block.c

本源码是H.26L标准的Visual C++源代码

  //  vertical

  for (i=0; i < BLOCK_SIZE; i++)
  {
    for (j=0; j < BLOCK_SIZE; j++)
    {
      m5[j]=img->m7[i][j];
    }
    m6[0]=(m5[0]+m5[2])*13;
    m6[1]=(m5[0]-m5[2])*13;
    m6[2]=m5[1]*7-m5[3]*17;
    m6[3]=m5[1]*17+m5[3]*7;

    for (j=0; j < 2; j++)
    {
      j1=3-j;
      img->m7[i][j] =min(255,max(0,(m6[j]+m6[j1]+img->mpr[i+block_x][j+block_y] *JQQ1+JQQ2)/JQQ1));
      img->m7[i][j1]=min(255,max(0,(m6[j]-m6[j1]+img->mpr[i+block_x][j1+block_y] *JQQ1+JQQ2)/JQQ1));
    }
  }

  //  Decoded block moved to frame memory

  for (j=0; j < BLOCK_SIZE; j++)
    for (i=0; i < BLOCK_SIZE; i++)
      imgY[img->pix_y+block_y+j][img->pix_x+block_x+i]=img->m7[i][j];


  return nonzero;
}

/*!
 ************************************************************************
 * \brief
 *    Transform,quantization,inverse transform for chroma.
 *    The main reason why this is done in a separate routine is the
 *    additional 2x2 transform of DC-coeffs. This routine is called
 *    ones for each of the chroma components.
 *
 * \para Input:
 *    uv    : Make difference between the U and V chroma component  \n
 *    cr_cbp: chroma coded block pattern
 *
 * \para Output:
 *    cr_cbp: Updated chroma coded block pattern.
 ************************************************************************
 */
#ifndef NO_RDQUANT

int dct_chroma(int uv,int cr_cbp)
{
  int i,j,i1,j2,ilev,n2,n1,j1,mb_y,coeff_ctr,qp_const,pos_x,pos_y,quant_set,level ,scan_pos,run;
  int m1[BLOCK_SIZE],m5[BLOCK_SIZE],m6[BLOCK_SIZE];
  int coeff[16];

  if (img->type == INTRA_IMG)
    qp_const=JQQ3;
  else
    qp_const=JQQ4;

  for (n2=0; n2 <= BLOCK_SIZE; n2 += BLOCK_SIZE)
  {
    for (n1=0; n1 <= BLOCK_SIZE; n1 += BLOCK_SIZE)
    {

      //  Horizontal transform.
      for (j=0; j < BLOCK_SIZE; j++)
      {
        mb_y=n2+j;
        for (i=0; i < 2; i++)
        {
          i1=3-i;
          m5[i]=img->m7[i+n1][mb_y]+img->m7[i1+n1][mb_y];
          m5[i1]=img->m7[i+n1][mb_y]-img->m7[i1+n1][mb_y];
        }
        img->m7[n1][mb_y]=(m5[0]+m5[1])*13;
        img->m7[n1+2][mb_y]=(m5[0]-m5[1])*13;
        img->m7[n1+1][mb_y]=m5[3]*17+m5[2]*7;
        img->m7[n1+3][mb_y]=m5[3]*7-m5[2]*17;
      }

      //  Vertical transform.

      for (i=0; i < BLOCK_SIZE; i++)
      {
        j1=n1+i;
        for (j=0; j < 2; j++)
        {
          j2=3-j;
          m5[j]=img->m7[j1][n2+j]+img->m7[j1][n2+j2];
          m5[j2]=img->m7[j1][n2+j]-img->m7[j1][n2+j2];
        }
        img->m7[j1][n2+0]=(m5[0]+m5[1])*13;
        img->m7[j1][n2+2]=(m5[0]-m5[1])*13;
        img->m7[j1][n2+1]=m5[3]*17+m5[2]*7;
        img->m7[j1][n2+3]=m5[3]*7-m5[2]*17;
      }
    }
  }

  //     2X2 transform of DC coeffs.
  m1[0]=(img->m7[0][0]+img->m7[4][0]+img->m7[0][4]+img->m7[4][4])/2;
  m1[1]=(img->m7[0][0]-img->m7[4][0]+img->m7[0][4]-img->m7[4][4])/2;
  m1[2]=(img->m7[0][0]+img->m7[4][0]-img->m7[0][4]-img->m7[4][4])/2;
  m1[3]=(img->m7[0][0]-img->m7[4][0]-img->m7[0][4]+img->m7[4][4])/2;

  //     Quant of chroma 2X2 coeffs.
  quant_set=QP_SCALE_CR[img->qp];
  run=-1;
  scan_pos=0;

  for (coeff_ctr=0; coeff_ctr < 4; coeff_ctr++)
    coeff[coeff_ctr]=m1[coeff_ctr];

  rd_quant(QUANT_CHROMA_DC,coeff);

  for (coeff_ctr=0; coeff_ctr < 4; coeff_ctr++)
  {
    run++;
    ilev=0;
    level =0;

    level =(absm(coeff[coeff_ctr]));
    if (level  != 0)
    {
      currMB->cbp_blk |= 0xf0000 << (uv << 2) ;  // if one of the 2x2-DC levels is != 0 the coded-bit
      cr_cbp=max(1,cr_cbp);                      // for all 4 4x4 blocks is set (bit 16-19 or 20-23)
      img->cofu[scan_pos][0][uv]=sign(level ,m1[coeff_ctr]);
      img->cofu[scan_pos][1][uv]=run;
      scan_pos++;
      run=-1;
      ilev=level*JQ[quant_set][1];
    }
    m1[coeff_ctr]=sign(ilev,m1[coeff_ctr]);
  }

  img->cofu[scan_pos][0][uv]=0;

  //  Invers transform of 2x2 DC levels

  img->m7[0][0]=(m1[0]+m1[1]+m1[2]+m1[3])/2;
  img->m7[4][0]=(m1[0]-m1[1]+m1[2]-m1[3])/2;
  img->m7[0][4]=(m1[0]+m1[1]-m1[2]-m1[3])/2;
  img->m7[4][4]=(m1[0]-m1[1]-m1[2]+m1[3])/2;

  //     Quant of chroma AC-coeffs.

  for (n2=0; n2 <= BLOCK_SIZE; n2 += BLOCK_SIZE)
  {
    for (n1=0; n1 <= BLOCK_SIZE; n1 += BLOCK_SIZE)
    {
      pos_x=n1/BLOCK_SIZE + 2*uv;
      pos_y=n2/BLOCK_SIZE + BLOCK_SIZE;
      run=-1;
      scan_pos=0;

      for (coeff_ctr=1; coeff_ctr < 16; coeff_ctr++)
      {
        i=SNGL_SCAN[coeff_ctr][0];
        j=SNGL_SCAN[coeff_ctr][1];
        coeff[coeff_ctr-1]=img->m7[n1+i][n2+j];
      }
      rd_quant(QUANT_CHROMA_AC,coeff);

      for (coeff_ctr=1; coeff_ctr < 16; coeff_ctr++)
      {
        i=SNGL_SCAN[coeff_ctr][0];
        j=SNGL_SCAN[coeff_ctr][1];
        ++run;
        ilev=0;

        level=absm(coeff[coeff_ctr-1]);
        if (level  != 0)
        {
          currMB->cbp_blk |=  1 << (16 + (uv << 2) + ((n2 >> 1) + (n1 >> 2))) ;
          cr_cbp = 2;
          img->cof[pos_x][pos_y][scan_pos][0][0]=sign(level,img->m7[n1+i][n2+j]);
          img->cof[pos_x][pos_y][scan_pos][1][0]=run;
          ++scan_pos;
          run=-1;
          ilev=level*JQ[quant_set][1];
        }
        img->m7[n1+i][n2+j]=sign(ilev,img->m7[n1+i][n2+j]); // for use in IDCT
      }
      img->cof[pos_x][pos_y][scan_pos][0][0]=0; // EOB


      //     IDCT.

      //     Horizontal.
      for (j=0; j < BLOCK_SIZE; j++)
      {
        for (i=0; i < BLOCK_SIZE; i++)
        {
          m5[i]=img->m7[n1+i][n2+j];
        }
        m6[0]=(m5[0]+m5[2])*13;
        m6[1]=(m5[0]-m5[2])*13;
        m6[2]=m5[1]*7-m5[3]*17;
        m6[3]=m5[1]*17+m5[3]*7;

        for (i=0; i < 2; i++)
        {
          i1=3-i;
          img->m7[n1+i][n2+j]=m6[i]+m6[i1];
          img->m7[n1+i1][n2+j]=m6[i]-m6[i1];
        }
      }

      //     Vertical.
      for (i=0; i < BLOCK_SIZE; i++)
      {
        for (j=0; j < BLOCK_SIZE; j++)
        {
          m5[j]=img->m7[n1+i][n2+j];
        }
        m6[0]=(m5[0]+m5[2])*13;
        m6[1]=(m5[0]-m5[2])*13;
        m6[2]=m5[1]*7-m5[3]*17;
        m6[3]=m5[1]*17+m5[3]*7;

        for (j=0; j < 2; j++)
        {
          j2=3-j;
          img->m7[n1+i][n2+j]=min(255,max(0,(m6[j]+m6[j2]+img->mpr[n1+i][n2+j]*JQQ1+JQQ2)/JQQ1));
          img->m7[n1+i][n2+j2]=min(255,max(0,(m6[j]-m6[j2]+img->mpr[n1+i][n2+j2]*JQQ1+JQQ2)/JQQ1));
        }
      }
    }
  }

  //  Decoded block moved to memory
  for (j=0; j < BLOCK_SIZE*2; j++)
    for (i=0; i < BLOCK_SIZE*2; i++)
    {
      imgUV[uv][img->pix_c_y+j][img->pix_c_x+i]= img->m7[i][j];
    }

  return cr_cbp;
}
#endif

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

#ifdef NO_RDQUANT
int dct_chroma(int uv,int cr_cbp)
{
  int i,j,i1,j2,ilev,n2,n1,j1,mb_y,coeff_ctr,qp_const,pos_x,pos_y,quant_set,level ,scan_pos,run;
  int m1[BLOCK_SIZE],m5[BLOCK_SIZE],m6[BLOCK_SIZE];
// int coeff[16];
  int coeff_cost;
  int cr_cbp_tmp;
  int nn0,nn1;
  int DCcoded=0 ;
  Macroblock *currMB = &img->mb_data[img->current_mb_nr];

  if (img->type == INTRA_IMG)
    qp_const=JQQ3;
  else
    qp_const=JQQ4;

  for (n2=0; n2 <= BLOCK_SIZE; n2 += BLOCK_SIZE)
  {
    for (n1=0; n1 <= BLOCK_SIZE; n1 += BLOCK_SIZE)
    {

      //  Horizontal transform.
      for (j=0; j < BLOCK_SIZE; j++)
      {
        mb_y=n2+j;
        for (i=0; i < 2; i++)
        {
          i1=3-i;
          m5[i]=img->m7[i+n1][mb_y]+img->m7[i1+n1][mb_y];
          m5[i1]=img->m7[i+n1][mb_y]-img->m7[i1+n1][mb_y];
        }
        img->m7[n1][mb_y]=(m5[0]+m5[1])*13;
        img->m7[n1+2][mb_y]=(m5[0]-m5[1])*13;
        img->m7[n1+1][mb_y]=m5[3]*17+m5[2]*7;
        img->m7[n1+3][mb_y]=m5[3]*7-m5[2]*17;
      }

      //  Vertical transform.

      for (i=0; i < BLOCK_SIZE; i++)
      {
        j1=n1+i;
        for (j=0; j < 2; j++)
        {
          j2=3-j;
          m5[j]=img->m7[j1][n2+j]+img->m7[j1][n2+j2];
          m5[j2]=img->m7[j1][n2+j]-img->m7[j1][n2+j2];
        }
        img->m7[j1][n2+0]=(m5[0]+m5[1])*13;
        img->m7[j1][n2+2]=(m5[0]-m5[1])*13;
        img->m7[j1][n2+1]=m5[3]*17+m5[2]*7;
        img->m7[j1][n2+3]=m5[3]*7-m5[2]*17;
      }
    }
  }

  //     2X2 transform of DC coeffs.
  m1[0]=(img->m7[0][0]+img->m7[4][0]+img->m7[0][4]+img->m7[4][4])/2;
  m1[1]=(img->m7[0][0]-img->m7[4][0]+img->m7[0][4]-img->m7[4][4])/2;
  m1[2]=(img->m7[0][0]+img->m7[4][0]-img->m7[0][4]-img->m7[4][4])/2;
  m1[3]=(img->m7[0][0]-img->m7[4][0]-img->m7[0][4]+img->m7[4][4])/2;

//     Quant of chroma 2X2 coeffs.
  quant_set=QP_SCALE_CR[img->qp];
  run=-1;
  scan_pos=0;

  for (coeff_ctr=0; coeff_ctr < 4; coeff_ctr++)
  {
    run++;
    ilev=0;

    level =(abs(m1[coeff_ctr])*JQ[quant_set][0]+qp_const)/JQQ1;// CHANGE rd_quant removed
    if (level  != 0)
    {
      currMB->cbp_blk |= 0xf0000 << (uv << 2) ;    // if one of the 2x2-DC levels is != 0 set the
      cr_cbp=max(1,cr_cbp);                     // coded-bit all 4 4x4 blocks (bit 16-19 or 20-23)
      DCcoded = 1 ;
      img->cofu[scan_pos][0][uv]=sign(level ,m1[coeff_ctr]);
      img->cofu[scan_pos][1][uv]=run;
      scan_pos++;
      run=-1;
      ilev=level*JQ[quant_set][1];
    }
    m1[coeff_ctr]=sign(ilev,m1[coeff_ctr]);
  }
  img->cofu[scan_pos][0][uv]=0;

  //  Invers transform of 2x2 DC levels

  img->m7[0][0]=(m1[0]+m1[1]+m1[2]+m1[3])/2;
  img->m7[4][0]=(m1[0]-m1[1]+m1[2]-m1[3])/2;
  img->m7[0][4]=(m1[0]+m1[1]-m1[2]-m1[3])/2;
  img->m7[4][4]=(m1[0]-m1[1]-m1[2]+m1[3])/2;

  //     Quant of chroma AC-coeffs.
  coeff_cost=0;
  cr_cbp_tmp=0;

  for (n2=0; n2 <= BLOCK_SIZE; n2 += BLOCK_SIZE)
  {
    for (n1=0; n1 <= BLOCK_SIZE; n1 += BLOCK_SIZE)
    {
      pos_x=n1/BLOCK_SIZE + 2*uv;
      pos_y=n2/BLOCK_SIZE + BLOCK_SIZE;
      run=-1;
      scan_pos=0;

      for (coeff_ctr=1; coeff_ctr < 16; coeff_ctr++)// start change rd_quant
      {
        i = SNGL_SCAN[coeff_ctr][0];
        j = SNGL_SCAN[coeff_ctr][1];
        ++run;
        ilev=0;

        level=(abs(img->m7[n1+i][n2+j])*JQ[quant_set][0]+qp_const)/JQQ1;// CHANGE rd_quant removed
        if (level  != 0)
        {
          currMB->cbp_blk |= 1 << (16 + (uv << 2) + ((n2 >> 1) + (n1 >> 2))) ;
          if (level > 1)
            coeff_cost += MAX_VALUE;                // set high cost, shall not be discarded
          else
            coeff_cost += COEFF_COST[run];

          cr_cbp_tmp=2;
          img->cof[pos_x][pos_y][scan_pos][0][0]=sign(level,img->m7[n1+i][n2+j]);
          img->cof[pos_x][pos_y][scan_pos][1][0]=run;
          ++scan_pos;
          run=-1;
          ilev=level*JQ[quant_set][1];
        }
        img->m7[n1+i][n2+j]=sign(ilev,img->m7[n1+i][n2+j]); // for use in IDCT
      }
      img->cof[pos_x][pos_y][scan_pos][0][0]=0; // EOB
    }
  }

  // * reset chroma coeffs
  if(coeff_cost<7)
  {
    cr_cbp_tmp = 0 ;
    for (n2=0; n2 <= BLOCK_SIZE; n2 += BLOCK_SIZE)
    {
      for (n1=0; n1 <= BLOCK_SIZE; n1 += BLOCK_SIZE)
      {
        if( DCcoded == 0)                                   // if no chroma DC's: then reset
         currMB->cbp_blk &= ~(0xf0000 << (uv << 2)) ; // coded-bits of this chroma subblock
        nn0 = (n1>>2) + (uv<<1);
        nn1 = 4 + (n2>>2) ;
        img->cof[nn0][nn1][0][0][0] = 0;// dc coeff
        for (coeff_ctr=1; coeff_ctr < 16; coeff_ctr++)// ac coeff
        {
          i=SNGL_SCAN[coeff_ctr][0];
          j=SNGL_SCAN[coeff_ctr][1];
          img->m7[n1+i][n2+j]=0;
          img->cof[nn0][nn1][coeff_ctr][0][0]=0;
        }
      }
    }
  }
  if(cr_cbp_tmp==2)
      cr_cbp = 2;
  //     IDCT.

      //     Horizontal.
  for (n2=0; n2 <= BLOCK_SIZE; n2 += BLOCK_SIZE)
  {
    for (n1=0; n1 <= BLOCK_SIZE; n1 += BLOCK_SIZE)
    {
      for (j=0; j < BLOCK_SIZE; j++)
      {
        for (i=0; i < BLOCK_SIZE; i++)
        {
          m5[i]=img->m7[n1+i][n2+j];
        }
        m6[0]=(m5[0]+m5[2])*13;
        m6[1]=(m5[0]-m5[2])*13;
        m6[2]=m5[1]*7-m5[3]*17;
        m6[3]=m5[1]*17+m5[3]*7;

        for (i=0; i < 2; i++)
        {
          i1=3-i;
          img->m7[n1+i][n2+j]=m6[i]+m6[i1];
          img->m7[n1+i1][n2+j]=m6[i]-m6[i1];
        }
      }

      //     Vertical.
      for (i=0; i < BLOCK_SIZE; i++)
      {
        for (j=0; j < BLOCK_SIZE; j++)
        {
          m5[j]=img->m7[n1+i][n2+j];
        }
        m6[0]=(m5[0]+m5[2])*13;
        m6[1]=(m5[0]-m5[2])*13;
        m6[2]=m5[1]*7-m5[3]*17;
        m6[3]=m5[1]*17+m5[3]*7;

        for (j=0; j < 2; j++)
        {
          j2=3-j;
          img->m7[n1+i][n2+j]=min(255,max(0,(m6[j]+m6[j2]+img->mpr[n1+i][n2+j]*JQQ1+JQQ2)/JQQ1));
          img->m7[n1+i][n2+j2]=min(255,max(0,(m6[j]-m6[j2]+img->mpr[n1+i][n2+j2]*JQQ1+JQQ2)/JQQ1));
        }
      }
    }
  }

  //  Decoded block moved to memory
  for (j=0; j < BLOCK_SIZE*2; j++)
    for (i=0; i < BLOCK_SIZE*2; i++)
      imgUV[uv][img->pix_c_y+j][img->pix_c_x+i]= img->m7[i][j];

  return cr_cbp;
}


#endif