// vertical
  for (i=0;i<4;i++)
  {
    for (j=0;j<4;j++)
      M5[j]=img->cof[i][j][0][0];

    M6[0]=M5[0]+M5[2];
    M6[1]=M5[0]-M5[2];
    M6[2]=M5[1]-M5[3];
    M6[3]=M5[1]+M5[3];

    for (j=0;j<2;j++)
    {
      j1=3-j;
      img->cof[i][j][0][0] = (((M6[j]+M6[j1])*dequant_coef[qp_rem][0][0]<<qp_per)+2)>>2;
      img->cof[i][j1][0][0]= (((M6[j]-M6[j1])*dequant_coef[qp_rem][0][0]<<qp_per)+2)>>2;
    }
  }
}


void itrans_sp(struct img_par *img,  //!< image parameters
               int ioff,             //!< index to 4x4 block
               int joff,             //!<
               int i0,               //!<
               int j0)               //!<
{
  int i,j,i1,j1;
  int m5[4];
  int m6[4];
  int predicted_block[BLOCK_SIZE][BLOCK_SIZE],ilev;
  
  int qp_per = (img->qp-MIN_QP)/6;
  int qp_rem = (img->qp-MIN_QP)%6;
  int q_bits    = Q_BITS+qp_per;

  int qp_per_sp = (img->qpsp-MIN_QP)/6;
  int qp_rem_sp = (img->qpsp-MIN_QP)%6;
  int q_bits_sp    = Q_BITS+qp_per_sp;
  int qp_const2=(1<<q_bits_sp)/2;  //sp_pred

  if (img->type == SI_SLICE) //ES  modified
  {
    qp_per = (img->qpsp-MIN_QP)/6;
    qp_rem = (img->qpsp-MIN_QP)%6;
    q_bits = Q_BITS+qp_per;
  }

  for (j=0; j< BLOCK_SIZE; j++)
  for (i=0; i< BLOCK_SIZE; i++)
      predicted_block[i][j]=img->mpr[i+ioff][j+joff];
  for (j=0; j < BLOCK_SIZE; j++)
  {
    for (i=0; i < 2; i++)
    {
      i1=3-i;
      m5[i]=predicted_block[i][j]+predicted_block[i1][j];
      m5[i1]=predicted_block[i][j]-predicted_block[i1][j];
    }
    predicted_block[0][j]=(m5[0]+m5[1]);
    predicted_block[2][j]=(m5[0]-m5[1]);
    predicted_block[1][j]=m5[3]*2+m5[2];
    predicted_block[3][j]=m5[3]-m5[2]*2;
  }

  //  Vertival transform

  for (i=0; i < BLOCK_SIZE; i++)
  {
    for (j=0; j < 2; j++)
    {
      j1=3-j;
      m5[j]=predicted_block[i][j]+predicted_block[i][j1];
      m5[j1]=predicted_block[i][j]-predicted_block[i][j1];
    }
    predicted_block[i][0]=(m5[0]+m5[1]);
    predicted_block[i][2]=(m5[0]-m5[1]);
    predicted_block[i][1]=m5[3]*2+m5[2];
    predicted_block[i][3]=m5[3]-m5[2]*2;
  }

  for (j=0;j<BLOCK_SIZE;j++)
  for (i=0;i<BLOCK_SIZE;i++)
  {
    // recovering coefficient since they are already dequantized earlier
    img->cof[i0][j0][i][j]=(img->cof[i0][j0][i][j] >> qp_per) / dequant_coef[qp_rem][i][j]; 
    if(img->sp_switch || img->type==SI_SLICE)  //M.W. patched for SI
    {
      ilev=(abs(predicted_block[i][j]) * quant_coef[qp_rem_sp][i][j] + qp_const2) >> q_bits_sp; //ES added
      ilev= sign(ilev,predicted_block[i][j])+ img->cof[i0][j0][i][j];                           //ES added
      img->cof[i0][j0][i][j] = sign(abs(ilev) * dequant_coef[qp_rem_sp][i][j] << qp_per_sp ,ilev) ; //ES added 
    }                                                                                             //ES added
    else
    {                                                                                          //ES added
      ilev=((img->cof[i0][j0][i][j]*dequant_coef[qp_rem][i][j]*A[i][j]<< qp_per) >>6)+predicted_block[i][j] ;
      img->cof[i0][j0][i][j]=sign((abs(ilev) * quant_coef[qp_rem_sp][i][j] + qp_const2) >> q_bits_sp, ilev) * dequant_coef[qp_rem_sp][i][j] << qp_per_sp;
    }
  }
  // horizontal
  for (j=0;j<BLOCK_SIZE;j++)
  {
    for (i=0;i<BLOCK_SIZE;i++)
    {
      m5[i]=img->cof[i0][j0][i][j];
    }
    m6[0]=(m5[0]+m5[2]);
    m6[1]=(m5[0]-m5[2]);
    m6[2]=(m5[1]>>1)-m5[3];
    m6[3]=m5[1]+(m5[3]>>1);

    for (i=0;i<2;i++)
    {
      i1=3-i;
      img->m7[i][j]=m6[i]+m6[i1];
      img->m7[i1][j]=m6[i]-m6[i1];
    }
  }
  // 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]);
    m6[1]=(m5[0]-m5[2]);
    m6[2]=(m5[1]>>1)-m5[3];
    m6[3]=m5[1]+(m5[3]>>1);

    for (j=0;j<2;j++)
    {
      j1=3-j;
      img->m7[i][j] =max(0,min(255,(m6[j]+m6[j1]+DQ_ROUND)>>DQ_BITS));
      img->m7[i][j1]=max(0,min(255,(m6[j]-m6[j1]+DQ_ROUND)>>DQ_BITS));
    }
  }
}

/*!
 ***********************************************************************
 * \brief
 *    The routine performs transform,quantization,inverse transform, adds the diff.
 *    to the prediction and writes the result to the decoded luma frame. Includes the
 *    RD constrained quantization also.
 *
 * \par Input:
 *    block_x,block_y: Block position inside a macro block (0,4,8,12).
 *
 * \par Output:
 *    nonzero: 0 if no levels are nonzero.  1 if there are nonzero levels. \n
 *    coeff_cost: Counter for nonzero coefficients, used to discard expencive levels.
 ************************************************************************
 */
void copyblock_sp(struct img_par *img,int block_x,int block_y)
{
  int sign(int a,int b);

  int i,j,i1,j1,m5[4],m6[4];

  int predicted_block[BLOCK_SIZE][BLOCK_SIZE];
  int qp_per = (img->qpsp-MIN_QP)/6;
  int qp_rem = (img->qpsp-MIN_QP)%6;
  int q_bits    = Q_BITS+qp_per;
  int qp_const2=(1<<q_bits)/2;  //sp_pred


  //  Horizontal transform
  for (j=0; j< BLOCK_SIZE; j++)
  for (i=0; i< BLOCK_SIZE; i++)
    predicted_block[i][j]=img->mpr[i+block_x][j+block_y];

  for (j=0; j < BLOCK_SIZE; j++)
  {
    for (i=0; i < 2; i++)
    {
      i1=3-i;
      m5[i]=predicted_block[i][j]+predicted_block[i1][j];
      m5[i1]=predicted_block[i][j]-predicted_block[i1][j];
    }
    predicted_block[0][j]=(m5[0]+m5[1]);
    predicted_block[2][j]=(m5[0]-m5[1]);
    predicted_block[1][j]=m5[3]*2+m5[2];
    predicted_block[3][j]=m5[3]-m5[2]*2;
  }

  //  Vertival transform

  for (i=0; i < BLOCK_SIZE; i++)
  {
    for (j=0; j < 2; j++)
    {
      j1=3-j;
      m5[j]=predicted_block[i][j]+predicted_block[i][j1];
      m5[j1]=predicted_block[i][j]-predicted_block[i][j1];
    }
    predicted_block[i][0]=(m5[0]+m5[1]);
    predicted_block[i][2]=(m5[0]-m5[1]);
    predicted_block[i][1]=m5[3]*2+m5[2];
    predicted_block[i][3]=m5[3]-m5[2]*2;
  }

  // Quant
  for (j=0;j < BLOCK_SIZE; j++)
  for (i=0; i < BLOCK_SIZE; i++)
    img->m7[i][j]=sign((abs(predicted_block[i][j])* quant_coef[qp_rem][i][j]+qp_const2)>> q_bits,predicted_block[i][j])*dequant_coef[qp_rem][i][j]<<qp_per;

  //     IDCT.
  //     horizontal

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

    for (i=0;i<2;i++)
    {
      i1=3-i;
      img->m7[i][j]=m6[i]+m6[i1];
      img->m7[i1][j]=m6[i]-m6[i1];
    }
  }
  // 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]);
    m6[1]=(m5[0]-m5[2]);
    m6[2]=(m5[1]>>1)-m5[3];
    m6[3]=m5[1]+(m5[3]>>1);

    for (j=0;j<2;j++)
    {
      j1=3-j;
      img->m7[i][j] =max(0,min(255,(m6[j]+m6[j1]+DQ_ROUND)>>DQ_BITS));
      img->m7[i][j1]=max(0,min(255,(m6[j]-m6[j1]+DQ_ROUND)>>DQ_BITS));
    }
  }

  //  Decoded block moved to frame memory

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

}

void itrans_sp_chroma(struct img_par *img,int ll)
{
  int i,j,i1,j2,ilev,n2,n1,j1,mb_y;
  int m5[BLOCK_SIZE];
  int predicted_chroma_block[MB_BLOCK_SIZE/2][MB_BLOCK_SIZE/2],mp1[BLOCK_SIZE];
  int qp_per,qp_rem,q_bits;
  int qp_per_sp,qp_rem_sp,q_bits_sp,qp_const2;

  qp_per    = ((img->qp<0?img->qp:QP_SCALE_CR[img->qp])-MIN_QP)/6;
  qp_rem    = ((img->qp<0?img->qp:QP_SCALE_CR[img->qp])-MIN_QP)%6;
  q_bits    = Q_BITS+qp_per;

  qp_per_sp    = ((img->qpsp<0?img->qpsp:QP_SCALE_CR[img->qpsp])-MIN_QP)/6;
  qp_rem_sp    = ((img->qpsp<0?img->qpsp:QP_SCALE_CR[img->qpsp])-MIN_QP)%6;
  q_bits_sp    = Q_BITS+qp_per_sp;
  qp_const2=(1<<q_bits_sp)/2;  //sp_pred

  if (img->type == SI_SLICE)
  {
    qp_per    = ((img->qpsp < 0 ? img->qpsp : QP_SCALE_CR[img->qpsp]) - MIN_QP) / 6;
    qp_rem    = ((img->qpsp < 0 ? img->qpsp : QP_SCALE_CR[img->qpsp]) - MIN_QP) % 6;
    q_bits    = Q_BITS + qp_per;
  }

  for (j=0; j < MB_BLOCK_SIZE/2; j++)
  for (i=0; i < MB_BLOCK_SIZE/2; i++)
  {
    predicted_chroma_block[i][j]=img->mpr[i][j];
    img->mpr[i][j]=0;
  }
  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]=predicted_chroma_block[i+n1][mb_y]+predicted_chroma_block[i1+n1][mb_y];
          m5[i1]=predicted_chroma_block[i+n1][mb_y]-predicted_chroma_block[i1+n1][mb_y];
        }
        predicted_chroma_block[n1][mb_y]  =(m5[0]+m5[1]);
        predicted_chroma_block[n1+2][mb_y]=(m5[0]-m5[1]);
        predicted_chroma_block[n1+1][mb_y]=m5[3]*2+m5[2];
        predicted_chroma_block[n1+3][mb_y]=m5[3]-m5[2]*2;
      }

      //  Vertical transform.

      for (i=0; i < BLOCK_SIZE; i++)
      {
        j1=n1+i;
        for (j=0; j < 2; j++)
        {
          j2=3-j;
          m5[j]=predicted_chroma_block[j1][n2+j]+predicted_chroma_block[j1][n2+j2];
          m5[j2]=predicted_chroma_block[j1][n2+j]-predicted_chroma_block[j1][n2+j2];
        }
        predicted_chroma_block[j1][n2+0]=(m5[0]+m5[1]);
        predicted_chroma_block[j1][n2+2]=(m5[0]-m5[1]);
        predicted_chroma_block[j1][n2+1]=m5[3]*2+m5[2];
        predicted_chroma_block[j1][n2+3]=m5[3]-m5[2]*2;
      }
    }
  }

  //     2X2 transform of DC coeffs.
  mp1[0]=(predicted_chroma_block[0][0]+predicted_chroma_block[4][0]+predicted_chroma_block[0][4]+predicted_chroma_block[4][4]);
  mp1[1]=(predicted_chroma_block[0][0]-predicted_chroma_block[4][0]+predicted_chroma_block[0][4]-predicted_chroma_block[4][4]);
  mp1[2]=(predicted_chroma_block[0][0]+predicted_chroma_block[4][0]-predicted_chroma_block[0][4]-predicted_chroma_block[4][4]);
  mp1[3]=(predicted_chroma_block[0][0]-predicted_chroma_block[4][0]-predicted_chroma_block[0][4]+predicted_chroma_block[4][4]);

  for (n1=0; n1 < 2; n1 ++)
  for (n2=0; n2 < 2; n2 ++)
  {
    if (img->sp_switch || img->type==SI_SLICE)  //M.W. patched for SI
    {
      //quantization fo predicted block
      ilev=(abs (mp1[n1+n2*2]) * quant_coef[qp_rem_sp][0][0] + 2 * qp_const2) >> (q_bits_sp + 1); 
      //addition 	  
      ilev=img->cof[n1+ll][4+n2][0][0]+sign(ilev,mp1[n1+n2*2]);                                   
      //dequantization
      mp1[n1+n2*2] =ilev*dequant_coef[qp_rem_sp][0][0]<<qp_per_sp;                                
    }   
    else
    {
      ilev=((img->cof[n1+ll][4+n2][0][0]*dequant_coef[qp_rem][0][0]*A[0][0]<< qp_per) >>5)+mp1[n1+n2*2] ;
      mp1[n1+n2*2]=sign((abs(ilev)* quant_coef[qp_rem_sp][0][0]+ 2 * qp_const2)>> (q_bits_sp+1),ilev)*dequant_coef[qp_rem_sp][0][0]<<qp_per_sp;
    }
  }


  for (n2=0; n2 < 2; n2 ++)
  for (n1=0; n1 < 2; n1 ++)
  for (i=0;i< BLOCK_SIZE; i++)
  for (j=0;j< BLOCK_SIZE; j++)
  {
  // recovering coefficient since they are already dequantized earlier
    img->cof[n1+ll][4+n2][i][j] = (img->cof[n1+ll][4+n2][i][j] >> qp_per) / dequant_coef[qp_rem][i][j];

    if (img->sp_switch || img->type==SI_SLICE)  //M.W. patched for SI
    {
      //quantization of the predicted block
      ilev =  (abs(predicted_chroma_block[n1*BLOCK_SIZE+i][n2*BLOCK_SIZE+j]) * quant_coef[qp_rem_sp][i][j] + qp_const2) >> q_bits_sp;
      //addition of the residual
      ilev = sign(ilev,predicted_chroma_block[n1*BLOCK_SIZE+i][n2*BLOCK_SIZE+j]) + img->cof[n1+ll][4+n2][i][j];
      // Inverse quantization 
      img->cof[n1+ll][4+n2][i][j] = ilev * dequant_coef[qp_rem_sp][i][j] << qp_per_sp  ;
    }
    else
    {
      //dequantization and addition of the predicted block
      ilev=((img->cof[n1+ll][4+n2][i][j]*dequant_coef[qp_rem][i][j]*A[i][j]<< qp_per) >>6)+predicted_chroma_block[n1*BLOCK_SIZE+i][n2*BLOCK_SIZE+j] ;
      //quantization and dequantization
      img->cof[n1+ll][4+n2][i][j] = sign((abs(ilev) * quant_coef[qp_rem_sp][i][j] + qp_const2)>> q_bits_sp,ilev)*dequant_coef[qp_rem_sp][i][j]<<qp_per_sp;
    }
  }
  img->cof[0+ll][4][0][0]=(mp1[0]+mp1[1]+mp1[2]+mp1[3])>>1;
  img->cof[1+ll][4][0][0]=(mp1[0]-mp1[1]+mp1[2]-mp1[3])>>1;
  img->cof[0+ll][5][0][0]=(mp1[0]+mp1[1]-mp1[2]-mp1[3])>>1;
  img->cof[1+ll][5][0][0]=(mp1[0]-mp1[1]-mp1[2]+mp1[3])>>1;
}

int sign(int a , int b)
{
  int x;

  x=abs(a);
  if (b>0)
    return(x);
  else return(-x);
}