/*!
 *************************************************************************************
 * \file block.c
 *
 * \brief
 *    Process one block
 *
 * \author
 *    Main contributors (see contributors.h for copyright, address and affiliation details)
 *    - Inge Lille-Lang鴜               <inge.lille-langoy@telenor.com>
 *    - Rickard Sjoberg                 <rickard.sjoberg@era.ericsson.se>
 *    - Stephan Wenger                  <stewe@cs.tu-berlin.de>
 *    - Jani Lainema                    <jani.lainema@nokia.com>
 *    - Detlev Marpe                    <marpe@hhi.de>
 *    - Thomas Wedi                     <wedi@tnt.uni-hannover.de>
 *    - Ragip Kurceren                  <ragip.kurceren@nokia.com>
 *    - Greg Conklin                    <gregc@real.com>
 *************************************************************************************
 */

#include "contributors.h"


#include <math.h>
#include <stdlib.h>
#include <assert.h>

#include "block.h"
#include "refbuf.h"
#include "vlc.h"
#include "mb_access.h"
#include "image.h"


#define Q_BITS          15
#define DQ_BITS         6
#define DQ_ROUND        (1<<(DQ_BITS-1))


static const int quant_coef[6][4][4] = {
  {{13107, 8066,13107, 8066},{ 8066, 5243, 8066, 5243},{13107, 8066,13107, 8066},{ 8066, 5243, 8066, 5243}},
  {{11916, 7490,11916, 7490},{ 7490, 4660, 7490, 4660},{11916, 7490,11916, 7490},{ 7490, 4660, 7490, 4660}},
  {{10082, 6554,10082, 6554},{ 6554, 4194, 6554, 4194},{10082, 6554,10082, 6554},{ 6554, 4194, 6554, 4194}},
  {{ 9362, 5825, 9362, 5825},{ 5825, 3647, 5825, 3647},{ 9362, 5825, 9362, 5825},{ 5825, 3647, 5825, 3647}},
  {{ 8192, 5243, 8192, 5243},{ 5243, 3355, 5243, 3355},{ 8192, 5243, 8192, 5243},{ 5243, 3355, 5243, 3355}},
  {{ 7282, 4559, 7282, 4559},{ 4559, 2893, 4559, 2893},{ 7282, 4559, 7282, 4559},{ 4559, 2893, 4559, 2893}}
};

static const int dequant_coef[6][4][4] = {
  {{10, 13, 10, 13},{ 13, 16, 13, 16},{10, 13, 10, 13},{ 13, 16, 13, 16}},
  {{11, 14, 11, 14},{ 14, 18, 14, 18},{11, 14, 11, 14},{ 14, 18, 14, 18}},
  {{13, 16, 13, 16},{ 16, 20, 16, 20},{13, 16, 13, 16},{ 16, 20, 16, 20}},
  {{14, 18, 14, 18},{ 18, 23, 18, 23},{14, 18, 14, 18},{ 18, 23, 18, 23}},
  {{16, 20, 16, 20},{ 20, 25, 20, 25},{16, 20, 16, 20},{ 20, 25, 20, 25}},
  {{18, 23, 18, 23},{ 23, 29, 23, 29},{18, 23, 18, 23},{ 23, 29, 23, 29}}
};
static const int A[4][4] = {
  { 16, 20, 16, 20},
  { 20, 25, 20, 25},
  { 16, 20, 16, 20},
  { 20, 25, 20, 25}
};


// Notation for comments regarding prediction and predictors.
// The pels of the 4x4 block are labelled a..p. The predictor pels above
// are labelled A..H, from the left I..P, and from above left X, as follows:
//
//  X A B C D E F G H
//  I a b c d
//  J e f g h
//  K i j k l
//  L m n o p
//

// Predictor array index definitions
#define P_X (PredPel[0])
#define P_A (PredPel[1])
#define P_B (PredPel[2])
#define P_C (PredPel[3])
#define P_D (PredPel[4])
#define P_E (PredPel[5])
#define P_F (PredPel[6])
#define P_G (PredPel[7])
#define P_H (PredPel[8])
#define P_I (PredPel[9])
#define P_J (PredPel[10])
#define P_K (PredPel[11])
#define P_L (PredPel[12])

/*!
 ************************************************************************
 * \brief
 *    Make intra 4x4 prediction according to all 9 prediction modes.
 *    The routine uses left and upper neighbouring points from
 *    previous coded blocks to do this (if available). Notice that
 *    inaccessible neighbouring points are signalled with a negative
 *    value in the predmode array .
 *
 *  \par Input:
 *     Starting point of current 4x4 block image posision
 *
 *  \par Output:
 *      none
 ************************************************************************
 */
void intrapred_luma(int img_x,int img_y, int *left_available, int *up_available, int *all_available)
{
  int i,j;
  int s0;
  int PredPel[13];  // array of predictor pels
  byte **imgY = enc_picture->imgY;  // For MB level frame/field coding tools -- set default to imgY

  int ioff = (img_x & 15);
  int joff = (img_y & 15);
  int mb_nr=img->current_mb_nr;

  PixelPos pix_a[4];
  PixelPos pix_b, pix_c, pix_d;

  int block_available_up;
  int block_available_left;
  int block_available_up_left;
  int block_available_up_right;

  for (i=0;i<4;i++)
  {
    getNeighbour(mb_nr, ioff -1 , joff +i , 1, &pix_a[i]);
  }
  
  
  getNeighbour(mb_nr, ioff    , joff -1 , 1, &pix_b);
  getNeighbour(mb_nr, ioff +4 , joff -1 , 1, &pix_c);
  getNeighbour(mb_nr, ioff -1 , joff -1 , 1, &pix_d);

  pix_c.available = pix_c.available && !(((ioff==4)||(ioff==12)) && ((joff==4)||(joff==12)));

  if (input->UseConstrainedIntraPred)
  {
    for (i=0, block_available_left=1; i<4;i++)
      block_available_left  &= pix_a[i].available ? img->intra_block[pix_a[i].mb_addr]: 0;
    block_available_up       = pix_b.available ? img->intra_block [pix_b.mb_addr] : 0;
    block_available_up_right = pix_c.available ? img->intra_block [pix_c.mb_addr] : 0;
    block_available_up_left  = pix_d.available ? img->intra_block [pix_d.mb_addr] : 0;
  }
  else
  {
    block_available_left     = pix_a[0].available;
    block_available_up       = pix_b.available;
    block_available_up_right = pix_c.available;
    block_available_up_left  = pix_d.available;
  }
  
  *left_available = block_available_left;
  *up_available   = block_available_up;
  *all_available  = block_available_up && block_available_left && block_available_up_left;

  i = (img_x & 15);
  j = (img_y & 15);

  // form predictor pels
  if (block_available_up)
  {
    P_A = imgY[pix_b.pos_y][pix_b.pos_x+0];
    P_B = imgY[pix_b.pos_y][pix_b.pos_x+1];
    P_C = imgY[pix_b.pos_y][pix_b.pos_x+2];
    P_D = imgY[pix_b.pos_y][pix_b.pos_x+3];

  }
  else
  {
    P_A = P_B = P_C = P_D = 128;
  }

  if (block_available_up_right)
  {
    P_E = imgY[pix_c.pos_y][pix_c.pos_x+0];
    P_F = imgY[pix_c.pos_y][pix_c.pos_x+1];
    P_G = imgY[pix_c.pos_y][pix_c.pos_x+2];
    P_H = imgY[pix_c.pos_y][pix_c.pos_x+3];
  }
  else
  {
    P_E = P_F = P_G = P_H = P_D;
  }

  if (block_available_left)
  {
    P_I = imgY[pix_a[0].pos_y][pix_a[0].pos_x];
    P_J = imgY[pix_a[1].pos_y][pix_a[1].pos_x];
    P_K = imgY[pix_a[2].pos_y][pix_a[2].pos_x];
    P_L = imgY[pix_a[3].pos_y][pix_a[3].pos_x];
  }
  else
  {
    P_I = P_J = P_K = P_L = 128;
  }

  if (block_available_up_left)
  {
    P_X = imgY[pix_d.pos_y][pix_d.pos_x];
  }
  else
  {
    P_X = 128;
  }

  for(i=0;i<9;i++)
    img->mprr[i][0][0]=-1;

  ///////////////////////////////
  // make DC prediction
  ///////////////////////////////
  s0 = 0;
  if (block_available_up && block_available_left)
  {   
    // no edge
    s0 = (P_A + P_B + P_C + P_D + P_I + P_J + P_K + P_L + 4)/(2*BLOCK_SIZE);
  }
  else if (!block_available_up && block_available_left)
  {
    // upper edge
    s0 = (P_I + P_J + P_K + P_L + 2)/BLOCK_SIZE;             
  }
  else if (block_available_up && !block_available_left)
  {
    // left edge
    s0 = (P_A + P_B + P_C + P_D + 2)/BLOCK_SIZE;             
  }
  else //if (!block_available_up && !block_available_left)
  {
    // top left corner, nothing to predict from
    s0 = 128;                           
  }

  for (j=0; j < BLOCK_SIZE; j++)
  {
    for (i=0; i < BLOCK_SIZE; i++)
    {
      // store DC prediction
      img->mprr[DC_PRED][i][j] = s0;
    }
  }

  ///////////////////////////////
  // make horiz and vert prediction
  ///////////////////////////////

  for (i=0; i < BLOCK_SIZE; i++)
  {
    img->mprr[VERT_PRED][0][i] = 
    img->mprr[VERT_PRED][1][i] = 
    img->mprr[VERT_PRED][2][i] = 
    img->mprr[VERT_PRED][3][i] = (&P_A)[i];
    img->mprr[HOR_PRED][i][0]  = 
    img->mprr[HOR_PRED][i][1]  = 
    img->mprr[HOR_PRED][i][2]  = 
    img->mprr[HOR_PRED][i][3]  = (&P_I)[i];
  }

  if(!block_available_up)img->mprr[VERT_PRED][0][0]=-1;
  if(!block_available_left)img->mprr[HOR_PRED][0][0]=-1;

  if (block_available_up) 
  {
    // Mode DIAG_DOWN_LEFT_PRED
    img->mprr[DIAG_DOWN_LEFT_PRED][0][0] = (P_A + P_C + 2*(P_B) + 2) / 4;
    img->mprr[DIAG_DOWN_LEFT_PRED][0][1] = 
    img->mprr[DIAG_DOWN_LEFT_PRED][1][0] = (P_B + P_D + 2*(P_C) + 2) / 4;
    img->mprr[DIAG_DOWN_LEFT_PRED][0][2] =
    img->mprr[DIAG_DOWN_LEFT_PRED][1][1] =
    img->mprr[DIAG_DOWN_LEFT_PRED][2][0] = (P_C + P_E + 2*(P_D) + 2) / 4;
    img->mprr[DIAG_DOWN_LEFT_PRED][0][3] = 
    img->mprr[DIAG_DOWN_LEFT_PRED][1][2] = 
    img->mprr[DIAG_DOWN_LEFT_PRED][2][1] = 
    img->mprr[DIAG_DOWN_LEFT_PRED][3][0] = (P_D + P_F + 2*(P_E) + 2) / 4;
    img->mprr[DIAG_DOWN_LEFT_PRED][1][3] = 
    img->mprr[DIAG_DOWN_LEFT_PRED][2][2] = 
    img->mprr[DIAG_DOWN_LEFT_PRED][3][1] = (P_E + P_G + 2*(P_F) + 2) / 4;
    img->mprr[DIAG_DOWN_LEFT_PRED][2][3] = 
    img->mprr[DIAG_DOWN_LEFT_PRED][3][2] = (P_F + P_H + 2*(P_G) + 2) / 4;
    img->mprr[DIAG_DOWN_LEFT_PRED][3][3] = (P_G + 3*(P_H) + 2) / 4;

    // Mode VERT_LEFT_PRED
    img->mprr[VERT_LEFT_PRED][0][0] = (P_A + P_B + 1) / 2;
    img->mprr[VERT_LEFT_PRED][0][1] = 
    img->mprr[VERT_LEFT_PRED][2][0] = (P_B + P_C + 1) / 2;
    img->mprr[VERT_LEFT_PRED][0][2] = 
    img->mprr[VERT_LEFT_PRED][2][1] = (P_C + P_D + 1) / 2;
    img->mprr[VERT_LEFT_PRED][0][3] = 
    img->mprr[VERT_LEFT_PRED][2][2] = (P_D + P_E + 1) / 2;
    img->mprr[VERT_LEFT_PRED][2][3] = (P_E + P_F + 1) / 2;
    img->mprr[VERT_LEFT_PRED][1][0] = (P_A + 2*P_B + P_C + 2) / 4;
    img->mprr[VERT_LEFT_PRED][1][1] = 
    img->mprr[VERT_LEFT_PRED][3][0] = (P_B + 2*P_C + P_D + 2) / 4;
    img->mprr[VERT_LEFT_PRED][1][2] = 
    img->mprr[VERT_LEFT_PRED][3][1] = (P_C + 2*P_D + P_E + 2) / 4;
    img->mprr[VERT_LEFT_PRED][1][3] = 
    img->mprr[VERT_LEFT_PRED][3][2] = (P_D + 2*P_E + P_F + 2) / 4;
    img->mprr[VERT_LEFT_PRED][3][3] = (P_E + 2*P_F + P_G + 2) / 4;

  }

  /*  Prediction according to 'diagonal' modes */
  if (block_available_left) 
  {
    // Mode HOR_UP_PRED
    img->mprr[HOR_UP_PRED][0][0] = (P_I + P_J + 1) / 2;
    img->mprr[HOR_UP_PRED][0][1] = (P_I + 2*P_J + P_K + 2) / 4;
    img->mprr[HOR_UP_PRED][0][2] = 
    img->mprr[HOR_UP_PRED][1][0] = (P_J + P_K + 1) / 2;
    img->mprr[HOR_UP_PRED][0][3] = 
    img->mprr[HOR_UP_PRED][1][1] = (P_J + 2*P_K + P_L + 2) / 4;
    img->mprr[HOR_UP_PRED][1][2] = 
    img->mprr[HOR_UP_PRED][2][0] = (P_K + P_L + 1) / 2;
    img->mprr[HOR_UP_PRED][1][3] = 
    img->mprr[HOR_UP_PRED][2][1] = (P_K + 2*P_L + P_L + 2) / 4;
    img->mprr[HOR_UP_PRED][3][0] = 
    img->mprr[HOR_UP_PRED][2][2] = 
    img->mprr[HOR_UP_PRED][2][3] = 
    img->mprr[HOR_UP_PRED][3][1] = 
    img->mprr[HOR_UP_PRED][3][2] = 
    img->mprr[HOR_UP_PRED][3][3] = P_L;
  }

  /*  Prediction according to 'diagonal' modes */
  if (block_available_up && block_available_left && block_available_up_left) 
  {
    // Mode DIAG_DOWN_RIGHT_PRED
    img->mprr[DIAG_DOWN_RIGHT_PRED][3][0] = (P_L + 2*P_K + P_J + 2) / 4; 
    img->mprr[DIAG_DOWN_RIGHT_PRED][2][0] =
    img->mprr[DIAG_DOWN_RIGHT_PRED][3][1] = (P_K + 2*P_J + P_I + 2) / 4; 
    img->mprr[DIAG_DOWN_RIGHT_PRED][1][0] =
    img->mprr[DIAG_DOWN_RIGHT_PRED][2][1] = 
    img->mprr[DIAG_DOWN_RIGHT_PRED][3][2] = (P_J + 2*P_I + P_X + 2) / 4; 
    img->mprr[DIAG_DOWN_RIGHT_PRED][0][0] =
    img->mprr[DIAG_DOWN_RIGHT_PRED][1][1] =
    img->mprr[DIAG_DOWN_RIGHT_PRED][2][2] =
    img->mprr[DIAG_DOWN_RIGHT_PRED][3][3] = (P_I + 2*P_X + P_A + 2) / 4; 
    img->mprr[DIAG_DOWN_RIGHT_PRED][0][1] =
    img->mprr[DIAG_DOWN_RIGHT_PRED][1][2] =
    img->mprr[DIAG_DOWN_RIGHT_PRED][2][3] = (P_X + 2*P_A + P_B + 2) / 4;
    img->mprr[DIAG_DOWN_RIGHT_PRED][0][2] =
    img->mprr[DIAG_DOWN_RIGHT_PRED][1][3] = (P_A + 2*P_B + P_C + 2) / 4;
    img->mprr[DIAG_DOWN_RIGHT_PRED][0][3] = (P_B + 2*P_C + P_D + 2) / 4;

     // Mode VERT_RIGHT_PRED
    img->mprr[VERT_RIGHT_PRED][0][0] = 
    img->mprr[VERT_RIGHT_PRED][2][1] = (P_X + P_A + 1) / 2;
    img->mprr[VERT_RIGHT_PRED][0][1] = 
    img->mprr[VERT_RIGHT_PRED][2][2] = (P_A + P_B + 1) / 2;
    img->mprr[VERT_RIGHT_PRED][0][2] = 
    img->mprr[VERT_RIGHT_PRED][2][3] = (P_B + P_C + 1) / 2;
    img->mprr[VERT_RIGHT_PRED][0][3] = (P_C + P_D + 1) / 2;
    img->mprr[VERT_RIGHT_PRED][1][0] = 
    img->mprr[VERT_RIGHT_PRED][3][1] = (P_I + 2*P_X + P_A + 2) / 4;
    img->mprr[VERT_RIGHT_PRED][1][1] = 
    img->mprr[VERT_RIGHT_PRED][3][2] = (P_X + 2*P_A + P_B + 2) / 4;
    img->mprr[VERT_RIGHT_PRED][1][2] = 
    img->mprr[VERT_RIGHT_PRED][3][3] = (P_A + 2*P_B + P_C + 2) / 4;
    img->mprr[VERT_RIGHT_PRED][1][3] = (P_B + 2*P_C + P_D + 2) / 4;
    img->mprr[VERT_RIGHT_PRED][2][0] = (P_X + 2*P_I + P_J + 2) / 4;
    img->mprr[VERT_RIGHT_PRED][3][0] = (P_I + 2*P_J + P_K + 2) / 4;

    // Mode HOR_DOWN_PRED
    img->mprr[HOR_DOWN_PRED][0][0] = 
    img->mprr[HOR_DOWN_PRED][1][2] = (P_X + P_I + 1) / 2;
    img->mprr[HOR_DOWN_PRED][0][1] = 
    img->mprr[HOR_DOWN_PRED][1][3] = (P_I + 2*P_X + P_A + 2) / 4;
    img->mprr[HOR_DOWN_PRED][0][2] = (P_X + 2*P_A + P_B + 2) / 4;
    img->mprr[HOR_DOWN_PRED][0][3] = (P_A + 2*P_B + P_C + 2) / 4;
    img->mprr[HOR_DOWN_PRED][1][0] = 
    img->mprr[HOR_DOWN_PRED][2][2] = (P_I + P_J + 1) / 2;
    img->mprr[HOR_DOWN_PRED][1][1] = 
    img->mprr[HOR_DOWN_PRED][2][3] = (P_X + 2*P_I + P_J + 2) / 4;
    img->mprr[HOR_DOWN_PRED][2][0] = 
    img->mprr[HOR_DOWN_PRED][3][2] = (P_J + P_K + 1) / 2;
    img->mprr[HOR_DOWN_PRED][2][1] = 
    img->mprr[HOR_DOWN_PRED][3][3] = (P_I + 2*P_J + P_K + 2) / 4;
    img->mprr[HOR_DOWN_PRED][3][0] = (P_K + P_L + 1) / 2;
    img->mprr[HOR_DOWN_PRED][3][1] = (P_J + 2*P_K + P_L + 2) / 4;
  }
}

/*!
 ************************************************************************
 * \brief
 *    16x16 based luma prediction
 *
 * \par Input:
 *    Image parameters
 *
 * \par Output:
 *    none
 ************************************************************************
 */
void intrapred_luma_16x16()
{
  int s0=0,s1,s2;
  int s[16][2];
  int i,j;

  int ih,iv;
  int ib,ic,iaa;

  byte   **imgY_pred = enc_picture->imgY;  // For Mb level field/frame coding tools -- default to frame pred
  int          mb_nr = img->current_mb_nr;

  PixelPos up;          //!< pixel position p(0,-1)
  PixelPos left[17];    //!< pixel positions p(-1, -1..15)

  int up_avail, left_avail, left_up_avail;

  for (i=0;i<17;i++)
  {
    getNeighbour(mb_nr, -1 ,  i-1 , 1, &left[i]);
  }
  
  getNeighbour(mb_nr, 0     ,  -1 , 1, &up);

  if (!(input->UseConstrainedIntraPred))
  {
    up_avail   = up.available;
    left_avail = left[1].available;
    left_up_avail = left[0].available;
  }
  else
  {
    up_avail      = up.available ? img->intra_block[up.mb_addr] : 0;
    for (i=1, left_avail=1; i<17;i++)
      left_avail  &= left[i].available ? img->intra_block[left[i].mb_addr]: 0;
    left_up_avail = left[0].available ? img->intra_block[left[0].mb_addr]: 0;
  }

  s1=s2=0;
  // make DC prediction
  for (i=0; i < MB_BLOCK_SIZE; i++)
  {
    if (up_avail)
      s1 += imgY_pred[up.pos_y][up.pos_x+i];    // sum hor pix
    if (left_avail)
      s2 += imgY_pred[left[i+1].pos_y][left[i+1].pos_x];    // sum vert pix
  }
  if (up_avail && left_avail)
    s0=(s1+s2+16)/(2*MB_BLOCK_SIZE);             // no edge
  if (!up_avail && left_avail)
    s0=(s2+8)/MB_BLOCK_SIZE;                     // upper edge
  if (up_avail && !left_avail)
    s0=(s1+8)/MB_BLOCK_SIZE;                     // left edge
  if (!up_avail && !left_avail)
    s0=128;                                      // top left corner, nothing to predict from

  for (i=0; i < MB_BLOCK_SIZE; i++)
  {
    // vertical prediction
    if (up_avail)
      s[i][0]=imgY_pred[up.pos_y][up.pos_x+i];
    // horizontal prediction
    if (left_avail)
      s[i][1]=imgY_pred[left[i+1].pos_y][left[i+1].pos_x];
  }

  for (j=0; j < MB_BLOCK_SIZE; j++)
  {
    for (i=0; i < MB_BLOCK_SIZE; i++)
    {
      img->mprr_2[VERT_PRED_16][j][i]=s[i][0]; // store vertical prediction
      img->mprr_2[HOR_PRED_16 ][j][i]=s[j][1]; // store horizontal prediction
      img->mprr_2[DC_PRED_16  ][j][i]=s0;      // store DC prediction
    }
  }
  if (!up_avail || !left_avail || !left_up_avail) // edge
    return;

  // 16 bit integer plan pred

  ih=0;
  iv=0;
  for (i=1;i<9;i++)
  {
    if (i<8)
      ih += i*(imgY_pred[up.pos_y][up.pos_x+7+i] - imgY_pred[up.pos_y][up.pos_x+7-i]);
    else
      ih += i*(imgY_pred[up.pos_y][up.pos_x+7+i] - imgY_pred[left[0].pos_y][left[0].pos_x]);
    
    iv += i*(imgY_pred[left[8+i].pos_y][left[8+i].pos_x] - imgY_pred[left[8-i].pos_y][left[8-i].pos_x]);
  }
  ib=(5*ih+32)>>6;
  ic=(5*iv+32)>>6;
  
  iaa=16*(imgY_pred[up.pos_y][up.pos_x+15]+imgY_pred[left[16].pos_y][left[16].pos_x]);

  for (j=0;j< MB_BLOCK_SIZE;j++)
  {
    for (i=0;i< MB_BLOCK_SIZE;i++)
    {
      img->mprr_2[PLANE_16][j][i]=max(0,min(255,(iaa+(i-7)*ib +(j-7)*ic + 16)/32));// store plane prediction
    }