_sq, &cbpcost); if (rd >= data->iMinSAD[0]) return; /* chroma V */ ptr = interpolate8x8_switch2(data->RefQ, data->RefP[5], 0, 0, xc, yc, data->iEdgedWidth/2, data->rounding); transfer_8to16subro(in, data->CurV, ptr, data->iEdgedWidth/2); - rd += Block_CalcBits_BVOP(coeff, in, data->dctSpace + 128, data->iQuant, data->quant_type, &cbp, 5, data->scan_table, data->lambda[5], data->mpeg_quant_matrices); - - if (cbp) rd += BITS_MULT * 7; + rd += Block_CalcBits_BVOP(coeff, in, data->dctSpace + 128, data->iQuant, data->quant_type, + &cbp, 5, data->scan_table, data->lambda[5], data->mpeg_quant_matrices, + data->quant_sq, &cbpcost); if (rd < data->iMinSAD[0]) { data->iMinSAD[0] = rd; @@ -194,7 +212,6 @@ } } - static void CheckCandidateRDDirect(const int x, const int y, SearchData * const data, const unsigned int Direction) { @@ -204,6 +221,7 @@ unsigned int cbp = 0; unsigned int k; VECTOR mvs, b_mvs; + int cbpcost = 6*BITS_MULT; /* how much to add if cbp turns out to be non-zero */ const uint8_t *ReferenceF, *ReferenceB; @@ -241,7 +259,9 @@ } transfer_8to16sub2ro(in, data->Cur + s, ReferenceF, ReferenceB, data->iEdgedWidth); - rd += Block_CalcBits_BVOP_direct(coeff, in, data->dctSpace + 128, data->iQuant, data->quant_type, &cbp, k, data->scan_table, data->lambda[k], data->mpeg_quant_matrices); + rd += Block_CalcBits_BVOP_direct(coeff, in, data->dctSpace + 128, data->iQuant, data->quant_type, + &cbp, k, data->scan_table, data->lambda[k], data->mpeg_quant_matrices, + data->quant_sq, &cbpcost); if (rd > *(data->iMinSAD)) return; } @@ -255,17 +275,19 @@ ReferenceF = interpolate8x8_switch2(data->RefQ, data->RefP[4], 0, 0, xcf, ycf, data->iEdgedWidth/2, data->rounding); ReferenceB = interpolate8x8_switch2(data->RefQ + 16, data->b_RefP[4], 0, 0, xcb, ycb, data->iEdgedWidth/2, data->rounding); transfer_8to16sub2ro(in, data->CurU, ReferenceF, ReferenceB, data->iEdgedWidth/2); - rd += Block_CalcBits_BVOP_direct(coeff, in, data->dctSpace + 128, data->iQuant, data->quant_type, &cbp, 4, data->scan_table, data->lambda[4], data->mpeg_quant_matrices); + rd += Block_CalcBits_BVOP_direct(coeff, in, data->dctSpace + 128, data->iQuant, data->quant_type, + &cbp, 4, data->scan_table, data->lambda[4], data->mpeg_quant_matrices, + data->quant_sq, &cbpcost); if (rd >= data->iMinSAD[0]) return; /* chroma V */ ReferenceF = interpolate8x8_switch2(data->RefQ, data->RefP[5], 0, 0, xcf, ycf, data->iEdgedWidth/2, data->rounding); ReferenceB = interpolate8x8_switch2(data->RefQ + 16, data->b_RefP[5], 0, 0, xcb, ycb, data->iEdgedWidth/2, data->rounding); transfer_8to16sub2ro(in, data->CurV, ReferenceF, ReferenceB, data->iEdgedWidth/2); - rd += Block_CalcBits_BVOP_direct(coeff, in, data->dctSpace + 128, data->iQuant, data->quant_type, &cbp, 5, data->scan_table, data->lambda[5], data->mpeg_quant_matrices); + rd += Block_CalcBits_BVOP_direct(coeff, in, data->dctSpace + 128, data->iQuant, data->quant_type, + &cbp, 5, data->scan_table, data->lambda[5], data->mpeg_quant_matrices, + data->quant_sq, &cbpcost); - if (cbp) - rd += BITS_MULT * 6; if (cbp || x != 0 || y != 0) rd += BITS_MULT * d_mv_bits(x, y, zeroMV, 1, 0); @@ -277,9 +299,6 @@ } } - - - static void CheckCandidateRDInt(const int x, const int y, SearchData * const data, const unsigned int Direction) { @@ -288,6 +307,7 @@ int16_t *in = data->dctSpace, *coeff = data->dctSpace + 64; unsigned int cbp = 0; unsigned int i; + int cbpcost = 7*BITS_MULT; /* how much to add if cbp turns out to be non-zero */ const uint8_t *ReferenceF, *ReferenceB; VECTOR *current; @@ -325,14 +345,15 @@ rd += BITS_MULT * (d_mv_bits(xf, yf, data->predMV, data->iFcode, data->qpel^data->qpel_precision) + d_mv_bits(xb, yb, data->bpredMV, data->iFcode, data->qpel^data->qpel_precision)); - for(i = 0; i < 4; i++) { int s = 8*((i&1) + (i>>1)*data->iEdgedWidth); if (rd >= *data->iMinSAD) return; transfer_8to16sub2ro(in, data->Cur + s, ReferenceF + s, ReferenceB + s, data->iEdgedWidth); - rd += Block_CalcBits_BVOP(coeff, in, data->dctSpace + 128, data->iQuant, data->quant_type, &cbp, i, data->scan_table, data->lambda[i], data->mpeg_quant_matrices); + rd += Block_CalcBits_BVOP(coeff, in, data->dctSpace + 128, data->iQuant, data->quant_type, &cbp, + i, data->scan_table, data->lambda[i], data->mpeg_quant_matrices, + data->quant_sq, &cbpcost); } - + /* chroma */ xcf = (xcf >> 1) + roundtab_79[xcf & 0x3]; ycf = (ycf >> 1) + roundtab_79[ycf & 0x3]; @@ -343,7 +364,9 @@ ReferenceF = interpolate8x8_switch2(data->RefQ, data->RefP[4], 0, 0, xcf, ycf, data->iEdgedWidth/2, data->rounding); ReferenceB = interpolate8x8_switch2(data->RefQ + 16, data->b_RefP[4], 0, 0, xcb, ycb, data->iEdgedWidth/2, data->rounding); transfer_8to16sub2ro(in, data->CurU, ReferenceF, ReferenceB, data->iEdgedWidth/2); - rd += Block_CalcBits_BVOP(coeff, in, data->dctSpace + 128, data->iQuant, data->quant_type, &cbp, 4, data->scan_table, data->lambda[4], data->mpeg_quant_matrices); + rd += Block_CalcBits_BVOP(coeff, in, data->dctSpace + 128, data->iQuant, data->quant_type, &cbp, + 4, data->scan_table, data->lambda[4], data->mpeg_quant_matrices, + data->quant_sq, &cbpcost); if (rd >= data->iMinSAD[0]) return; @@ -351,9 +374,9 @@ ReferenceF = interpolate8x8_switch2(data->RefQ, data->RefP[5], 0, 0, xcf, ycf, data->iEdgedWidth/2, data->rounding); ReferenceB = interpolate8x8_switch2(data->RefQ + 16, data->b_RefP[5], 0, 0, xcb, ycb, data->iEdgedWidth/2, data->rounding); transfer_8to16sub2ro(in, data->CurV, ReferenceF, ReferenceB, data->iEdgedWidth/2); - rd += Block_CalcBits_BVOP(coeff, in, data->dctSpace + 128, data->iQuant, data->quant_type, &cbp, 5, data->scan_table, data->lambda[5], data->mpeg_quant_matrices); - - if (cbp) rd += BITS_MULT * 7; + rd += Block_CalcBits_BVOP(coeff, in, data->dctSpace + 128, data->iQuant, data->quant_type, &cbp, + 5, data->scan_table, data->lambda[5], data->mpeg_quant_matrices, + data->quant_sq, &cbpcost); if (rd < *(data->iMinSAD)) { *data->iMinSAD = rd;