/***************************************************************************** * * XVID MPEG-4 VIDEO CODEC * - Macro Block coding functions - * * Copyright(C) 2002 Michael Militzer * * This file is part of XviD, a free MPEG-4 video encoder/decoder * * XviD 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 * * Under section 8 of the GNU General Public License, the copyright * holders of XVID explicitly forbid distribution in the following * countries: * * - Japan * - United States of America * * Linking XviD statically or dynamically with other modules is making a * combined work based on XviD. Thus, the terms and conditions of the * GNU General Public License cover the whole combination. * * As a special exception, the copyright holders of XviD give you * permission to link XviD with independent modules that communicate with * XviD solely through the VFW1.1 and DShow interfaces, regardless of the * license terms of these independent modules, and to copy and distribute * the resulting combined work under terms of your choice, provided that * every copy of the combined work is accompanied by a complete copy of * the source code of XviD (the version of XviD used to produce the * combined work), being distributed under the terms of the GNU General * Public License plus this exception. An independent module is a module * which is not derived from or based on XviD. * * Note that people who make modified versions of XviD are not obligated * to grant this special exception for their modified versions; it is * their choice whether to do so. The GNU General Public License gives * permission to release a modified version without this exception; this * exception also makes it possible to release a modified version which * carries forward this exception. * * $Id: mbcoding.c,v 1.41 2003-02-13 17:31:33 edgomez Exp $ * ****************************************************************************/ #include #include "../portab.h" #include "bitstream.h" #include "zigzag.h" #include "vlc_codes.h" #include "mbcoding.h" #include "../utils/mbfunctions.h" #define ABS(X) (((X)>0)?(X):-(X)) #define CLIP(X,A) (X > A) ? (A) : (X) /* #define BIGLUT */ #ifdef BIGLUT #define LEVELOFFSET 2048 #else #define LEVELOFFSET 32 #endif /***************************************************************************** * Local data ****************************************************************************/ static REVERSE_EVENT DCT3D[2][4096]; #ifdef BIGLUT static VLC coeff_VLC[2][2][4096][64]; static VLC *intra_table, *inter_table; #else static VLC coeff_VLC[2][2][64][64]; #endif /***************************************************************************** * Vector Length Coding Initialization ****************************************************************************/ void init_vlc_tables(void) { uint32_t i, j, intra, last, run, run_esc, level, level_esc, escape, escape_len, offset; #ifdef BIGLUT intra_table = (VLC*)coeff_VLC[1]; inter_table = (VLC*)coeff_VLC[0]; #endif for (intra = 0; intra < 2; intra++) for (i = 0; i < 4096; i++) DCT3D[intra][i].event.level = 0; for (intra = 0; intra < 2; intra++) for (last = 0; last < 2; last++) { for (run = 0; run < 63 + last; run++) for (level = 0; level < (uint32_t)(32 << intra); level++) { #ifdef BIGLUT offset = LEVELOFFSET; #else offset = !intra * LEVELOFFSET; #endif coeff_VLC[intra][last][level + offset][run].len = 128; } } for (intra = 0; intra < 2; intra++) for (i = 0; i < 102; i++) { #ifdef BIGLUT offset = LEVELOFFSET; #else offset = !intra * LEVELOFFSET; #endif for (j = 0; j < (uint32_t)(1 << (12 - coeff_tab[intra][i].vlc.len)); j++) { DCT3D[intra][(coeff_tab[intra][i].vlc.code << (12 - coeff_tab[intra][i].vlc.len)) | j].len = coeff_tab[intra][i].vlc.len; DCT3D[intra][(coeff_tab[intra][i].vlc.code << (12 - coeff_tab[intra][i].vlc.len)) | j].event = coeff_tab[intra][i].event; } coeff_VLC[intra][coeff_tab[intra][i].event.last][coeff_tab[intra][i].event.level + offset][coeff_tab[intra][i].event.run].code = coeff_tab[intra][i].vlc.code << 1; coeff_VLC[intra][coeff_tab[intra][i].event.last][coeff_tab[intra][i].event.level + offset][coeff_tab[intra][i].event.run].len = coeff_tab[intra][i].vlc.len + 1; #ifndef BIGLUT if (!intra) #endif { coeff_VLC[intra][coeff_tab[intra][i].event.last][offset - coeff_tab[intra][i].event.level][coeff_tab[intra][i].event.run].code = (coeff_tab[intra][i].vlc.code << 1) | 1; coeff_VLC[intra][coeff_tab[intra][i].event.last][offset - coeff_tab[intra][i].event.level][coeff_tab[intra][i].event.run].len = coeff_tab[intra][i].vlc.len + 1; } } for (intra = 0; intra < 2; intra++) for (last = 0; last < 2; last++) for (run = 0; run < 63 + last; run++) { for (level = 1; level < (uint32_t)(32 << intra); level++) { if (level <= max_level[intra][last][run] && run <= max_run[intra][last][level]) continue; #ifdef BIGLUT offset = LEVELOFFSET; #else offset = !intra * LEVELOFFSET; #endif level_esc = level - max_level[intra][last][run]; run_esc = run - 1 - max_run[intra][last][level]; if (level_esc <= max_level[intra][last][run] && run <= max_run[intra][last][level_esc]) { escape = ESCAPE1; escape_len = 7 + 1; run_esc = run; } else { if (run_esc <= max_run[intra][last][level] && level <= max_level[intra][last][run_esc]) { escape = ESCAPE2; escape_len = 7 + 2; level_esc = level; } else { #ifndef BIGLUT if (!intra) #endif { coeff_VLC[intra][last][level + offset][run].code = (ESCAPE3 << 21) | (last << 20) | (run << 14) | (1 << 13) | ((level & 0xfff) << 1) | 1; coeff_VLC[intra][last][level + offset][run].len = 30; coeff_VLC[intra][last][offset - level][run].code = (ESCAPE3 << 21) | (last << 20) | (run << 14) | (1 << 13) | ((-level & 0xfff) << 1) | 1; coeff_VLC[intra][last][offset - level][run].len = 30; } continue; } } coeff_VLC[intra][last][level + offset][run].code = (escape << coeff_VLC[intra][last][level_esc + offset][run_esc].len) | coeff_VLC[intra][last][level_esc + offset][run_esc].code; coeff_VLC[intra][last][level + offset][run].len = coeff_VLC[intra][last][level_esc + offset][run_esc].len + escape_len; #ifndef BIGLUT if (!intra) #endif { coeff_VLC[intra][last][offset - level][run].code = (escape << coeff_VLC[intra][last][level_esc + offset][run_esc].len) | coeff_VLC[intra][last][level_esc + offset][run_esc].code | 1; coeff_VLC[intra][last][offset - level][run].len = coeff_VLC[intra][last][level_esc + offset][run_esc].len + escape_len; } } #ifdef BIGLUT for (level = (uint32_t)(32 << intra); level < 2048; level++) { coeff_VLC[intra][last][level + offset][run].code = (ESCAPE3 << 21) | (last << 20) | (run << 14) | (1 << 13) | ((level & 0xfff) << 1) | 1; coeff_VLC[intra][last][level + offset][run].len = 30; coeff_VLC[intra][last][offset - level][run].code = (ESCAPE3 << 21) | (last << 20) | (run << 14) | (1 << 13) | ((-level & 0xfff) << 1) | 1; coeff_VLC[intra][last][offset - level][run].len = 30; } #else if (!intra) { coeff_VLC[intra][last][0][run].code = (ESCAPE3 << 21) | (last << 20) | (run << 14) | (1 << 13) | ((-32 & 0xfff) << 1) | 1; coeff_VLC[intra][last][0][run].len = 30; } #endif } /* Shut up the compiler -- gcc 3.3 pre release */ i = dc_threshold[0]; } /***************************************************************************** * Local inlined functions for MB coding ****************************************************************************/ static __inline void CodeVector(Bitstream * bs, int32_t value, int32_t f_code, Statistics * pStat) { const int scale_factor = 1 << (f_code - 1); const int cmp = scale_factor << 5; if (value < (-1 * cmp)) value += 64 * scale_factor; if (value > (cmp - 1)) value -= 64 * scale_factor; pStat->iMvSum += value * value; pStat->iMvCount++; if (value == 0) { BitstreamPutBits(bs, mb_motion_table[32].code, mb_motion_table[32].len); } else { uint16_t length, code, mv_res, sign; length = 16 << f_code; f_code--; sign = (value < 0); if (value >= length) value -= 2 * length; else if (value < -length) value += 2 * length; if (sign) value = -value; value--; mv_res = value & ((1 << f_code) - 1); code = ((value - mv_res) >> f_code) + 1; if (sign) code = -code; code += 32; BitstreamPutBits(bs, mb_motion_table[code].code, mb_motion_table[code].len); if (f_code) BitstreamPutBits(bs, mv_res, f_code); } } #ifdef BIGLUT static __inline void CodeCoeff(Bitstream * bs, const int16_t qcoeff[64], VLC * table, const uint16_t * zigzag, uint16_t intra) { uint32_t j, last; short v; VLC *vlc; j = intra; last = intra; while (j < 64 && (v = qcoeff[zigzag[j]]) == 0) j++; do { vlc = table + 64 * 2048 + (v << 6) + j - last; last = ++j; /* count zeroes */ while (j < 64 && (v = qcoeff[zigzag[j]]) == 0) j++; /* write code */ if (j != 64) { BitstreamPutBits(bs, vlc->code, vlc->len); } else { vlc += 64 * 4096; BitstreamPutBits(bs, vlc->code, vlc->len); break; } } while (1); } #else static __inline void CodeCoeffInter(Bitstream * bs, const int16_t qcoeff[64], const uint16_t * zigzag) { uint32_t i, run, prev_run, code, len; int32_t level, prev_level, level_shifted; i = 0; run = 0; while (!(level = qcoeff[zigzag[i++]])) run++; prev_level = level; prev_run = run; run = 0; while (i < 64) { if ((level = qcoeff[zigzag[i++]]) != 0) { level_shifted = prev_level + 32; if (!(level_shifted & -64)) { code = coeff_VLC[0][0][level_shifted][prev_run].code; len = coeff_VLC[0][0][level_shifted][prev_run].len; } else { code = (ESCAPE3 << 21) | (prev_run << 14) | (1 << 13) | ((prev_level & 0xfff) << 1) | 1; len = 30; } BitstreamPutBits(bs, code, len); prev_level = level; prev_run = run; run = 0; } else run++; } level_shifted = prev_level + 32; if (!(level_shifted & -64)) { code = coeff_VLC[0][1][level_shifted][prev_run].code; len = coeff_VLC[0][1][level_shifted][prev_run].len; } else { code = (ESCAPE3 << 21) | (1 << 20) | (prev_run << 14) | (1 << 13) | ((prev_level & 0xfff) << 1) | 1; len = 30; } BitstreamPutBits(bs, code, len); } static __inline void CodeCoeffIntra(Bitstream * bs, const int16_t qcoeff[64], const uint16_t * zigzag) { uint32_t i, abs_level, run, prev_run, code, len; int32_t level, prev_level; i = 1; run = 0; while (!(level = qcoeff[zigzag[i++]])) run++; prev_level = level; prev_run = run; run = 0; while (i < 64) { if ((level = qcoeff[zigzag[i++]]) != 0) { abs_level = ABS(prev_level); abs_level = abs_level < 64 ? abs_level : 0; code = coeff_VLC[1][0][abs_level][prev_run].code; len = coeff_VLC[1][0][abs_level][prev_run].len; if (len != 128) code |= (prev_level < 0); else { code = (ESCAPE3 << 21) | (prev_run << 14) | (1 << 13) | ((prev_level & 0xfff) << 1) | 1; len = 30; } BitstreamPutBits(bs, code, len); prev_level = level; prev_run = run; run = 0; } else run++; } abs_level = ABS(prev_level); abs_level = abs_level < 64 ? abs_level : 0; code = coeff_VLC[1][1][abs_level][prev_run].code; len = coeff_VLC[1][1][abs_level][prev_run].len; if (len != 128) code |= (prev_level < 0); else { code = (ESCAPE3 << 21) | (1 << 20) | (prev_run << 14) | (1 << 13) | ((prev_level & 0xfff) << 1) | 1; len = 30; } BitstreamPutBits(bs, code, len); } #endif /***************************************************************************** * Local functions ****************************************************************************/ static void CodeBlockIntra(const FRAMEINFO * frame, const MACROBLOCK * pMB, int16_t qcoeff[6 * 64], Bitstream * bs, Statistics * pStat) { uint32_t i, mcbpc, cbpy, bits; cbpy = pMB->cbp >> 2; /* write mcbpc */ if (frame->coding_type == I_VOP) { mcbpc = ((pMB->mode >> 1) & 3) | ((pMB->cbp & 3) << 2); BitstreamPutBits(bs, mcbpc_intra_tab[mcbpc].code, mcbpc_intra_tab[mcbpc].len); } else { mcbpc = (pMB->mode & 7) | ((pMB->cbp & 3) << 3); BitstreamPutBits(bs, mcbpc_inter_tab[mcbpc].code, mcbpc_inter_tab[mcbpc].len); } /* ac prediction flag */ if (pMB->acpred_directions[0]) BitstreamPutBits(bs, 1, 1); else BitstreamPutBits(bs, 0, 1); /* write cbpy */ BitstreamPutBits(bs, cbpy_tab[cbpy].code, cbpy_tab[cbpy].len); /* write dquant */ if (pMB->mode == MODE_INTRA_Q) BitstreamPutBits(bs, pMB->dquant, 2); /* write interlacing */ if (frame->global_flags & XVID_INTERLACING) { BitstreamPutBit(bs, pMB->field_dct); } /* code block coeffs */ for (i = 0; i < 6; i++) { if (i < 4) BitstreamPutBits(bs, dcy_tab[qcoeff[i * 64 + 0] + 255].code, dcy_tab[qcoeff[i * 64 + 0] + 255].len); else BitstreamPutBits(bs, dcc_tab[qcoeff[i * 64 + 0] + 255].code, dcc_tab[qcoeff[i * 64 + 0] + 255].len); if (pMB->cbp & (1 << (5 - i))) { bits = BitstreamPos(bs); #ifdef BIGLUT CodeCoeff(bs, &qcoeff[i * 64], intra_table, scan_tables[pMB->acpred_directions[i]], 1); #else CodeCoeffIntra(bs, &qcoeff[i * 64], scan_tables[pMB->acpred_directions[i]]); #endif bits = BitstreamPos(bs) - bits; pStat->iTextBits += bits; } } } static void CodeBlockInter(const FRAMEINFO * frame, const MACROBLOCK * pMB, int16_t qcoeff[6 * 64], Bitstream * bs, Statistics * pStat) { int32_t i; uint32_t bits, mcbpc, cbpy; mcbpc = (pMB->mode & 7) | ((pMB->cbp & 3) << 3); cbpy = 15 - (pMB->cbp >> 2); /* write mcbpc */ BitstreamPutBits(bs, mcbpc_inter_tab[mcbpc].code, mcbpc_inter_tab[mcbpc].len); /* write cbpy */ BitstreamPutBits(bs, cbpy_tab[cbpy].code, cbpy_tab[cbpy].len); /* write dquant */ if (pMB->mode == MODE_INTER_Q) BitstreamPutBits(bs, pMB->dquant, 2); /* interlacing */ if (frame->global_flags & XVID_INTERLACING) { if (pMB->cbp) { BitstreamPutBit(bs, pMB->field_dct); DPRINTF(DPRINTF_DEBUG, "codep: field_dct: %d", pMB->field_dct); } /* if inter block, write field ME flag */ if (pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q) { BitstreamPutBit(bs, pMB->field_pred); DPRINTF(DPRINTF_DEBUG, "codep: field_pred: %d", pMB->field_pred); /* write field prediction references */ if (pMB->field_pred) { BitstreamPutBit(bs, pMB->field_for_top); BitstreamPutBit(bs, pMB->field_for_bot); } } } /* code motion vector(s) */ for (i = 0; i < (pMB->mode == MODE_INTER4V ? 4 : 1); i++) { CodeVector(bs, pMB->pmvs[i].x, frame->fcode, pStat); CodeVector(bs, pMB->pmvs[i].y, frame->fcode, pStat); } bits = BitstreamPos(bs); /* code block coeffs */ for (i = 0; i < 6; i++) if (pMB->cbp & (1 << (5 - i))) #ifdef BIGLUT CodeCoeff(bs, &qcoeff[i * 64], inter_table, scan_tables[0], 0); #else CodeCoeffInter(bs, &qcoeff[i * 64], scan_tables[0]); #endif bits = BitstreamPos(bs) - bits; pStat->iTextBits += bits; } /***************************************************************************** * Macro Block bitstream encoding functions ****************************************************************************/ void MBCoding(const FRAMEINFO * frame, MACROBLOCK * pMB, int16_t qcoeff[6 * 64], Bitstream * bs, Statistics * pStat) { if (frame->coding_type == P_VOP) { BitstreamPutBit(bs, 0); /* coded */ } if (pMB->mode == MODE_INTRA || pMB->mode == MODE_INTRA_Q) CodeBlockIntra(frame, pMB, qcoeff, bs, pStat); else CodeBlockInter(frame, pMB, qcoeff, bs, pStat); } void MBSkip(Bitstream * bs) { BitstreamPutBit(bs, 1); /* not coded */ return; } /***************************************************************************** * decoding stuff starts here ****************************************************************************/ /* * For IVOP addbits == 0 * For PVOP addbits == fcode - 1 * For BVOP addbits == max(fcode,bcode) - 1 * returns true or false */ int check_resync_marker(Bitstream * bs, int addbits) { uint32_t nbits; uint32_t code; uint32_t nbitsresyncmarker = NUMBITS_VP_RESYNC_MARKER + addbits; nbits = BitstreamNumBitsToByteAlign(bs); code = BitstreamShowBits(bs, nbits); if (code == (((uint32_t)1 << (nbits - 1)) - 1)) { return BitstreamShowBitsFromByteAlign(bs, nbitsresyncmarker) == RESYNC_MARKER; } return 0; } int get_mcbpc_intra(Bitstream * bs) { uint32_t index; index = BitstreamShowBits(bs, 9); index >>= 3; BitstreamSkip(bs, mcbpc_intra_table[index].len); return mcbpc_intra_table[index].code; } int get_mcbpc_inter(Bitstream * bs) { uint32_t index; index = CLIP(BitstreamShowBits(bs, 9), 256); BitstreamSkip(bs, mcbpc_inter_table[index].len); return mcbpc_inter_table[index].code; } int get_cbpy(Bitstream * bs, int intra) { int cbpy; uint32_t index = BitstreamShowBits(bs, 6); BitstreamSkip(bs, cbpy_table[index].len); cbpy = cbpy_table[index].code; if (!intra) cbpy = 15 - cbpy; return cbpy; } int get_mv_data(Bitstream * bs) { uint32_t index; if (BitstreamGetBit(bs)) return 0; index = BitstreamShowBits(bs, 12); if (index >= 512) { index = (index >> 8) - 2; BitstreamSkip(bs, TMNMVtab0[index].len); return TMNMVtab0[index].code; } if (index >= 128) { index = (index >> 2) - 32; BitstreamSkip(bs, TMNMVtab1[index].len); return TMNMVtab1[index].code; } index -= 4; BitstreamSkip(bs, TMNMVtab2[index].len); return TMNMVtab2[index].code; } int get_mv(Bitstream * bs, int fcode) { int data; int res; int mv; int scale_fac = 1 << (fcode - 1); data = get_mv_data(bs); if (scale_fac == 1 || data == 0) return data; res = BitstreamGetBits(bs, fcode - 1); mv = ((ABS(data) - 1) * scale_fac) + res + 1; return data < 0 ? -mv : mv; } int get_dc_dif(Bitstream * bs, uint32_t dc_size) { int code = BitstreamGetBits(bs, dc_size); int msb = code >> (dc_size - 1); if (msb == 0) return (-1 * (code ^ ((1 << dc_size) - 1))); return code; } int get_dc_size_lum(Bitstream * bs) { int code, i; code = BitstreamShowBits(bs, 11); for (i = 11; i > 3; i--) { if (code == 1) { BitstreamSkip(bs, i); return i + 1; } code >>= 1; } BitstreamSkip(bs, dc_lum_tab[code].len); return dc_lum_tab[code].code; } int get_dc_size_chrom(Bitstream * bs) { uint32_t code, i; code = BitstreamShowBits(bs, 12); for (i = 12; i > 2; i--) { if (code == 1) { BitstreamSkip(bs, i); return i; } code >>= 1; } return 3 - BitstreamGetBits(bs, 2); } /***************************************************************************** * Local inlined function to "decode" written vlc codes ****************************************************************************/ static __inline int get_coeff(Bitstream * bs, int *run, int *last, int intra, int short_video_header) { uint32_t mode; int32_t level; REVERSE_EVENT *reverse_event; if (short_video_header) /* inter-VLCs will be used for both intra and inter blocks */ intra = 0; if (BitstreamShowBits(bs, 7) != ESCAPE) { reverse_event = &DCT3D[intra][BitstreamShowBits(bs, 12)]; if ((level = reverse_event->event.level) == 0) goto error; *last = reverse_event->event.last; *run = reverse_event->event.run; BitstreamSkip(bs, reverse_event->len); return BitstreamGetBits(bs, 1) ? -level : level; } BitstreamSkip(bs, 7); if (short_video_header) { /* escape mode 4 - H.263 type, only used if short_video_header = 1 */ *last = BitstreamGetBit(bs); *run = BitstreamGetBits(bs, 6); level = BitstreamGetBits(bs, 8); if (level == 0 || level == 128) DPRINTF(DPRINTF_ERROR, "Illegal LEVEL for ESCAPE mode 4: %d", level); return (level << 24) >> 24; } mode = BitstreamShowBits(bs, 2); if (mode < 3) { BitstreamSkip(bs, (mode == 2) ? 2 : 1); reverse_event = &DCT3D[intra][BitstreamShowBits(bs, 12)]; if ((level = reverse_event->event.level) == 0) goto error; *last = reverse_event->event.last; *run = reverse_event->event.run; BitstreamSkip(bs, reverse_event->len); if (mode < 2) /* first escape mode, level is offset */ level += max_level[intra][*last][*run]; else /* second escape mode, run is offset */ *run += max_run[intra][*last][level] + 1; return BitstreamGetBits(bs, 1) ? -level : level; } /* third escape mode - fixed length codes */ BitstreamSkip(bs, 2); *last = BitstreamGetBits(bs, 1); *run = BitstreamGetBits(bs, 6); BitstreamSkip(bs, 1); /* marker */ level = BitstreamGetBits(bs, 12); BitstreamSkip(bs, 1); /* marker */ return (level << 20) >> 20; error: *run = VLC_ERROR; return 0; } /***************************************************************************** * MB reading functions ****************************************************************************/ void get_intra_block(Bitstream * bs, int16_t * block, int direction, int coeff) { const uint16_t *scan = scan_tables[direction]; int level; int run; int last; do { level = get_coeff(bs, &run, &last, 1, 0); if (run == -1) { DPRINTF(DPRINTF_DEBUG, "fatal: invalid run"); break; } coeff += run; block[scan[coeff]] = level; DPRINTF(DPRINTF_COEFF,"block[%i] %i", scan[coeff], level); /*DPRINTF(DPRINTF_COEFF,"block[%i] %i %08x", scan[coeff], level, BitstreamShowBits(bs, 32)); */ if (level < -2047 || level > 2047) { DPRINTF(DPRINTF_DEBUG, "warning: intra_overflow: %d", level); } coeff++; } while (!last); } void get_inter_block(Bitstream * bs, int16_t * block) { const uint16_t *scan = scan_tables[0]; int p; int level; int run; int last; p = 0; do { level = get_coeff(bs, &run, &last, 0, 0); if (run == -1) { DPRINTF(DPRINTF_ERROR, "fatal: invalid run"); break; } p += run; block[scan[p]] = level; DPRINTF(DPRINTF_COEFF,"block[%i] %i", scan[p], level); if (level < -2047 || level > 2047) { DPRINTF(DPRINTF_DEBUG, "warning: inter_overflow: %d", level); } p++; } while (!last); }