}
}

/*!
 ************************************************************************
 * \brief
 *    For new intra pred routines
 *
 * \par Input:
 *    Image par, 16x16 based intra mode
 *
 * \par Output:
 *    none
 ************************************************************************
 */
int dct_luma_16x16(int new_intra_mode)
{
  int qp_const;
  int i,j;
  int ii,jj;
  int i1,j1;
  int M1[16][16];
  int M4[4][4];
  int M5[4],M6[4];
  int M0[4][4][4][4];
  int run,scan_pos,coeff_ctr,level;
  int qp_per,qp_rem,q_bits;
  int ac_coef = 0;

  Macroblock *currMB = &img->mb_data[img->current_mb_nr];

  int   b8, b4;
  int*  DCLevel = img->cofDC[0][0];
  int*  DCRun   = img->cofDC[0][1];
  int*  ACLevel;
  int*  ACRun;

  qp_per    = (currMB->qp-MIN_QP)/6;
  qp_rem    = (currMB->qp-MIN_QP)%6;

  q_bits    = Q_BITS+qp_per;
  
  qp_const  = (1<<q_bits)/3;

  for (j=0;j<16;j++)
  {
    for (i=0;i<16;i++)
    {
      M1[i][j]=imgY_org[img->opix_y+j][img->opix_x+i]-img->mprr_2[new_intra_mode][j][i];
      M0[i%4][i/4][j%4][j/4]=M1[i][j];
    }
  }

  for (jj=0;jj<4;jj++)
  {
    for (ii=0;ii<4;ii++)
    {
      for (j=0;j<4;j++)
      {
        for (i=0;i<2;i++)
        {
          i1=3-i;
          M5[i]=  M0[i][ii][j][jj]+M0[i1][ii][j][jj];
          M5[i1]= M0[i][ii][j][jj]-M0[i1][ii][j][jj];
        }
        M0[0][ii][j][jj]=M5[0]+M5[1];
        M0[2][ii][j][jj]=M5[0]-M5[1];
        M0[1][ii][j][jj]=M5[3]*2+M5[2];
        M0[3][ii][j][jj]=M5[3]-M5[2]*2;
      }
      // vertical
      for (i=0;i<4;i++)
      {
        for (j=0;j<2;j++)
        {
          j1=3-j;
          M5[j] = M0[i][ii][j][jj]+M0[i][ii][j1][jj];
          M5[j1]= M0[i][ii][j][jj]-M0[i][ii][j1][jj];
        }
        M0[i][ii][0][jj]=M5[0]+M5[1];
        M0[i][ii][2][jj]=M5[0]-M5[1];
        M0[i][ii][1][jj]=M5[3]*2+M5[2];
        M0[i][ii][3][jj]=M5[3] -M5[2]*2;
      }
    }
  }

  // pick out DC coeff

  for (j=0;j<4;j++)
    for (i=0;i<4;i++)
      M4[i][j]= M0[0][i][0][j];

  for (j=0;j<4;j++)
  {
    for (i=0;i<2;i++)
    {
      i1=3-i;
      M5[i]= M4[i][j]+M4[i1][j];
      M5[i1]=M4[i][j]-M4[i1][j];
    }
    M4[0][j]=M5[0]+M5[1];
    M4[2][j]=M5[0]-M5[1];
    M4[1][j]=M5[3]+M5[2];
    M4[3][j]=M5[3]-M5[2];
  }

  // vertical

  for (i=0;i<4;i++)
  {
    for (j=0;j<2;j++)
    {
      j1=3-j;
      M5[j]= M4[i][j]+M4[i][j1];
      M5[j1]=M4[i][j]-M4[i][j1];
    }
    M4[i][0]=(M5[0]+M5[1])>>1;
    M4[i][2]=(M5[0]-M5[1])>>1;
    M4[i][1]=(M5[3]+M5[2])>>1;
    M4[i][3]=(M5[3]-M5[2])>>1;
  }

  // quant

  run=-1;
  scan_pos=0;

  for (coeff_ctr=0;coeff_ctr<16;coeff_ctr++)
  {
    if (img->field_picture || ( mb_adaptive && img->field_mode )) 
    {  // Alternate scan for field coding
        i=FIELD_SCAN[coeff_ctr][0];
        j=FIELD_SCAN[coeff_ctr][1];
    }
    else 
    {
        i=SNGL_SCAN[coeff_ctr][0];
        j=SNGL_SCAN[coeff_ctr][1];
    }

    run++;

    level= (abs(M4[i][j]) * quant_coef[qp_rem][0][0] + 2*qp_const)>>(q_bits+1);

    if (level != 0)
    {
      DCLevel[scan_pos] = sign(level,M4[i][j]);
      DCRun  [scan_pos] = run;
      ++scan_pos;
      run=-1;
    }
    M4[i][j]=sign(level,M4[i][j]);
  }
  DCLevel[scan_pos]=0;

  // invers DC transform
  for (j=0;j<4;j++)
  {
    for (i=0;i<4;i++)
      M5[i]=M4[i][j];

    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 (i=0;i<2;i++)
    {
      i1=3-i;
      M4[i][j]= M6[i]+M6[i1];
      M4[i1][j]=M6[i]-M6[i1];
    }
  }

  for (i=0;i<4;i++)
  {
    for (j=0;j<4;j++)
      M5[j]=M4[i][j];

    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;
      M0[0][i][0][j] = (((M6[j]+M6[j1])*dequant_coef[qp_rem][0][0]<<qp_per)+2)>>2;
      M0[0][i][0][j1]= (((M6[j]-M6[j1])*dequant_coef[qp_rem][0][0]<<qp_per)+2)>>2;
    }
  }

  // AC invers trans/quant for MB
  for (jj=0;jj<4;jj++)
  {
    for (ii=0;ii<4;ii++)
    {
      run      = -1;
      scan_pos =  0;
      b8       = 2*(jj/2) + (ii/2);
      b4       = 2*(jj%2) + (ii%2);
      ACLevel  = img->cofAC [b8][b4][0];
      ACRun    = img->cofAC [b8][b4][1];

      for (coeff_ctr=1;coeff_ctr<16;coeff_ctr++) // set in AC coeff
      {

        if (img->field_picture || ( mb_adaptive && img->field_mode )) 
        {  // Alternate scan for field coding
          i=FIELD_SCAN[coeff_ctr][0];
          j=FIELD_SCAN[coeff_ctr][1];
        }
        else 
        {
          i=SNGL_SCAN[coeff_ctr][0];
          j=SNGL_SCAN[coeff_ctr][1];
        }
        run++;

        level= ( abs( M0[i][ii][j][jj]) * quant_coef[qp_rem][i][j] + qp_const) >> q_bits;

        if (level != 0)
        {
          ac_coef = 15;
          ACLevel[scan_pos] = sign(level,M0[i][ii][j][jj]);
          ACRun  [scan_pos] = run;
          ++scan_pos;
          run=-1;
        }
        M0[i][ii][j][jj]=sign(level*dequant_coef[qp_rem][i][j]<<qp_per,M0[i][ii][j][jj]);
      }
      ACLevel[scan_pos] = 0;


      // IDCT horizontal

      for (j=0;j<4;j++)
      {
        for (i=0;i<4;i++)
        {
          M5[i]=M0[i][ii][j][jj];
        }

        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;
          M0[i][ii][j][jj] =M6[i]+M6[i1];
          M0[i1][ii][j][jj]=M6[i]-M6[i1];
        }
      }

      // vert
      for (i=0;i<4;i++)
      {
        for (j=0;j<4;j++)
          M5[j]=M0[i][ii][j][jj];

        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;
          M0[i][ii][ j][jj]=M6[j]+M6[j1];
          M0[i][ii][j1][jj]=M6[j]-M6[j1];

        }
      }

    }
  }

  for (j=0;j<16;j++)
  {
    for (i=0;i<16;i++)
    {
      M1[i][j]=M0[i%4][i/4][j%4][j/4];
    }
  }

  for (j=0;j<16;j++)
    for (i=0;i<16;i++)
      enc_picture->imgY[img->pix_y+j][img->pix_x+i]=(byte)min(255,max(0,(M1[i][j]+(img->mprr_2[new_intra_mode][j][i]<<DQ_BITS)+DQ_ROUND)>>DQ_BITS));

    return ac_coef;
}


/*!
 ************************************************************************
 * \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.
 ************************************************************************
 */
int dct_luma(int block_x,int block_y,int *coeff_cost, int old_intra_mode)
{
  int sign(int a,int b);

  int i,j,i1,j1,ilev,m5[4],m6[4],coeff_ctr;
  int qp_const,level,scan_pos,run;
  int nonzero;
  int qp_per,qp_rem,q_bits;

  int   pos_x   = block_x/BLOCK_SIZE;
  int   pos_y   = block_y/BLOCK_SIZE;
  int   b8      = 2*(pos_y/2) + (pos_x/2);
  int   b4      = 2*(pos_y%2) + (pos_x%2);
  int*  ACLevel = img->cofAC[b8][b4][0];
  int*  ACRun   = img->cofAC[b8][b4][1];

  Macroblock *currMB = &img->mb_data[img->current_mb_nr];

  qp_per    = (currMB->qp-MIN_QP)/6;
  qp_rem    = (currMB->qp-MIN_QP)%6;
  q_bits    = Q_BITS+qp_per;

  if (img->type == I_SLICE)
    qp_const=(1<<q_bits)/3;    // intra
  else
    qp_const=(1<<q_bits)/6;    // inter

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

  //  Vertical transform
  for (i=0; i < BLOCK_SIZE; i++)
  {
    for (j=0; j < 2; j++)
    {
      j1=3-j;
      m5[j]=img->m7[i][j]+img->m7[i][j1];
      m5[j1]=img->m7[i][j]-img->m7[i][j1];
    }
    img->m7[i][0]=(m5[0]+m5[1]);
    img->m7[i][2]=(m5[0]-m5[1]);
    img->m7[i][1]=m5[3]*2+m5[2];
    img->m7[i][3]=m5[3]-m5[2]*2;
  }

  // Quant

  nonzero=FALSE;

  run=-1;
  scan_pos=0;
  
  for (coeff_ctr=0;coeff_ctr < 16;coeff_ctr++)
  {

    if (img->field_picture || ( img->MbaffFrameFlag && currMB->mb_field )) 
    {  // Alternate scan for field coding
        i=FIELD_SCAN[coeff_ctr][0];
        j=FIELD_SCAN[coeff_ctr][1];
    }
    else 
    {
        i=SNGL_SCAN[coeff_ctr][0];
        j=SNGL_SCAN[coeff_ctr][1];
    }
    
    run++;
    ilev=0;
    
    level = (abs (img->m7[i][j]) * quant_coef[qp_rem][i][j] + qp_const) >> q_bits;
    
    if (level != 0)
    {
      nonzero=TRUE;
      if (level > 1)
        *coeff_cost += MAX_VALUE;                // set high cost, shall not be discarded
      else
        *coeff_cost += COEFF_COST[run];
      ACLevel[scan_pos] = sign(level,img->m7[i][j]);
      ACRun  [scan_pos] = run;
      ++scan_pos;
      run=-1;                     // reset zero level counter
      ilev=level*dequant_coef[qp_rem][i][j]<<qp_per;
    }
    img->m7[i][j]=sign(ilev,img->m7[i][j]);
  }
  ACLevel[scan_pos] = 0;
  
  
  //     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] =min(255,max(0,(m6[j]+m6[j1]+(img->mpr[i+block_x][j+block_y] <<DQ_BITS)+DQ_ROUND)>>DQ_BITS));
      img->m7[i][j1]=min(255,max(0,(m6[j]-m6[j1]+(img->mpr[i+block_x][j1+block_y]<<DQ_BITS)+DQ_ROUND)>>DQ_BITS));
    }
  }

  //  Decoded block moved to frame memory

  for (j=0; j < BLOCK_SIZE; j++)
    for (i=0; i < BLOCK_SIZE; i++)
      enc_picture->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.
 *
 * \par Input:
 *    uv    : Make difference between the U and V chroma component  \n
 *    cr_cbp: chroma coded block pattern
 *
 * \par Output:
 *    cr_cbp: Updated chroma coded block pattern.
 ************************************************************************
 */
int dct_chroma(int uv,int cr_cbp)
{
  int i,j,i1,j2,ilev,n2,n1,j1,mb_y,coeff_ctr,qp_const,level ,scan_pos,run;
  int m1[BLOCK_SIZE],m5[BLOCK_SIZE],m6[BLOCK_SIZE];
  int coeff_cost;
  int cr_cbp_tmp;
  int nn0,nn1;
  int DCcoded=0 ;
  Macroblock *currMB = &img->mb_data[img->current_mb_nr];

  int qp_per,qp_rem,q_bits;

  int   b4;
  int*  DCLevel = img->cofDC[uv+1][0];
  int*  DCRun   = img->cofDC[uv+1][1];
  int*  ACLevel;