jfdct_bin_b1.c,v

JPEG Image compression using IJG standards followed

head	1.1;
access;
symbols;
locks
	jliang:1.1; strict;
comment	@ * @;


1.1
date	2000.07.23.15.37.11;	author jliang;	state Exp;
branches;
next	;


desc
@@


1.1
log
@Initial revision
@
text
@/*
 * jfdct_bin_b1.c
 *
 * Variation 4 of the binDCT from Chen's factorization (see the SPIE paper):
 * Cost:        19 shifts, 37 Adds. 
 * Coding Gain: 8.8220 dB, 
 * MSE:         8.49E-5.
 *
 *
 */

/*
 ************************************************
 *
 * $Log$
 *
 ************************************************
 */

/*
**********************************************************************
*
* Modification History:
* Date       Programmer   Description
* --------   ----------   --------------------------------------------

*
***********************************************************************
*/

#define JPEG_INTERNALS
#include "jinclude.h"
#include "jpeglib.h"
#include "jdct.h"		/* Private declarations for DCT subsystem */

#ifdef DCT_BIN_B1_SUPPORTED

/*
 * This module is specialized to the case DCTSIZE = 8.
 */

#if DCTSIZE != 8
  Sorry, this code only copes with 8x8 DCTs. /* deliberate syntax err */
#endif

#if BITS_IN_JSAMPLE == 8
#define CONST_BITS  13
#define PASS1_BITS  2
#else
#define CONST_BITS  13
#define PASS1_BITS  1		/* lose a little precision to avoid overflow */
#endif

/* Multiply an INT32 variable by an INT32 constant to yield an INT32 result.
 * For 8-bit samples with the recommended scaling, all the variable
 * and constant values involved are no more than 16 bits wide, so a
 * 16x16->32 bit multiply can be used instead of a full 32x32 multiply.
 * For 12-bit samples, a full 32-bit multiplication will be needed.
 */

#if BITS_IN_JSAMPLE == 8
#define MULTIPLY(var,const)  MULTIPLY16C16(var,const)
#else
#define MULTIPLY(var,const)  ((var) * (const))
#endif


//Jie 07/09/00: lossless binDCT flag, defined in cjpeg.
extern boolean lossless_codec;


/*
 * Perform the forward binDCT-A on one block of samples.
 * Cost: 17 shifts.
 */

GLOBAL(void)
jpeg_fdct_bin_b1 (DCTELEM * data)
{
  INT32 tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
  INT32 tmp10, tmp11, tmp12, tmp13;
  INT32 z1, z2, z3, z4, z5;
  DCTELEM *dataptr;
  int ctr;
  SHIFT_TEMPS

	/*	fprintf(stderr, "jpeg_fdct_bin_a1 ...\n"); */

  /* Pass 1: process rows. */
  /* Note results are scaled up by sqrt(8) compared to a true DCT; */
  /* furthermore, we scale the results by 2**PASS1_BITS. */

	//fprintf(stderr, "\nOriginal image:\n");

//Case 1: lossless binDCT not required. Use original butterflies.
if (!lossless_codec){

  dataptr = data;
  for (ctr = DCTSIZE-1; ctr >= 0; ctr--) {


/*******************/
/* Jie: test code */
	//	for (tmp0 = 0; tmp0 < 8; tmp0 ++) {
	//	  fprintf(stderr, "%10d", dataptr[tmp0]+128);
	//	}
	//	fprintf(stderr, "\n");

    tmp0 = dataptr[0] + dataptr[7];
    tmp7 = dataptr[0] - dataptr[7];
    tmp1 = dataptr[1] + dataptr[6];
    tmp6 = dataptr[1] - dataptr[6];
    tmp2 = dataptr[2] + dataptr[5];
    tmp5 = dataptr[2] - dataptr[5];
    tmp3 = dataptr[3] + dataptr[4];
    tmp4 = dataptr[3] - dataptr[4];
    
    /* Even part */
    tmp10 = (tmp0 + tmp3) ;	/* phase 2 */
    tmp13 = tmp0 - tmp3;
    tmp11 = (tmp1 + tmp2) ;
    tmp12 = tmp1 - tmp2;
    
    dataptr[0] = (tmp10 + tmp11);             /* phase 3 */
	dataptr[4] = ((dataptr[0] + 1) >> 1) - tmp11;   /* Jie 05/18/00 */
	//dataptr[4] = tmp10 - tmp11;

	/*3/8, -3/8: alter the sign to get positive scaling factor */   
	/* new version: 7/16 and -3/8 */
    dataptr[6] = ( ((tmp13 << 3) - tmp13 + 8) >> 4) - tmp12;     /* 1/2 - 1/16 */
    dataptr[2] = tmp13 - (((dataptr[6] << 1) + dataptr[6] + 4) >> 3);
    
    /* Odd part */

/* pi/4 = -7/16u 11/16d -3/8u*/
	tmp10 = tmp5 - (( (tmp6 << 3) - tmp6 + 8) >> 4);
	tmp6 = tmp6 + tmp10 - (((tmp10 << 2) + tmp10 + 8) >> 4);
	tmp5 = (((tmp6 << 1) + tmp6 + 4) >> 3) - tmp10;

    tmp10 = tmp4 + tmp5;
    tmp11 = tmp4 - tmp5;
    tmp12 = tmp7 - tmp6;
    tmp13 = tmp7 + tmp6;

    /* 7pi/16 = 3/16u -3/16d: alter the sign to get positive scaling factor */
    dataptr[7] = (((tmp13 << 1) + tmp13 + 8) >> 4) - tmp10;
	dataptr[1] = tmp13 - (((dataptr[7] << 1) + dataptr[7] + 8) >> 4);

    /* 3pi/16 = */
	/* new version: 5/8, -7/16 */
    dataptr[5] = tmp11 + (((tmp12 << 2) + tmp12 + 4) >> 3);
    dataptr[3] = tmp12 - (((dataptr[5] << 3) - dataptr[5] + 8) >> 4);
   
    dataptr += DCTSIZE;		/* advance pointer to next row */
  }

  /* Pass 2: process columns.
   * We remove the PASS1_BITS scaling, but leave the results scaled up
   * by an overall factor of 8.
   */

  dataptr = data;
  for (ctr = DCTSIZE-1; ctr >= 0; ctr--) {
    tmp0 = (dataptr[DCTSIZE*0] + dataptr[DCTSIZE*7]);
    tmp7 = (dataptr[DCTSIZE*0] - dataptr[DCTSIZE*7]);
    tmp1 = (dataptr[DCTSIZE*1] + dataptr[DCTSIZE*6]);
    tmp6 = (dataptr[DCTSIZE*1] - dataptr[DCTSIZE*6]);
    tmp2 = (dataptr[DCTSIZE*2] + dataptr[DCTSIZE*5]);
    tmp5 = (dataptr[DCTSIZE*2] - dataptr[DCTSIZE*5]);
    tmp3 = (dataptr[DCTSIZE*3] + dataptr[DCTSIZE*4]);
    tmp4 = (dataptr[DCTSIZE*3] - dataptr[DCTSIZE*4]);

    /* Even part */
    
    tmp10 = (tmp0 + tmp3);	/* phase 2 */
    tmp13 = tmp0 - tmp3;
    tmp11 = (tmp1 + tmp2);
    tmp12 = tmp1 - tmp2;
    
    dataptr[DCTSIZE*0] = (tmp10 + tmp11); /* phase 3 */
	dataptr[DCTSIZE*4] = ((dataptr[DCTSIZE*0] + 1) >> 1) - tmp11;   /* Jie 05/18/00 */
    //dataptr[DCTSIZE*4] = (tmp10 - tmp11); /* phase 3 */

	// 7/16, 3/8
    dataptr[DCTSIZE*6] = (((tmp13 << 3) - tmp13 + 8) >> 4) - tmp12;
    dataptr[DCTSIZE*2] = tmp13 - (((dataptr[DCTSIZE*6] << 1) + dataptr[DCTSIZE*6] + 4) >> 3);
 
    /* Odd part */

/* pi/4 = -7/16u 11/16d -3/8u*/
	tmp10 = tmp5 - (((tmp6 << 3) - tmp6 + 8) >> 4);
	tmp6 = tmp6 + tmp10 - (((tmp10 << 2) + tmp10 + 8) >> 4);
	tmp5 = (((tmp6 << 1) + tmp6 + 4) >> 3) - tmp10;

    tmp10 = tmp4 + tmp5;
    tmp11 = tmp4 - tmp5;
    tmp12 = tmp7 - tmp6;
    tmp13 = tmp7 + tmp6;

    /* 7pi/16 = 3/16u -3/16d: alter sign to get positive scaling factor */
    dataptr[DCTSIZE*7] = (((tmp13 << 1) + tmp13 + 8) >> 4) - tmp10;
	dataptr[DCTSIZE*1] = tmp13 - (((dataptr[DCTSIZE*7] << 1) + dataptr[DCTSIZE*7] + 8) >> 4);

    /* 3pi/16 = */
	/* new : 5/8 and -7/16 */
    dataptr[DCTSIZE*5] = tmp11 + (((tmp12 << 2) + tmp12 + 4) >> 3);
    dataptr[DCTSIZE*3] = tmp12 - (((dataptr[DCTSIZE*5] << 3) - dataptr[DCTSIZE*5] + 8) >> 4);

     dataptr++;			/* advance pointer to next column */

  }

/****************************************************************************/
   
} else {

/****************************************************************************/
  //Case 2: lossless binDCT: Use new butterflies.

  dataptr = data;
  for (ctr = DCTSIZE-1; ctr >= 0; ctr--) {


/*******************/
/* Jie: test code */
	//	for (tmp0 = 0; tmp0 < 8; tmp0 ++) {
	//	  fprintf(stderr, "%10d", dataptr[tmp0]+128);
	//	}
	//	fprintf(stderr, "\n");

    tmp7 = dataptr[0] - dataptr[7];
    tmp0 = dataptr[0] - ((tmp7 + 1) >> 1);

    tmp6 = dataptr[1] - dataptr[6];
    tmp1 = dataptr[1] - ((tmp6 + 1) >> 1);

    tmp5 = dataptr[2] - dataptr[5];
    tmp2 = dataptr[2] - ((tmp5 + 1) >> 1);

    tmp4 = dataptr[3] - dataptr[4];
    tmp3 = dataptr[3] - ((tmp4 + 1) >> 1);
    
    /* Even part */
	//old method: larger dynamic range.
	//    tmp10 = (tmp0 + tmp3) ;	/* phase 2 */
	//    tmp13 = ((tmp10 + 1) >> 1) - tmp3;
	tmp13 = tmp0 - tmp3;
 	tmp10 = tmp0 - ((tmp13 + 1) >> 1);	

	//    tmp11 = (tmp1 + tmp2) ;
	//    tmp12 = ((tmp11 + 1) >> 1) - tmp2;
	tmp12 = tmp1 - tmp2;
	tmp11 = tmp1 - ((tmp12 + 1) >> 1);   

    //    dataptr[0] = (tmp10 + tmp11);             /* phase 3 */
	//    dataptr[4] = ((dataptr[0] + 1) >> 1) - tmp11;   /* Jie 05/18/00 */
	dataptr[4] = tmp10 - tmp11;
    dataptr[0] = tmp10 - ((dataptr[4] + 1) >> 1);

	/*3/8, -3/8: alter the sign to get positive scaling factor */   
	/* new version: 7/16 and -3/8 */
    dataptr[6] = ( ((tmp13 << 3) - tmp13 + 8) >> 4) - tmp12;     /* 1/2 - 1/16 */
    dataptr[2] = tmp13 - (((dataptr[6] << 1) + dataptr[6] + 4) >> 3);
    
    /* Odd part */

/* pi/4 = -7/16u 11/16d -3/8u*/
	tmp10 = tmp5 - (( (tmp6 << 3) - tmp6 + 8) >> 4);
	tmp6 = tmp6 + tmp10 - (((tmp10 << 2) + tmp10 + 8) >> 4);
	tmp5 = (((tmp6 << 1) + tmp6 + 4) >> 3) - tmp10;

	//butterflies:
    //tmp10 = tmp4 + tmp5;
    //tmp11 = ((tmp10 + 1) >> 1) - tmp5;
	tmp11 = tmp4 - tmp5;
	tmp10 = tmp4 - ((tmp11 + 1) >> 1);

    //tmp13 = tmp6 + tmp7;
    //tmp12 = ((tmp13 + 1) >> 1) - tmp6;
	tmp12 = tmp7 - tmp6;
	tmp13 = tmp7 - (( tmp12 + 1) >> 1);

    /* 7pi/16 = 3/16u -3/16d: alter the sign to get positive scaling factor */
    dataptr[7] = (((tmp13 << 1) + tmp13 + 8) >> 4) - tmp10;
	dataptr[1] = tmp13 - (((dataptr[7] << 1) + dataptr[7] + 8) >> 4);

    /* 3pi/16 = */
	/* new version: 5/8, -7/16 */
    dataptr[5] = tmp11 + (((tmp12 << 2) + tmp12 + 4) >> 3);
    dataptr[3] = tmp12 - (((dataptr[5] << 3) - dataptr[5] + 8) >> 4);
    
    dataptr += DCTSIZE;		/* advance pointer to next row */
  }

  /* Pass 2: process columns.
   * We remove the PASS1_BITS scaling, but leave the results scaled up
   * by an overall factor of 8.
   */

  dataptr = data;
  for (ctr = DCTSIZE-1; ctr >= 0; ctr--) {

    tmp7 = dataptr[DCTSIZE*0] - dataptr[DCTSIZE*7];
    tmp0 = dataptr[DCTSIZE*0] - ((tmp7 + 1) >> 1);

    tmp6 = dataptr[DCTSIZE*1] - dataptr[DCTSIZE*6];
    tmp1 = dataptr[DCTSIZE*1] - ((tmp6 + 1) >> 1);

    tmp5 = dataptr[DCTSIZE*2] - dataptr[DCTSIZE*5];
    tmp2 = dataptr[DCTSIZE*2] - ((tmp5 + 1) >> 1);

    tmp4 = dataptr[DCTSIZE*3] - dataptr[DCTSIZE*4];
    tmp3 = dataptr[DCTSIZE*3] - ((tmp4 + 1) >> 1);

    /* Even part */
    
    //tmp10 = (tmp0 + tmp3); 
    //tmp13 = ((tmp10 + 1) >> 1) - tmp3;
	tmp13 = tmp0 - tmp3;
 	tmp10 = tmp0 - ((tmp13 + 1) >> 1);

    //tmp11 = (tmp1 + tmp2);
    //tmp12 = ((tmp11 + 1) >> 1) - tmp2;
	tmp12 = tmp1 - tmp2;
	tmp11 = tmp1 - ((tmp12 + 1) >> 1);

	//old method: larger dynamic range
    //dataptr[DCTSIZE*0] = (tmp10 + tmp11); /* phase 3 */
	//dataptr[DCTSIZE*4] = ((dataptr[DCTSIZE*0] + 1) >> 1) - tmp11;   /* Jie 05/18/00 */
	dataptr[DCTSIZE*4] = tmp10 - tmp11;
    dataptr[DCTSIZE*0] = tmp10 - ((dataptr[DCTSIZE*4] + 1) >> 1);

	// 7/16, 3/8
    dataptr[DCTSIZE*6] = (((tmp13 << 3) - tmp13 + 8) >> 4) - tmp12;
    dataptr[DCTSIZE*2] = tmp13 - (((dataptr[DCTSIZE*6] << 1) + dataptr[DCTSIZE*6] + 4) >> 3);
 
    /* Odd part */

/* pi/4 = -7/16u 11/16d -3/8u*/
	tmp10 = tmp5 - (((tmp6 << 3) - tmp6 + 8) >> 4);
	tmp6 = tmp6 + tmp10 - (((tmp10 << 2) + tmp10 + 8) >> 4);
	tmp5 = (((tmp6 << 1) + tmp6 + 4) >> 3) - tmp10;

	//butterflies
	//old
    //tmp10 = tmp4 + tmp5;
    //tmp11 = ((tmp10 + 1) >> 1) - tmp5;
	tmp11 = tmp4 - tmp5;
	tmp10 = tmp4 - ((tmp11 + 1) >> 1);

	//old
    //tmp13 = tmp7 + tmp6;
    //tmp12 = ((tmp13 + 1) >> 1) - tmp6;
	tmp12 = tmp7 - tmp6;
	tmp13 = tmp7 - (( tmp12 + 1) >> 1);

    /* 7pi/16 = 3/16u -3/16d: alter sign to get positive scaling factor */
    dataptr[DCTSIZE*7] = (((tmp13 << 1) + tmp13 + 8) >> 4) - tmp10;
	dataptr[DCTSIZE*1] = tmp13 - (((dataptr[DCTSIZE*7] << 1) + dataptr[DCTSIZE*7] + 8) >> 4);

    /* 3pi/16 = */
	/* new : 5/8 and -7/16 */
    dataptr[DCTSIZE*5] = tmp11 + (((tmp12 << 2) + tmp12 + 4) >> 3);
    dataptr[DCTSIZE*3] = tmp12 - (((dataptr[DCTSIZE*5] << 3) - dataptr[DCTSIZE*5] + 8) >> 4);

    dataptr++;			/* advance pointer to next column */
  }

} // lossless ot not.
 
}

/**************************************
 * problem with Sachin's code:
 * 1: it's verison binDCT-C.
 * 2: X[0] and X[4]: butterfly, not lifting.
 * 3: Neg scaling required.
 **************************************
 */

#endif /* DCT_BIN_B1_SUPPORTED */
@