uvg266/src/inter.c

/*****************************************************************************
 * This file is part of Kvazaar HEVC encoder.
 *
 * Copyright (C) 2013-2015 Tampere University of Technology and others (see
 * COPYING file).
 *
 * Kvazaar is free software: you can redistribute it and/or modify it under
 * the terms of the GNU Lesser General Public License as published by the
 * Free Software Foundation; either version 2.1 of the License, or (at your
 * option) any later version.
 *
 * Kvazaar is distributed in the hope that it will be useful, but WITHOUT ANY
 * WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
 * FOR A PARTICULAR PURPOSE.  See the GNU Lesser General Public License for
 * more details.
 *
 * You should have received a copy of the GNU General Public License along
 * with Kvazaar.  If not, see <http://www.gnu.org/licenses/>.
 ****************************************************************************/

/*
 * \file
 */

#include "inter.h"

#include <stdio.h>
#include <stdlib.h>
#include <string.h>

#include "config.h"
#include "filter.h"
#include "strategies/strategies-ipol.h"

/**
 * \brief Set block info to the CU structure
 * \param pic picture to use
 * \param x_cu x CU position (smallest CU)
 * \param y_cu y CU position (smallest CU)
 * \param depth current CU depth
 * \param cur_cu CU to take the settings from
 * \returns Void
*/
void inter_set_block(videoframe_t* frame, uint32_t x_cu, uint32_t y_cu, uint8_t depth, cu_info_t* cur_cu)
{
  uint32_t x, y;
  // Width in smallest CU
  int block_scu_width = (LCU_WIDTH>>depth)/(LCU_WIDTH>>MAX_DEPTH);
  int tr_depth = (depth == 0 ? 1 : depth);
  // Loop through all the block in the area of cur_cu
  for (y = y_cu; y < y_cu + block_scu_width; y++) {
    for (x = x_cu; x < x_cu + block_scu_width; x++) {
      cu_info_t * const cu = videoframe_get_cu(frame, x, y);
      // Set all SCU's to this blocks values at the bottom most depth.
      cu->depth = depth;
      cu->type  = CU_INTER;
      cu->part_size = SIZE_2Nx2N;
      cu->inter.mode   = cur_cu->inter.mode;
      cu->inter.mv[0]  = cur_cu->inter.mv[0];
      cu->inter.mv[1]  = cur_cu->inter.mv[1];
      cu->inter.mv_dir = cur_cu->inter.mv_dir;
      cu->inter.mv_ref = cur_cu->inter.mv_ref;
      cu->inter.mv_ref_coded = cur_cu->inter.mv_ref_coded;
      cu->tr_depth = tr_depth;
    }
  }
}

/**
 * \brief Reconstruct inter block
 * \param ref picture to copy the data from
 * \param xpos block x position
 * \param ypos block y position
 * \param width block width
 * \param mv[2] motion vector
 * \param lcu destination lcu
 * \returns Void
*/
void inter_recon_lcu(const encoder_state_t * const state, const image_t * const ref, int32_t xpos, int32_t ypos,int32_t width, const int16_t mv_param[2], lcu_t *lcu)
{
  int x,y,coord_x,coord_y;
  int16_t mv[2] = { mv_param[0], mv_param[1] };

  int32_t dst_width_c = LCU_WIDTH>>1; //!< Destination picture width in chroma pixels
  int32_t ref_width_c = ref->width>>1; //!< Reference picture width in chroma pixels

  // negative overflow flag
  int8_t overflow_neg_x = (state->tile->lcu_offset_x * LCU_WIDTH + xpos + (mv[0]>>2) < 0)?1:0;
  int8_t overflow_neg_y = (state->tile->lcu_offset_y * LCU_WIDTH + ypos + (mv[1]>>2) < 0)?1:0;

  // positive overflow flag
  int8_t overflow_pos_x = (state->tile->lcu_offset_x * LCU_WIDTH + xpos + (mv[0]>>2) + width > ref->width )?1:0;
  int8_t overflow_pos_y = (state->tile->lcu_offset_y * LCU_WIDTH + ypos + (mv[1]>>2) + width > ref->height)?1:0;

  // Chroma half-pel
  #define HALFPEL_CHROMA_WIDTH ((LCU_WIDTH>>1) + 8)
  int8_t chroma_halfpel = ((mv[0]>>2)&1) || ((mv[1]>>2)&1); //!< (luma integer mv) lsb is set -> chroma is half-pel
  pixel_t halfpel_src_u[HALFPEL_CHROMA_WIDTH * HALFPEL_CHROMA_WIDTH]; //!< U source block for interpolation
  pixel_t halfpel_src_v[HALFPEL_CHROMA_WIDTH * HALFPEL_CHROMA_WIDTH]; //!< V source block for interpolation
  pixel_t *halfpel_src_off_u = &halfpel_src_u[HALFPEL_CHROMA_WIDTH * 4 + 4]; //!< halfpel_src_u with offset (4,4)
  pixel_t *halfpel_src_off_v = &halfpel_src_v[HALFPEL_CHROMA_WIDTH * 4 + 4]; //!< halfpel_src_v with offset (4,4)
  pixel_t halfpel_u[LCU_WIDTH * LCU_WIDTH]; //!< interpolated 2W x 2H block (u)
  pixel_t halfpel_v[LCU_WIDTH * LCU_WIDTH]; //!< interpolated 2W x 2H block (v)

  // Luma quarter-pel
    int8_t fractional_mv = (mv[0]&1) || (mv[1]&1) || (mv[0]&2) || (mv[1]&2); // either of 2 lowest bits of mv set -> mv is fractional

    if(fractional_mv) {
      int y_off_x = (mv[0]&3);
      int y_off_y = (mv[1]&3);

      int c_off_x = (mv[0]&7);
      int c_off_y = (mv[1]&7);

      int y,x;

      #define FILTER_SIZE_Y 8 //Luma filter size
      #define FILTER_SIZE_C 4 //Chroma filter size

      // Fractional luma 1/4-pel
      pixel_t qpel_src_y[(LCU_WIDTH+FILTER_SIZE_Y) * (LCU_WIDTH+FILTER_SIZE_Y)];
      pixel_t* qpel_src_off_y = &qpel_src_y[(width+FILTER_SIZE_Y)*(FILTER_SIZE_Y>>1)+(FILTER_SIZE_Y>>1)];
      pixel_t qpel_dst_y[LCU_WIDTH*LCU_WIDTH*16];

      // Fractional chroma 1/8-pel
      int width_c = width>>1;
      pixel_t octpel_src_u[((LCU_WIDTH>>1)+FILTER_SIZE_C) * ((LCU_WIDTH>>1)+FILTER_SIZE_C)];
      pixel_t* octpel_src_off_u = &octpel_src_u[(width_c+FILTER_SIZE_C)*(FILTER_SIZE_C>>1)+(FILTER_SIZE_C>>1)];
      pixel_t octpel_dst_u[(LCU_WIDTH >> 1)*(LCU_WIDTH >> 1) * 64];

      pixel_t octpel_src_v[((LCU_WIDTH >> 1) + FILTER_SIZE_C) * ((LCU_WIDTH >> 1) + FILTER_SIZE_C)];
      pixel_t* octpel_src_off_v = &octpel_src_v[(width_c + FILTER_SIZE_C)*(FILTER_SIZE_C >> 1) + (FILTER_SIZE_C >> 1)];
      pixel_t octpel_dst_v[(LCU_WIDTH >> 1)*(LCU_WIDTH >> 1) * 64];

      // Fractional luma
      extend_borders(xpos, ypos, mv[0]>>2, mv[1]>>2, state->tile->lcu_offset_x * LCU_WIDTH, state->tile->lcu_offset_y * LCU_WIDTH,
          ref->y, ref->width, ref->height, FILTER_SIZE_Y, width, width, qpel_src_y);

      filter_inter_quarterpel_luma(state->encoder_control, qpel_src_off_y, width+FILTER_SIZE_Y, width,
                                   width, qpel_dst_y, width*4, y_off_x, y_off_y);

      //Fractional chroma U
      extend_borders(xpos>>1, ypos>>1, (mv[0]>>2)>>1, (mv[1]>>2)>>1, state->tile->lcu_offset_x * (LCU_WIDTH>>1), state->tile->lcu_offset_y * (LCU_WIDTH>>1),
          ref->u, ref->width>>1, ref->height>>1, FILTER_SIZE_C, width_c, width_c, octpel_src_u);

      filter_inter_octpel_chroma(state->encoder_control, octpel_src_off_u, width_c+FILTER_SIZE_C, width_c,
                                 width_c, octpel_dst_u, width_c*8, c_off_x, c_off_y);

      //Fractional chroma V
      extend_borders(xpos>>1, ypos>>1, (mv[0]>>2)>>1, (mv[1]>>2)>>1, state->tile->lcu_offset_x * (LCU_WIDTH>>1), state->tile->lcu_offset_y * (LCU_WIDTH>>1),
          ref->v, ref->width>>1, ref->height>>1, FILTER_SIZE_C, width_c, width_c, octpel_src_v);

      filter_inter_octpel_chroma(state->encoder_control, octpel_src_off_v, width_c+FILTER_SIZE_C, width_c,
                   width_c, octpel_dst_v, width_c*8, c_off_x, c_off_y);

      //Sample fractional pixels for luma
      for(y = 0; y < width; ++y) {
        int y_in_lcu = ((y+ypos) & ((LCU_WIDTH)-1));
        int qpel_y = y*4+y_off_y;
        for(x = 0; x < width; ++x) {
          int x_in_lcu = ((x+xpos) & ((LCU_WIDTH)-1));
          int qpel_x = x*4+y_off_x;
          lcu->rec.y[y_in_lcu * LCU_WIDTH + x_in_lcu] = (pixel_t)qpel_dst_y[qpel_y*(width*4)+qpel_x];
        }
      }
      //Sample fractional pixels for chroma
      for(y = 0; y < width_c; ++y) {
        int y_in_lcu = ((y+(ypos>>1)) & ((LCU_WIDTH>>1)-1));
        int qpel_y = y*8+c_off_y;
        for(x = 0; x < width_c; ++x) {
          int x_in_lcu = ((x+(xpos>>1)) & ((LCU_WIDTH>>1)-1));
          int qpel_x = x*8+c_off_x;
          lcu->rec.u[y_in_lcu * dst_width_c + x_in_lcu] = (pixel_t)octpel_dst_u[qpel_y*(width_c*8)+qpel_x];
          lcu->rec.v[y_in_lcu * dst_width_c + x_in_lcu] = (pixel_t)octpel_dst_v[qpel_y*(width_c*8)+qpel_x];
        }
      }
    }

  mv[0] >>= 2;
  mv[1] >>= 2;

  // Chroma half-pel
  // get half-pel interpolated block and push it to output
  if(!fractional_mv) {
    if(chroma_halfpel) {
      int halfpel_y, halfpel_x;
      int abs_mv_x = mv[0]&1;
      int abs_mv_y = mv[1]&1;
      int8_t overflow_neg_y_temp,overflow_pos_y_temp,overflow_neg_x_temp,overflow_pos_x_temp;
      // Fill source blocks with data from reference, -4...width+4
      for (halfpel_y = 0, y = (ypos>>1) - 4; y < ((ypos + width)>>1) + 4; halfpel_y++, y++) {
        // calculate y-pixel offset
        coord_y = (y + state->tile->lcu_offset_y * (LCU_WIDTH>>1)) + (mv[1]>>1);

        // On y-overflow set coord_y accordingly
        overflow_neg_y_temp = (coord_y < 0) ? 1 : 0;
        overflow_pos_y_temp = (coord_y >= ref->height>>1) ? 1 : 0;
        if (overflow_neg_y_temp)      coord_y = 0;
        else if (overflow_pos_y_temp) coord_y = (ref->height>>1) - 1;
        coord_y *= ref_width_c;

        for (halfpel_x = 0, x = (xpos>>1) - 4; x < ((xpos + width)>>1) + 4; halfpel_x++, x++) {
          coord_x = (x + state->tile->lcu_offset_x * (LCU_WIDTH>>1)) + (mv[0]>>1);

          // On x-overflow set coord_x accordingly
          overflow_neg_x_temp = (coord_x < 0) ? 1 : 0;
          overflow_pos_x_temp = (coord_x >= ref_width_c) ? 1 : 0;
          if (overflow_neg_x_temp)      coord_x = 0;
          else if (overflow_pos_x_temp) coord_x = ref_width_c - 1;

          // Store source block data (with extended borders)
          halfpel_src_u[halfpel_y*HALFPEL_CHROMA_WIDTH + halfpel_x] = ref->u[coord_y + coord_x];
          halfpel_src_v[halfpel_y*HALFPEL_CHROMA_WIDTH + halfpel_x] = ref->v[coord_y + coord_x];
        }
      }

      // Filter the block to half-pel resolution
      filter_inter_halfpel_chroma(state->encoder_control, halfpel_src_off_u, HALFPEL_CHROMA_WIDTH, width>>1, width>>1, halfpel_u, LCU_WIDTH, abs_mv_x, abs_mv_y);
      filter_inter_halfpel_chroma(state->encoder_control, halfpel_src_off_v, HALFPEL_CHROMA_WIDTH, width>>1, width>>1, halfpel_v, LCU_WIDTH, abs_mv_x, abs_mv_y);

      // Assign filtered pixels to output, take every second half-pel sample with offset of abs_mv_y/x
      for (halfpel_y = abs_mv_y, y = ypos>>1; y < (ypos + width)>>1; halfpel_y += 2, y++) {
        for (halfpel_x = abs_mv_x, x = xpos>>1; x < (xpos + width)>>1; halfpel_x += 2, x++) {
          int x_in_lcu = (x & ((LCU_WIDTH>>1)-1));
          int y_in_lcu = (y & ((LCU_WIDTH>>1)-1));
          lcu->rec.u[y_in_lcu*dst_width_c + x_in_lcu] = (pixel_t)halfpel_u[halfpel_y*LCU_WIDTH + halfpel_x];
          lcu->rec.v[y_in_lcu*dst_width_c + x_in_lcu] = (pixel_t)halfpel_v[halfpel_y*LCU_WIDTH + halfpel_x];
        }
      }
    }

    // With overflow present, more checking
    if (overflow_neg_x || overflow_neg_y || overflow_pos_x || overflow_pos_y) {
      // Copy Luma with boundary checking
      for (y = ypos; y < ypos + width; y++) {
        for (x = xpos; x < xpos + width; x++) {
          int x_in_lcu = (x & ((LCU_WIDTH)-1));
          int y_in_lcu = (y & ((LCU_WIDTH)-1));

          coord_x = (x + state->tile->lcu_offset_x * LCU_WIDTH) + mv[0];
          coord_y = (y + state->tile->lcu_offset_y * LCU_WIDTH) + mv[1];
          overflow_neg_x = (coord_x < 0)?1:0;
          overflow_neg_y = (coord_y < 0)?1:0;

          overflow_pos_x = (coord_x >= ref->width )?1:0;
          overflow_pos_y = (coord_y >= ref->height)?1:0;

          // On x-overflow set coord_x accordingly
          if (overflow_neg_x) {
            coord_x = 0;
          } else if (overflow_pos_x) {
            coord_x = ref->width - 1;
          }

          // On y-overflow set coord_y accordingly
          if (overflow_neg_y) {
            coord_y = 0;
          } else if (overflow_pos_y) {
            coord_y = ref->height - 1;
          }

          // set destination to (corrected) pixel value from the reference
          lcu->rec.y[y_in_lcu * LCU_WIDTH + x_in_lcu] = ref->y[coord_y*ref->width + coord_x];
        }
      }

      if(!chroma_halfpel) {
        // Copy Chroma with boundary checking
        // TODO: chroma fractional pixel interpolation
        for (y = ypos>>1; y < (ypos + width)>>1; y++) {
          for (x = xpos>>1; x < (xpos + width)>>1; x++) {
            int x_in_lcu = (x & ((LCU_WIDTH>>1)-1));
            int y_in_lcu = (y & ((LCU_WIDTH>>1)-1));

            coord_x = (x + state->tile->lcu_offset_x * (LCU_WIDTH >> 1)) + (mv[0]>>1);
            coord_y = (y + state->tile->lcu_offset_y * (LCU_WIDTH >> 1)) + (mv[1]>>1);

            overflow_neg_x = (coord_x < 0)?1:0;
            overflow_neg_y = (y + (mv[1]>>1) < 0)?1:0;

            overflow_pos_x = (coord_x >= ref->width>>1 )?1:0;
            overflow_pos_y = (coord_y >= ref->height>>1)?1:0;

            // On x-overflow set coord_x accordingly
            if(overflow_neg_x) {
              coord_x = 0;
            } else if(overflow_pos_x) {
              coord_x = (ref->width>>1) - 1;
            }

            // On y-overflow set coord_y accordingly
            if(overflow_neg_y) {
              coord_y = 0;
            } else if(overflow_pos_y) {
              coord_y = (ref->height>>1) - 1;
            }

            // set destinations to (corrected) pixel value from the reference
            lcu->rec.u[y_in_lcu*dst_width_c + x_in_lcu] = ref->u[coord_y * ref_width_c + coord_x];
            lcu->rec.v[y_in_lcu*dst_width_c + x_in_lcu] = ref->v[coord_y * ref_width_c + coord_x];
          }
        }
      }
    } else { //If no overflow, we can copy without checking boundaries
      // Copy Luma
      for (y = ypos; y < ypos + width; y++) {
        int y_in_lcu = (y & ((LCU_WIDTH)-1));
        coord_y = ((y + state->tile->lcu_offset_y * LCU_WIDTH) + mv[1]) * ref->width; // pre-calculate
        for (x = xpos; x < xpos + width; x++) {
          int x_in_lcu = (x & ((LCU_WIDTH)-1));

          lcu->rec.y[y_in_lcu * LCU_WIDTH + x_in_lcu] = ref->y[coord_y + (x + state->tile->lcu_offset_x * LCU_WIDTH) + mv[0]];
        }
      }

      if(!chroma_halfpel) {
        // Copy Chroma
        // TODO: chroma fractional pixel interpolation
        for (y = ypos>>1; y < (ypos + width)>>1; y++) {
          int y_in_lcu = (y & ((LCU_WIDTH>>1)-1));
          coord_y = ((y + state->tile->lcu_offset_y * (LCU_WIDTH>>1)) + (mv[1]>>1)) * ref_width_c; // pre-calculate
          for (x = xpos>>1; x < (xpos + width)>>1; x++) {
            int x_in_lcu = (x & ((LCU_WIDTH>>1)-1));
            lcu->rec.u[y_in_lcu*dst_width_c + x_in_lcu] = ref->u[coord_y + (x + state->tile->lcu_offset_x * (LCU_WIDTH>>1)) + (mv[0]>>1)];
            lcu->rec.v[y_in_lcu*dst_width_c + x_in_lcu] = ref->v[coord_y + (x + state->tile->lcu_offset_x * (LCU_WIDTH>>1)) + (mv[0]>>1)];
          }
        }
      }
    }
  }
}

/**
 * \brief Get merge candidates for current block
 * \param encoder encoder control struct to use
 * \param x_cu block x position in SCU
 * \param y_cu block y position in SCU
 * \param depth current block depth
 * \param b0 candidate b0
 * \param b1 candidate b1
 * \param b2 candidate b2
 * \param a0 candidate a0
 * \param a1 candidate a1
 */
void inter_get_spatial_merge_candidates(int32_t x, int32_t y, int8_t depth, cu_info_t **b0, cu_info_t **b1,
                                        cu_info_t **b2,cu_info_t **a0,cu_info_t **a1, lcu_t *lcu)
{
  uint8_t cur_block_in_scu = (LCU_WIDTH>>depth) / CU_MIN_SIZE_PIXELS; //!< the width of the current block on SCU
  /*
  Predictor block locations
  ____      _______
  |B2|______|B1|B0|
     |         |
     |  Cur CU |
   __|         |
  |A1|_________|
  |A0|
  */
  int32_t x_cu = (x & (LCU_WIDTH - 1)) >> MAX_DEPTH; //!< coordinates from top-left of this LCU
  int32_t y_cu = (y & (LCU_WIDTH - 1)) >> MAX_DEPTH;
  cu_info_t* cu = &lcu->cu[LCU_CU_OFFSET];
  // A0 and A1 availability testing
  if (x != 0) {
    *a1 = &cu[x_cu - 1 + (y_cu + cur_block_in_scu - 1) * LCU_T_CU_WIDTH];
    if (!(*a1)->coded) *a1 = NULL;

    if (y_cu + cur_block_in_scu < LCU_WIDTH>>3) {
      *a0 = &cu[x_cu - 1 + (y_cu + cur_block_in_scu) * LCU_T_CU_WIDTH];
      if (!(*a0)->coded) *a0 = NULL;
    }
  }

  // B0, B1 and B2 availability testing
  if (y != 0) {
    if (x_cu + cur_block_in_scu < LCU_WIDTH>>3) {
      *b0 = &cu[x_cu + cur_block_in_scu + (y_cu - 1) * LCU_T_CU_WIDTH];
      if (!(*b0)->coded) *b0 = NULL;
    } else if(y_cu == 0) {
      // Special case, top-right cu from LCU is the last in lcu->cu array
      *b0 = &lcu->cu[LCU_T_CU_WIDTH*LCU_T_CU_WIDTH];
      if (!(*b0)->coded) *b0 = NULL;
    }

    *b1 = &cu[x_cu + cur_block_in_scu - 1 + (y_cu - 1) * LCU_T_CU_WIDTH];
    if (!(*b1)->coded) *b1 = NULL;

    if (x != 0) {
      *b2 = &cu[x_cu - 1 + (y_cu - 1) * LCU_T_CU_WIDTH];
      if(!(*b2)->coded) *b2 = NULL;
    }
  }
}

/**
 * \brief Get MV prediction for current block
 * \param encoder encoder control struct to use
 * \param x_cu block x position in SCU
 * \param y_cu block y position in SCU
 * \param depth current block depth
 * \param mv_pred[2][2] 2x motion vector prediction
 */
void inter_get_mv_cand(const encoder_state_t * const state, int32_t x, int32_t y, int8_t depth, int16_t mv_cand[2][2], cu_info_t* cur_cu, lcu_t *lcu)
{
  uint8_t candidates = 0;
  uint8_t b_candidates = 0;

  cu_info_t *b0, *b1, *b2, *a0, *a1;
  b0 = b1 = b2 = a0 = a1 = NULL;
  inter_get_spatial_merge_candidates(x, y, depth, &b0, &b1, &b2, &a0, &a1, lcu);

 #define CALCULATE_SCALE(cu,tb,td) ((tb * ((0x4000 + (abs(td)>>1))/td) + 32) >> 6)
#define APPLY_MV_SCALING(cu, cand) {int td = state->global->poc - state->global->ref->images[(cu)->inter.mv_ref]->poc;\
                                   int tb = state->global->poc - state->global->ref->images[cur_cu->inter.mv_ref]->poc;\
                                   if (td != tb) { \
                                      int scale = CALCULATE_SCALE(cu,tb,td); \
                                       mv_cand[cand][0] = ((scale * (cu)->inter.mv[0] + 127 + (scale * (cu)->inter.mv[0] < 0)) >> 8 ); \
                                       mv_cand[cand][1] = ((scale * (cu)->inter.mv[1] + 127 + (scale * (cu)->inter.mv[1] < 0)) >> 8 ); }}

  // Left predictors
  if (a0 && a0->type == CU_INTER && a0->inter.mv_ref == cur_cu->inter.mv_ref) {
    mv_cand[candidates][0] = a0->inter.mv[0];
    mv_cand[candidates][1] = a0->inter.mv[1];
    candidates++;
  } else if (a1 && a1->type == CU_INTER && a1->inter.mv_ref == cur_cu->inter.mv_ref) {
    mv_cand[candidates][0] = a1->inter.mv[0];
    mv_cand[candidates][1] = a1->inter.mv[1];
    candidates++;
  }

  if(!candidates) {
      // Left predictors
    if (a0 && a0->type == CU_INTER) {
      mv_cand[candidates][0] = a0->inter.mv[0];
      mv_cand[candidates][1] = a0->inter.mv[1];
      APPLY_MV_SCALING(a0, candidates);
      candidates++;
    } else if (a1 && a1->type == CU_INTER) {
      mv_cand[candidates][0] = a1->inter.mv[0];
      mv_cand[candidates][1] = a1->inter.mv[1];
      APPLY_MV_SCALING(a1, candidates);
      candidates++;
    }
  }

  // Top predictors
  if (b0 && b0->type == CU_INTER && b0->inter.mv_ref == cur_cu->inter.mv_ref) {
    mv_cand[candidates][0] = b0->inter.mv[0];
    mv_cand[candidates][1] = b0->inter.mv[1];
    b_candidates++;
  } else if (b1 && b1->type == CU_INTER && b1->inter.mv_ref == cur_cu->inter.mv_ref) {
    mv_cand[candidates][0] = b1->inter.mv[0];
    mv_cand[candidates][1] = b1->inter.mv[1];
    b_candidates++;
  } else if(b2 && b2->type == CU_INTER && b2->inter.mv_ref == cur_cu->inter.mv_ref) {
    mv_cand[candidates][0] = b2->inter.mv[0];
    mv_cand[candidates][1] = b2->inter.mv[1];
    b_candidates++;
  }
  candidates += b_candidates;

  // When a1 or a0 is available, we dont check for secondary B candidates
  if((a1 && a1->type == CU_INTER) || (a0 && a0->type == CU_INTER)) {
    b_candidates = 1;
  } else if(candidates != 2) {
    b_candidates = 0;
  }

  if(!b_candidates) {
    // Top predictors
    if (b0 && b0->type == CU_INTER) {
      mv_cand[candidates][0] = b0->inter.mv[0];
      mv_cand[candidates][1] = b0->inter.mv[1];
      APPLY_MV_SCALING(b0, candidates);
      candidates++;
    } else if (b1 && b1->type == CU_INTER) {
      mv_cand[candidates][0] = b1->inter.mv[0];
      mv_cand[candidates][1] = b1->inter.mv[1];
      APPLY_MV_SCALING(b1, candidates);
      candidates++;
    } else if(b2 && b2->type == CU_INTER) {
      mv_cand[candidates][0] = b2->inter.mv[0];
      mv_cand[candidates][1] = b2->inter.mv[1];
      APPLY_MV_SCALING(b2, candidates);
      candidates++;
    }
  }

  // Remove identical candidate
  if(candidates == 2 && mv_cand[0][0] == mv_cand[1][0] && mv_cand[0][1] == mv_cand[1][1]) {
    candidates = 1;
  }

#if ENABLE_TEMPORAL_MVP
  if(candidates < AMVP_MAX_NUM_CANDS) {
    //TODO: add temporal mv predictor
  }
#endif

  // Fill with (0,0)
  while (candidates < AMVP_MAX_NUM_CANDS) {
    mv_cand[candidates][0] = 0;
    mv_cand[candidates][1] = 0;
    candidates++;
  }
#undef CALCULATE_SCALE
#undef APPLY_MV_SCALING
}

/**
 * \brief Get merge predictions for current block
 * \param encoder encoder control struct to use
 * \param x_cu block x position in SCU
 * \param y_cu block y position in SCU
 * \param depth current block depth
 * \param mv_pred[MRG_MAX_NUM_CANDS][2] MRG_MAX_NUM_CANDS motion vector prediction
 */
uint8_t inter_get_merge_cand(int32_t x, int32_t y, int8_t depth, int16_t mv_cand[MRG_MAX_NUM_CANDS][3], lcu_t *lcu)
{
  uint8_t candidates = 0;
  int8_t duplicate = 0;

  cu_info_t *b0, *b1, *b2, *a0, *a1;
  int8_t zero_idx = 0;
  b0 = b1 = b2 = a0 = a1 = NULL;
  inter_get_spatial_merge_candidates(x, y, depth, &b0, &b1, &b2, &a0, &a1, lcu);


#define CHECK_DUPLICATE(CU1,CU2) {duplicate = 0; if ((CU2) && (CU2)->type == CU_INTER && \
                                                     (CU1)->inter.mv[0] == (CU2)->inter.mv[0] && \
                                                     (CU1)->inter.mv[1] == (CU2)->inter.mv[1] && \
                                                     (CU1)->inter.mv_ref == (CU2)->inter.mv_ref) duplicate = 1; }

  if (a1 && a1->type == CU_INTER) {
      mv_cand[candidates][0] = a1->inter.mv[0];
      mv_cand[candidates][1] = a1->inter.mv[1];
      mv_cand[candidates][2] = a1->inter.mv_ref;
      candidates++;
  }

  if (b1 && b1->type == CU_INTER) {
    if(candidates) CHECK_DUPLICATE(b1, a1);
    if(!duplicate) {
      mv_cand[candidates][0] = b1->inter.mv[0];
      mv_cand[candidates][1] = b1->inter.mv[1];
      mv_cand[candidates][2] = b1->inter.mv_ref;
      candidates++;
    }
  }

  if (b0 && b0->type == CU_INTER) {
    if(candidates) CHECK_DUPLICATE(b0,b1);
    if(!duplicate) {
      mv_cand[candidates][0] = b0->inter.mv[0];
      mv_cand[candidates][1] = b0->inter.mv[1];
      mv_cand[candidates][2] = b0->inter.mv_ref;
      candidates++;
    }
  }

  if (a0 && a0->type == CU_INTER) {
    if(candidates) CHECK_DUPLICATE(a0,a1);
    if(!duplicate) {
      mv_cand[candidates][0] = a0->inter.mv[0];
      mv_cand[candidates][1] = a0->inter.mv[1];
      mv_cand[candidates][2] = a0->inter.mv_ref;
      candidates++;
    }
  }

  if (candidates != 4) {
    if(b2 && b2->type == CU_INTER) {
      CHECK_DUPLICATE(b2,a1);
      if(!duplicate) {
        CHECK_DUPLICATE(b2,b1);
        if(!duplicate) {
          mv_cand[candidates][0] = b2->inter.mv[0];
          mv_cand[candidates][1] = b2->inter.mv[1];
          mv_cand[candidates][2] = b2->inter.mv_ref;
          candidates++;
        }
      }
    }
  }


#if ENABLE_TEMPORAL_MVP
  if(candidates < AMVP_MAX_NUM_CANDS) {
    //TODO: add temporal mv predictor
  }
#endif

  // Add (0,0) prediction
  if (candidates != 5) {
    mv_cand[candidates][0] = 0;
    mv_cand[candidates][1] = 0;
    mv_cand[candidates][2] = zero_idx;
    zero_idx++;
    candidates++;
  }

  return candidates;
}