mbtransquant.c

mpeg4代码,比较具体

/*****************************************************************************
 *
 *  XVID MPEG-4 VIDEO CODEC
 *  - MacroBlock transfer and quantization -
 *
 *  Copyright(C) 2002-2001 Christoph Lampert <[e-mail]>
 *               2002-2001 Michael Militzer <[e-mail]>
 *               2002-2001 Peter Ross <[e-mail]>
 *               2002      Daniel Smith <[e-mail]>
 *
 *  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: mbtransquant.c,v 1.20 2002/11/26 23:44:11 edgomez Exp $
 *
 ****************************************************************************/

#include <string.h>

#include "../portab.h"
#include "mbfunctions.h"

#include "../global.h"
#include "mem_transfer.h"
#include "timer.h"
#include "../dct/fdct.h"
#include "../dct/idct.h"
#include "../quant/quant_mpeg4.h"
#include "../quant/quant_h263.h"
#include "../encoder.h"

#define MIN(X, Y) ((X)<(Y)?(X):(Y))
#define MAX(X, Y) ((X)>(Y)?(X):(Y))

#define TOOSMALL_LIMIT 3		/* skip blocks having a coefficient sum below this value */

/* this isnt pretty, but its better than 20 ifdefs */

void
MBTransQuantIntra(const MBParam * pParam,
				  FRAMEINFO * frame,
				  MACROBLOCK * pMB,
				  const uint32_t x_pos,
				  const uint32_t y_pos,
				  int16_t data[6 * 64],
				  int16_t qcoeff[6 * 64])
{

	uint32_t stride = pParam->edged_width;
	uint32_t stride2 = stride / 2;
	uint32_t next_block = stride * 8;
	uint32_t i;
	uint32_t iQuant = frame->quant;
	uint8_t *pY_Cur, *pU_Cur, *pV_Cur;
	IMAGE *pCurrent = &frame->image;

	pY_Cur = pCurrent->y + (y_pos << 4) * stride + (x_pos << 4);
	pU_Cur = pCurrent->u + (y_pos << 3) * stride2 + (x_pos << 3);
	pV_Cur = pCurrent->v + (y_pos << 3) * stride2 + (x_pos << 3);

	start_timer();
	transfer_8to16copy(&data[0 * 64], pY_Cur, stride);
	transfer_8to16copy(&data[1 * 64], pY_Cur + 8, stride);
	transfer_8to16copy(&data[2 * 64], pY_Cur + next_block, stride);
	transfer_8to16copy(&data[3 * 64], pY_Cur + next_block + 8, stride);
	transfer_8to16copy(&data[4 * 64], pU_Cur, stride2);
	transfer_8to16copy(&data[5 * 64], pV_Cur, stride2);
	stop_transfer_timer();

	start_timer();
	pMB->field_dct = 0;
	if ((frame->global_flags & XVID_INTERLACING) &&
		(x_pos>0) && (x_pos<pParam->mb_width-1) &&
		(y_pos>0) && (y_pos<pParam->mb_height-1)) {
		pMB->field_dct = MBDecideFieldDCT(data);
	}
	stop_interlacing_timer();

	for (i = 0; i < 6; i++) {
		uint32_t iDcScaler = get_dc_scaler(iQuant, i < 4);

		start_timer();
		fdct(&data[i * 64]);
		stop_dct_timer();

		if (pParam->m_quant_type == H263_QUANT) {
			start_timer();
			quant_intra(&qcoeff[i * 64], &data[i * 64], iQuant, iDcScaler);
			stop_quant_timer();

			start_timer();
			dequant_intra(&data[i * 64], &qcoeff[i * 64], iQuant, iDcScaler);
			stop_iquant_timer();
		} else {
			start_timer();
			quant4_intra(&qcoeff[i * 64], &data[i * 64], iQuant, iDcScaler);
			stop_quant_timer();

			start_timer();
			dequant4_intra(&data[i * 64], &qcoeff[i * 64], iQuant, iDcScaler);
			stop_iquant_timer();
		}

		start_timer();
		idct(&data[i * 64]);
		stop_idct_timer();
	}

	if (pMB->field_dct) {
		next_block = stride;
		stride *= 2;
	}

	start_timer();
	transfer_16to8copy(pY_Cur, &data[0 * 64], stride);
	transfer_16to8copy(pY_Cur + 8, &data[1 * 64], stride);
	transfer_16to8copy(pY_Cur + next_block, &data[2 * 64], stride);
	transfer_16to8copy(pY_Cur + next_block + 8, &data[3 * 64], stride);
	transfer_16to8copy(pU_Cur, &data[4 * 64], stride2);
	transfer_16to8copy(pV_Cur, &data[5 * 64], stride2);
	stop_transfer_timer();

}


uint8_t
MBTransQuantInter(const MBParam * pParam,
				  FRAMEINFO * frame,
				  MACROBLOCK * pMB,
				  const uint32_t x_pos,
				  const uint32_t y_pos,
				  int16_t data[6 * 64],
				  int16_t qcoeff[6 * 64])
{

	uint32_t stride = pParam->edged_width;
	uint32_t stride2 = stride / 2;
	uint32_t next_block = stride * 8;
	uint32_t i;
	uint32_t iQuant = frame->quant;
	uint8_t *pY_Cur, *pU_Cur, *pV_Cur;
	uint8_t cbp = 0;
	uint32_t sum;
	IMAGE *pCurrent = &frame->image;

	pY_Cur = pCurrent->y + (y_pos << 4) * stride + (x_pos << 4);
	pU_Cur = pCurrent->u + (y_pos << 3) * stride2 + (x_pos << 3);
	pV_Cur = pCurrent->v + (y_pos << 3) * stride2 + (x_pos << 3);

	start_timer();
	pMB->field_dct = 0;
	if ((frame->global_flags & XVID_INTERLACING) &&
		(x_pos>0) && (x_pos<pParam->mb_width-1) &&
		(y_pos>0) && (y_pos<pParam->mb_height-1)) {
		pMB->field_dct = MBDecideFieldDCT(data);
	}
	stop_interlacing_timer();

	for (i = 0; i < 6; i++) {
		/* 
		 *  no need to transfer 8->16-bit
		 * (this is performed already in motion compensation) 
		 */
		start_timer();
		fdct(&data[i * 64]);
		stop_dct_timer();

		if (pParam->m_quant_type == 0) {
			start_timer();
			sum = quant_inter(&qcoeff[i * 64], &data[i * 64], iQuant);
			stop_quant_timer();
		} else {
			start_timer();
			sum = quant4_inter(&qcoeff[i * 64], &data[i * 64], iQuant);
			stop_quant_timer();
		}

		if ((sum >= TOOSMALL_LIMIT) || (qcoeff[i*64] != 0) ||
			(qcoeff[i*64+1] != 0) || (qcoeff[i*64+8] != 0)) {

			if (pParam->m_quant_type == H263_QUANT) {
				start_timer();
				dequant_inter(&data[i * 64], &qcoeff[i * 64], iQuant);
				stop_iquant_timer();
			} else {
				start_timer();
				dequant4_inter(&data[i * 64], &qcoeff[i * 64], iQuant);
				stop_iquant_timer();
			}

			cbp |= 1 << (5 - i);

			start_timer();
			idct(&data[i * 64]);
			stop_idct_timer();
		}
	}

	if (pMB->field_dct) {
		next_block = stride;
		stride *= 2;
	}

	start_timer();
	if (cbp & 32)
		transfer_16to8add(pY_Cur, &data[0 * 64], stride);
	if (cbp & 16)
		transfer_16to8add(pY_Cur + 8, &data[1 * 64], stride);
	if (cbp & 8)
		transfer_16to8add(pY_Cur + next_block, &data[2 * 64], stride);
	if (cbp & 4)
		transfer_16to8add(pY_Cur + next_block + 8, &data[3 * 64], stride);
	if (cbp & 2)
		transfer_16to8add(pU_Cur, &data[4 * 64], stride2);
	if (cbp & 1)
		transfer_16to8add(pV_Cur, &data[5 * 64], stride2);
	stop_transfer_timer();

	return cbp;

}

void 
MBTransQuantIntra2(const MBParam * pParam,
				  FRAMEINFO * frame,
				  MACROBLOCK * pMB,
				  const uint32_t x_pos,
				  const uint32_t y_pos,
				  int16_t data[6 * 64],
				  int16_t qcoeff[6 * 64])
{
	MBTrans(pParam,frame,pMB,x_pos,y_pos,data);
	MBfDCT(pParam,frame,pMB,data);
	MBQuantIntra(pParam,frame,pMB,data,qcoeff);
	MBDeQuantIntra(pParam,frame->quant,data,qcoeff);
	MBiDCT(data,0x3F);
	MBTransAdd(pParam,frame,pMB,x_pos,y_pos,data,0x3F);
}


uint8_t
MBTransQuantInter2(const MBParam * pParam,
				  FRAMEINFO * frame,
				  MACROBLOCK * pMB,
				  const uint32_t x_pos,
				  const uint32_t y_pos,
				  int16_t data[6 * 64],
				  int16_t qcoeff[6 * 64])
{
	uint8_t cbp;
	
/* there is no MBTrans for Inter block, that's done in motion compensation already */

	MBfDCT(pParam,frame,pMB,data);
	cbp = MBQuantInter(pParam,frame->quant,data,qcoeff);
	MBDeQuantInter(pParam,frame->quant,data,qcoeff,cbp);
	MBiDCT(data,cbp);
	MBTransAdd(pParam,frame,pMB,x_pos,y_pos,data,cbp);
	
	return cbp;
}

uint8_t
MBTransQuantInterBVOP(const MBParam * pParam,
				  FRAMEINFO * frame,
				  MACROBLOCK * pMB,
				  int16_t data[6 * 64],
				  int16_t qcoeff[6 * 64])
{
	uint8_t cbp;
	
/* there is no MBTrans for Inter block, that's done in motion compensation already */

	MBfDCT(pParam,frame,pMB,data);
	cbp = MBQuantInter(pParam,frame->quant,data,qcoeff);

/* we don't have to DeQuant, iDCT and Transfer back data for B-frames */

	return cbp;
}


void
MBfDCT(const MBParam * pParam,
				  FRAMEINFO * frame,
				  MACROBLOCK * pMB,
				  int16_t data[6 * 64])
{	
	int i;

	start_timer();
	pMB->field_dct = 0;
	if ((frame->global_flags & XVID_INTERLACING)) {
		pMB->field_dct = MBDecideFieldDCT(data);
	}
	stop_interlacing_timer();

	for (i = 0; i < 6; i++) {
		start_timer();
		fdct(&data[i * 64]);
		stop_dct_timer();
	}
}

void
MBQuantDeQuantIntra(const MBParam * pParam,
				  	FRAMEINFO * frame,
				  	MACROBLOCK * pMB,