uvg266/src/intra.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
 * \brief Functions for handling intra frames.
 */

#include "intra.h"

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

#include "config.h"
#include "encoder.h"
#include "transform.h"
#include "rdo.h"


/**
 * \brief Function for deriving intra luma predictions
 * \param pic picture to use
 * \param x_cu x CU position (smallest CU)
 * \param y_cu y CU position (smallest CU)
 * \param preds output buffer for 3 predictions
 * \returns (predictions are found)?1:0
 */
int8_t kvz_intra_get_dir_luma_predictor(const uint32_t x, const uint32_t y, int8_t* preds,
                                    const cu_info_t * const cur_cu, const cu_info_t * const left_cu, const cu_info_t * const above_cu)
{
  int y_cu = y>>3;

  // The default mode if block is not coded yet is INTRA_DC.
  int8_t left_intra_dir  = 1;
  int8_t above_intra_dir = 1;

  if (x & 4) {
    // If current CU is NxN and PU is on the right half, take mode from the
    // left half of the same CU.
    left_intra_dir = cur_cu->intra[PU_INDEX(0, y >> 2)].mode;
  } else if (left_cu && left_cu->type == CU_INTRA) {
    // Otherwise take the mode from the right side of the CU on the left.
    left_intra_dir = left_cu->intra[PU_INDEX(1, y >> 2)].mode;
  }

  if (y & 4) {
    // If current CU is NxN and PU is on the bottom half, take mode from the
    // top half of the same CU.
    above_intra_dir = cur_cu->intra[PU_INDEX(x >> 2, 0)].mode;
  } else if (above_cu && above_cu->type == CU_INTRA &&
             (y_cu * (LCU_WIDTH>>MAX_DEPTH)) % LCU_WIDTH != 0)
  {
    // Otherwise take the mode from the bottom half of the CU above.
    above_intra_dir = above_cu->intra[PU_INDEX(x >> 2, 1)].mode;
  }

  // If the predictions are the same, add new predictions
  if (left_intra_dir == above_intra_dir) {
    if (left_intra_dir > 1) { // angular modes
      preds[0] = left_intra_dir;
      preds[1] = ((left_intra_dir + 29) % 32) + 2;
      preds[2] = ((left_intra_dir - 1 ) % 32) + 2;
    } else { //non-angular
      preds[0] = 0;//PLANAR_IDX;
      preds[1] = 1;//DC_IDX;
      preds[2] = 26;//VER_IDX;
    }
  } else { // If we have two distinct predictions
    preds[0] = left_intra_dir;
    preds[1] = above_intra_dir;

    // add planar mode if it's not yet present
    if (left_intra_dir && above_intra_dir ) {
      preds[2] = 0; // PLANAR_IDX;
    } else {  // Add DC mode if it's not present, otherwise 26.
      preds[2] =  (left_intra_dir+above_intra_dir)<2? 26 : 1;
    }
  }

  return 1;
}


static void intra_filter_reference(int_fast8_t log2_width, kvz_intra_references *refs)
{
  if (refs->filtered_initialized) {
    return;
  } else {
    refs->filtered_initialized = true;
  }

  const int_fast8_t ref_width = 2 * (1 << log2_width) + 1;
  kvz_intra_ref *ref = &refs->ref;
  kvz_intra_ref *filtered_ref = &refs->filtered_ref;

  filtered_ref->left[0] = (ref->left[1] + 2 * ref->left[0] + ref->top[1] + 2) / 4;
  filtered_ref->top[0] = filtered_ref->left[0];

  for (int_fast8_t y = 1; y < ref_width - 1; ++y) {
    kvz_pixel *p = &ref->left[y];
    filtered_ref->left[y] = (p[-1] + 2 * p[0] + p[1] + 2) / 4;
  }
  filtered_ref->left[ref_width - 1] = ref->left[ref_width - 1];

  for (int_fast8_t x = 1; x < ref_width - 1; ++x) {
    kvz_pixel *p = &ref->top[x];
    filtered_ref->top[x] = (p[-1] + 2 * p[0] + p[1] + 2) / 4;
  }
  filtered_ref->top[ref_width - 1] = ref->top[ref_width - 1];
}


static void post_process_intra_angular(
  unsigned width,
  unsigned stride,
  const kvz_pixel *ref,
  kvz_pixel *block)
{
  kvz_pixel ref2 = ref[0];
  for (unsigned i = 0; i < width; i++) {
    kvz_pixel val = block[i * stride];
    kvz_pixel ref1 = ref[i + 1];
    block[i * stride] = CLIP_TO_PIXEL(val + ((ref1 - ref2) >> 1));
  }
}


/**
 * \brief Generage angular predictions.
 * \param log2_width    Log2 of width, range 2..5.
 * \param intra_mode    Angular mode in range 2..34.
 * \param in_ref_above  Pointer to -1 index of above reference, length=width*2+1.
 * \param in_ref_left   Pointer to -1 index of left reference, length=width*2+1.
 * \param dst           Buffer of size width*width.
 */
static void kvz_intra_pred_angular(
  const int_fast8_t log2_width,
  const int_fast8_t intra_mode,
  const kvz_pixel *const in_ref_above,
  const kvz_pixel *const in_ref_left,
  kvz_pixel *const dst)
{
  assert(log2_width >= 2 && log2_width <= 5);
  assert(intra_mode >= 2 && intra_mode <= 34);

  static const int8_t modedisp2sampledisp[9] = {0, 2, 5, 9, 13, 17, 21, 26, 32};
  static const int16_t modedisp2invsampledisp[9] = {0, 4096, 1638, 910, 630, 482, 390, 315, 256}; // (256 * 32) / sampledisp

  // Temporary buffer for modes 11-25.
  // It only needs to be big enough to hold indices from -width to width-1.
  kvz_pixel tmp_ref[2 * 32];
  const int_fast8_t width = 1 << log2_width;

  // Whether to swap references to always project on the left reference row.
  const bool vertical_mode = intra_mode >= 18;
  // Modes distance to horizontal or vertical mode.
  const int_fast8_t mode_disp = vertical_mode ? intra_mode - 26 : 10 - intra_mode;
  // Sample displacement per column in fractions of 32.
  const int_fast8_t sample_disp = (mode_disp < 0 ? -1 : 1) * modedisp2sampledisp[abs(mode_disp)];

  // Pointer for the reference we are interpolating from.
  const kvz_pixel *ref_main;
  // Pointer for the other reference.
  const kvz_pixel *ref_side;

  // Set ref_main and ref_side such that, when indexed with 0, they point to
  // index 0 in block coordinates.
  if (sample_disp < 0) {
    // Negative sample_disp means, we need to use both references.

    ref_side = (vertical_mode ? in_ref_left : in_ref_above) + 1;
    ref_main = (vertical_mode ? in_ref_above : in_ref_left) + 1;

    // Move the reference pixels to start from the middle to the later half of
    // the tmp_ref, so there is room for negative indices.
    for (int_fast8_t x = -1; x < width; ++x) {
      tmp_ref[x + width] = ref_main[x];
    }
    // Get a pointer to block index 0 in tmp_ref.
    ref_main = &tmp_ref[width];

    // Extend the side reference to the negative indices of main reference.
    int_fast32_t col_sample_disp = 128; // rounding for the ">> 8"
    int_fast16_t inv_abs_sample_disp = modedisp2invsampledisp[abs(mode_disp)];
    int_fast8_t most_negative_index = (width * sample_disp) >> 5;
    for (int_fast8_t x = -2; x >= most_negative_index; --x) {
      col_sample_disp += inv_abs_sample_disp;
      int_fast8_t side_index = col_sample_disp >> 8;
      tmp_ref[x + width] = ref_side[side_index - 1];
    }
  } else {
    // sample_disp >= 0 means we don't need to refer to negative indices,
    // which means we can just use the references as is.
    ref_main = (vertical_mode ? in_ref_above : in_ref_left) + 1;
    ref_side = (vertical_mode ? in_ref_left : in_ref_above) + 1;
  }

  if (sample_disp != 0) {
    // The mode is not horizontal or vertical, we have to do interpolation.

    int_fast16_t delta_pos = 0;
    for (int_fast8_t y = 0; y < width; ++y) {
      delta_pos += sample_disp;
      int_fast8_t delta_int = delta_pos >> 5;
      int_fast8_t delta_fract = delta_pos & (32 - 1);

      if (delta_fract) {
        // Do linear filtering
        for (int_fast8_t x = 0; x < width; ++x) {
          kvz_pixel ref1 = ref_main[x + delta_int];
          kvz_pixel ref2 = ref_main[x + delta_int + 1];
          dst[y * width + x] = ((32 - delta_fract) * ref1 + delta_fract * ref2 + 16) >> 5;
        }
      } else {
        // Just copy the integer samples
        for (int_fast8_t x = 0; x < width; x++) {
          dst[y * width + x] = ref_main[x + delta_int];
        }
      }
    }
  } else {
    // Mode is horizontal or vertical, just copy the pixels.

    for (int_fast8_t y = 0; y < width; ++y) {
      for (int_fast8_t x = 0; x < width; ++x) {
        dst[y * width + x] = ref_main[x];
      }
    }
  }

  // Flip the block if this is was a horizontal mode.
  if (!vertical_mode) {
    for (int_fast8_t y = 0; y < width - 1; ++y) {
      for (int_fast8_t x = y + 1; x < width; ++x) {
        SWAP(dst[y * width + x], dst[x * width + y], kvz_pixel);
      }
    }
  }
}


/**
 * \brief Generage planar prediction.
 * \param log2_width    Log2 of width, range 2..5.
 * \param in_ref_above  Pointer to -1 index of above reference, length=width*2+1.
 * \param in_ref_left   Pointer to -1 index of left reference, length=width*2+1.
 * \param dst           Buffer of size width*width.
 */
static void kvz_intra_pred_planar(
  const int_fast8_t log2_width,
  const kvz_pixel *const ref_top,
  const kvz_pixel *const ref_left,
  kvz_pixel *const dst)
{
  assert(log2_width >= 2 && log2_width <= 5);

  const int_fast8_t width = 1 << log2_width;
  const kvz_pixel top_right = ref_top[width + 1];
  const kvz_pixel bottom_left = ref_left[width + 1];

#if 0
  // Unoptimized version for reference.
  for (int y = 0; y < width; ++y) {
    for (int x = 0; x < width; ++x) {
      int_fast16_t hor = (width - 1 - x) * ref_left[y + 1] + (x + 1) * top_right;
      int_fast16_t ver = (width - 1 - y) * ref_top[x + 1] + (y + 1) * bottom_left;
      dst[y * width + x] = (ver + hor + width) >> (log2_width + 1);
    }
  }
#else
  int_fast16_t top[32];
  for (int i = 0; i < width; ++i) {
    top[i] = ref_top[i + 1] << log2_width;
  }

  for (int y = 0; y < width; ++y) {
    int_fast16_t hor = (ref_left[y + 1] << log2_width) + width;
    for (int x = 0; x < width; ++x) {
      hor += top_right - ref_left[y + 1];
      top[x] += bottom_left - ref_top[x + 1];
      dst[y * width + x] = (hor + top[x]) >> (log2_width + 1);
    }
  }
#endif
}


/**
* \brief Generage planar prediction.
* \param log2_width    Log2 of width, range 2..5.
* \param in_ref_above  Pointer to -1 index of above reference, length=width*2+1.
* \param in_ref_left   Pointer to -1 index of left reference, length=width*2+1.
* \param dst           Buffer of size width*width.
*/
static void kvz_intra_pred_dc(
  const int_fast8_t log2_width,
  const kvz_pixel *const ref_top,
  const kvz_pixel *const ref_left,
  kvz_pixel *const out_block)
{
  int_fast8_t width = 1 << log2_width;

  int_fast16_t sum = 0;
  for (int_fast8_t i = 0; i < width; ++i) {
    sum += ref_top[i + 1];
    sum += ref_left[i + 1];
  }

  const kvz_pixel dc_val = (sum + width) >> (log2_width + 1);
  const int_fast16_t block_size = 1 << (log2_width * 2);

  for (int_fast16_t i = 0; i < block_size; ++i) {
    out_block[i] = dc_val;
  }
}


/**
* \brief Generage intra DC prediction with post filtering applied.
* \param log2_width    Log2 of width, range 2..5.
* \param in_ref_above  Pointer to -1 index of above reference, length=width*2+1.
* \param in_ref_left   Pointer to -1 index of left reference, length=width*2+1.
* \param dst           Buffer of size width*width.
*/
static void kvz_intra_pred_filtered_dc(
  const int_fast8_t log2_width,
  const kvz_pixel *const ref_top,
  const kvz_pixel *const ref_left,
  kvz_pixel *const out_block)
{
  assert(log2_width >= 2 && log2_width <= 5);

  const int_fast8_t width = 1 << log2_width;

  int_fast16_t sum = 0;
  for (int_fast8_t i = 0; i < width; ++i) {
    sum += ref_top[i + 1];
    sum += ref_left[i + 1];
  }

  const kvz_pixel dc_val = (sum + width) >> (log2_width + 1);

  // Filter top-left with ([1 2 1] / 4)
  out_block[0] = (ref_left[1] + 2 * dc_val + ref_top[1] + 2) / 4;

  // Filter rest of the boundary with ([1 3] / 4)
  for (int_fast8_t x = 1; x < width; ++x) {
    out_block[x] = (ref_top[x + 1] + 3 * dc_val + 2) / 4;
  }
  for (int_fast8_t y = 1; y < width; ++y) {
    out_block[y * width] = (ref_left[y + 1] + 3 * dc_val + 2) / 4;
    for (int_fast8_t x = 1; x < width; ++x) {
      out_block[y * width + x] = dc_val;
    }
  }
}


void kvz_intra_get_pred_new(
  kvz_intra_references *refs,
  int_fast8_t log2_width,
  int_fast8_t mode,
  color_t color,
  kvz_pixel *dst)
{
  const int_fast8_t width = 1 << log2_width;

  const kvz_intra_ref *used_ref = &refs->ref;
  if (color != COLOR_Y || mode == 1 || width == 4) {
    // For chroma, DC and 4x4 blocks, always use unfiltered reference.
  } else if (mode == 0) {
    // Otherwise, use filtered for planar.
    used_ref = &refs->filtered_ref;
  } else {
    // Angular modes use smoothed reference pixels, unless the mode is close
    // to being either vertical or horizontal.
    static const int kvz_intra_hor_ver_dist_thres[5] = { 0, 7, 1, 0, 0 };
    int filter_threshold = kvz_intra_hor_ver_dist_thres[g_to_bits[width]];
    int dist_from_vert_or_hor = MIN(abs(mode - 26), abs(mode - 10));
    if (dist_from_vert_or_hor > filter_threshold) {
      used_ref = &refs->filtered_ref;
    }
  }

  if (used_ref == &refs->filtered_ref && !refs->filtered_initialized) {
    intra_filter_reference(log2_width, refs);
  }

  if (mode == 0) {
    kvz_intra_pred_planar(log2_width, used_ref->top, used_ref->left, dst);
  } else if (mode == 1) {
    // Do extra post filtering for edge pixels of luma DC mode.
    if (color == COLOR_Y && width < 32) {
      kvz_intra_pred_filtered_dc(log2_width, used_ref->top, used_ref->left, dst);
    } else {
      kvz_intra_pred_dc(log2_width, used_ref->top, used_ref->left, dst);
    }
  } else {
    kvz_intra_pred_angular(log2_width, mode, used_ref->top, used_ref->left, dst);
    if (color == COLOR_Y && width < 32) {
      if (mode == 10) {
        post_process_intra_angular(width, 1, used_ref->top, dst);
      } else if (mode == 26) {
        post_process_intra_angular(width, width, used_ref->left, dst);
      }
    }
  }
}


void kvz_intra_build_reference(
  const int_fast8_t log2_width,
  const color_t color,
  const vector2d_t *const luma_px,
  const vector2d_t *const pic_px,
  const lcu_t *const lcu,
  kvz_intra_references *const refs)
{
  assert(log2_width >= 2 && log2_width <= 5);

  // Tables for looking up the number of intra reference pixels based on
  // prediction units coordinate within an LCU.
  // generated by "tools/generate_ref_pixel_tables.py".
  static const uint8_t num_ref_pixels_top[16][16] = {
    { 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64 },
    {  8,  4,  8,  4,  8,  4,  8,  4,  8,  4,  8,  4,  8,  4,  8,  4 },
    { 16, 12,  8,  4, 16, 12,  8,  4, 16, 12,  8,  4, 16, 12,  8,  4 },
    {  8,  4,  8,  4,  8,  4,  8,  4,  8,  4,  8,  4,  8,  4,  8,  4 },
    { 32, 28, 24, 20, 16, 12,  8,  4, 32, 28, 24, 20, 16, 12,  8,  4 },
    {  8,  4,  8,  4,  8,  4,  8,  4,  8,  4,  8,  4,  8,  4,  8,  4 },
    { 16, 12,  8,  4, 16, 12,  8,  4, 16, 12,  8,  4, 16, 12,  8,  4 },
    {  8,  4,  8,  4,  8,  4,  8,  4,  8,  4,  8,  4,  8,  4,  8,  4 },
    { 64, 60, 56, 52, 48, 44, 40, 36, 32, 28, 24, 20, 16, 12,  8,  4 },
    {  8,  4,  8,  4,  8,  4,  8,  4,  8,  4,  8,  4,  8,  4,  8,  4 },
    { 16, 12,  8,  4, 16, 12,  8,  4, 16, 12,  8,  4, 16, 12,  8,  4 },
    {  8,  4,  8,  4,  8,  4,  8,  4,  8,  4,  8,  4,  8,  4,  8,  4 },
    { 32, 28, 24, 20, 16, 12,  8,  4, 32, 28, 24, 20, 16, 12,  8,  4 },
    {  8,  4,  8,  4,  8,  4,  8,  4,  8,  4,  8,  4,  8,  4,  8,  4 },
    { 16, 12,  8,  4, 16, 12,  8,  4, 16, 12,  8,  4, 16, 12,  8,  4 },
    {  8,  4,  8,  4,  8,  4,  8,  4,  8,  4,  8,  4,  8,  4,  8,  4 }
  };
  static const uint8_t num_ref_pixels_left[16][16] = {
    { 64,  4,  8,  4, 16,  4,  8,  4, 32,  4,  8,  4, 16,  4,  8,  4 },
    { 60,  4,  4,  4, 12,  4,  4,  4, 28,  4,  4,  4, 12,  4,  4,  4 },
    { 56,  4,  8,  4,  8,  4,  8,  4, 24,  4,  8,  4,  8,  4,  8,  4 },
    { 52,  4,  4,  4,  4,  4,  4,  4, 20,  4,  4,  4,  4,  4,  4,  4 },
    { 48,  4,  8,  4, 16,  4,  8,  4, 16,  4,  8,  4, 16,  4,  8,  4 },
    { 44,  4,  4,  4, 12,  4,  4,  4, 12,  4,  4,  4, 12,  4,  4,  4 },
    { 40,  4,  8,  4,  8,  4,  8,  4,  8,  4,  8,  4,  8,  4,  8,  4 },
    { 36,  4,  4,  4,  4,  4,  4,  4,  4,  4,  4,  4,  4,  4,  4,  4 },
    { 32,  4,  8,  4, 16,  4,  8,  4, 32,  4,  8,  4, 16,  4,  8,  4 },
    { 28,  4,  4,  4, 12,  4,  4,  4, 28,  4,  4,  4, 12,  4,  4,  4 },
    { 24,  4,  8,  4,  8,  4,  8,  4, 24,  4,  8,  4,  8,  4,  8,  4 },
    { 20,  4,  4,  4,  4,  4,  4,  4, 20,  4,  4,  4,  4,  4,  4,  4 },
    { 16,  4,  8,  4, 16,  4,  8,  4, 16,  4,  8,  4, 16,  4,  8,  4 },
    { 12,  4,  4,  4, 12,  4,  4,  4, 12,  4,  4,  4, 12,  4,  4,  4 },
    { 8,  4,  8,  4,  8,  4,  8,  4,  8,  4,  8,  4,  8,  4,  8,  4 },
    { 4,  4,  4,  4,  4,  4,  4,  4,  4,  4,  4,  4,  4,  4,  4,  4 }
  };

  refs->filtered_initialized = false;
  kvz_pixel *out_left_ref = &refs->ref.left[0];
  kvz_pixel *out_top_ref = &refs->ref.top[0];

  const kvz_pixel dc_val = 1 << (KVZ_BIT_DEPTH - 1);
  const int is_chroma = color != COLOR_Y ? 1 : 0;
  const int_fast8_t width = 1 << log2_width;

  // Convert luma coordinates to chroma coordinates for chroma.
  const vector2d_t lcu_px = {
    luma_px->x % LCU_WIDTH,
    luma_px->y % LCU_WIDTH
  };
  const vector2d_t px = {
    lcu_px.x >> is_chroma,
    lcu_px.y >> is_chroma,
  };

  // Init pointers to LCUs reconstruction buffers, such that index 0 refers to block coordinate 0.
  const kvz_pixel *left_ref = !color ? &lcu->left_ref.y[1] : (color == 1) ? &lcu->left_ref.u[1] : &lcu->left_ref.v[1];
  const kvz_pixel *top_ref = !color ? &lcu->top_ref.y[1] : (color == 1) ? &lcu->top_ref.u[1] : &lcu->top_ref.v[1];
  const kvz_pixel *rec_ref = !color ? lcu->rec.y : (color == 1) ? lcu->rec.u : lcu->rec.v;

  // Init top borders pointer to point to the correct place in the correct reference array.
  const kvz_pixel *top_border;
  if (px.y) {
    top_border = &rec_ref[px.x + (px.y - 1) * (LCU_WIDTH >> is_chroma)];
  } else {
    top_border = &top_ref[px.x];
  }

  // Init left borders pointer to point to the correct place in the correct reference array.
  const kvz_pixel *left_border;
  int left_stride; // Distance between reference samples.
  if (px.x) {
    left_border = &rec_ref[px.x - 1 + px.y * (LCU_WIDTH >> is_chroma)];
    left_stride = LCU_WIDTH >> is_chroma;
  } else {
    left_border = &left_ref[px.y];
    left_stride = 1;
  }

  // Generate left reference.
  if (luma_px->x > 0) {
    // Get the number of reference pixels based on the PU coordinate within the LCU.
    int px_available_left = num_ref_pixels_left[lcu_px.y / 4][lcu_px.x / 4] >> is_chroma;

    // Limit the number of available pixels based on block size and dimensions
    // of the picture.
    px_available_left = MIN(px_available_left, width * 2);
    px_available_left = MIN(px_available_left, (pic_px->y - luma_px->y) >> is_chroma);

    // Copy pixels from coded CUs.
    for (int i = 0; i < px_available_left; ++i) {
      out_left_ref[i + 1] = left_border[i * left_stride];
    }
    // Extend the last pixel for the rest of the reference values.
    kvz_pixel nearest_pixel = out_left_ref[px_available_left];
    for (int i = px_available_left; i < width * 2; ++i) {
      out_left_ref[i + 1] = nearest_pixel;
    }
  } else {
    // If we are on the left edge, extend the first pixel of the top row.
    kvz_pixel nearest_pixel = luma_px->y > 0 ? top_border[0] : dc_val;
    for (int i = 0; i < width * 2; i++) {
      out_left_ref[i + 1] = nearest_pixel;
    }
  }

  // Generate top-left reference.
  if (luma_px->x > 0 && luma_px->y > 0) {
    // If the block is at an LCU border, the top-left must be copied from
    // the border that points to the LCUs 1D reference buffer.
    if (px.x == 0) {
      out_left_ref[0] = left_border[-1 * left_stride];
      out_top_ref[0] = left_border[-1 * left_stride];
    } else {
      out_left_ref[0] = top_border[-1];
      out_top_ref[0] = top_border[-1];
    }
  } else {
    // Copy reference clockwise.
    out_left_ref[0] = out_left_ref[1];
    out_top_ref[0] = out_left_ref[1];
  }

  // Generate top reference.
  if (luma_px->y > 0) {
    // Get the number of reference pixels based on the PU coordinate within the LCU.
    int px_available_top = num_ref_pixels_top[lcu_px.y / 4][lcu_px.x / 4] >> is_chroma;

    // Limit the number of available pixels based on block size and dimensions
    // of the picture.
    px_available_top = MIN(px_available_top, width * 2);
    px_available_top = MIN(px_available_top, (pic_px->x - luma_px->x) >> is_chroma);

    // Copy all the pixels we can.
    for (int i = 0; i < px_available_top; ++i) {
      out_top_ref[i + 1] = top_border[i];
    }
    // Extend the last pixel for the rest of the reference values.
    kvz_pixel nearest_pixel = top_border[px_available_top - 1];
    for (int i = px_available_top; i < width * 2; ++i) {
      out_top_ref[i + 1] = nearest_pixel;
    }
  } else {
    // Extend nearest pixel.
    kvz_pixel nearest_pixel = luma_px->x > 0 ? left_border[0] : dc_val;
    for (int i = 0; i < width * 2; i++) {
      out_top_ref[i + 1] = nearest_pixel;
    }
  }
}


void kvz_intra_recon_lcu_luma(encoder_state_t * const state, int x, int y, int depth, int8_t intra_mode, cu_info_t *cur_cu, lcu_t *lcu)
{
  const vector2d_t lcu_px = { x & 0x3f, y & 0x3f };
  if (cur_cu == NULL) {
    cur_cu = &lcu->cu[LCU_CU_OFFSET + (lcu_px.x >> 3) + (lcu_px.y >> 3)*LCU_T_CU_WIDTH];
  }
  const int8_t width = LCU_WIDTH >> depth;

  if (depth == 0 || cur_cu->tr_depth > depth) {
    int offset = width / 2;

    kvz_intra_recon_lcu_luma(state, x,          y,          depth+1, intra_mode, NULL, lcu);
    kvz_intra_recon_lcu_luma(state, x + offset, y,          depth+1, intra_mode, NULL, lcu);
    kvz_intra_recon_lcu_luma(state, x,          y + offset, depth+1, intra_mode, NULL, lcu);
    kvz_intra_recon_lcu_luma(state, x + offset, y + offset, depth+1, intra_mode, NULL, lcu);

    if (depth < MAX_DEPTH) {
      cu_info_t *cu_a = &lcu->cu[LCU_CU_OFFSET + ((lcu_px.x + offset) >> 3) + (lcu_px.y >> 3)        *LCU_T_CU_WIDTH];
      cu_info_t *cu_b = &lcu->cu[LCU_CU_OFFSET + (lcu_px.x >> 3) + ((lcu_px.y + offset) >> 3)*LCU_T_CU_WIDTH];
      cu_info_t *cu_c = &lcu->cu[LCU_CU_OFFSET + ((lcu_px.x + offset) >> 3) + ((lcu_px.y + offset) >> 3)*LCU_T_CU_WIDTH];
      if (cbf_is_set(cu_a->cbf.y, depth+1) || cbf_is_set(cu_b->cbf.y, depth+1) || cbf_is_set(cu_c->cbf.y, depth+1)) {
        cbf_set(&cur_cu->cbf.y, depth);
      }
    }

    return;
  }

  // Perform intra prediction and put the result in correct place lcu.
  vector2d_t pic_px = { state->tile->frame->width, state->tile->frame->height };
  vector2d_t luma_px = { x, y };
  kvz_intra_references refs;
  const int_fast8_t log2_width = kvz_g_convert_to_bit[width] + 2;
  kvz_intra_build_reference(log2_width, COLOR_Y, &luma_px, &pic_px, lcu, &refs);

  kvz_pixel pred[32 * 32];
  kvz_intra_get_pred_new(&refs, log2_width, intra_mode, COLOR_Y, pred);
  
  kvz_pixel *block_in_lcu = &lcu->rec.y[lcu_px.x + lcu_px.y * LCU_WIDTH];
  kvz_pixels_blit(pred, block_in_lcu, width, width, width, LCU_WIDTH);

  kvz_quantize_lcu_luma_residual(state, x, y, depth, cur_cu, lcu);
}

void kvz_intra_recon_lcu_chroma(encoder_state_t * const state, int x, int y, int depth, int8_t intra_mode, cu_info_t *cur_cu, lcu_t *lcu)
{
  const vector2d_t lcu_px = { x & 0x3f, y & 0x3f };
  const int8_t width = LCU_WIDTH >> depth;
  const int8_t width_c = (depth == MAX_PU_DEPTH ? width : width / 2);

  if (cur_cu == NULL) {
    cur_cu = &lcu->cu[LCU_CU_OFFSET + (lcu_px.x >> 3) + (lcu_px.y >> 3)*LCU_T_CU_WIDTH];
  }

  if (depth == 0 || cur_cu->tr_depth > depth) {
    int offset = width / 2;

    kvz_intra_recon_lcu_chroma(state, x,          y,          depth+1, intra_mode, NULL, lcu);
    kvz_intra_recon_lcu_chroma(state, x + offset, y,          depth+1, intra_mode, NULL, lcu);
    kvz_intra_recon_lcu_chroma(state, x,          y + offset, depth+1, intra_mode, NULL, lcu);
    kvz_intra_recon_lcu_chroma(state, x + offset, y + offset, depth+1, intra_mode, NULL, lcu);

    if (depth < MAX_DEPTH) {
      cu_info_t *cu_a = &lcu->cu[LCU_CU_OFFSET + ((lcu_px.x + offset) >> 3) + (lcu_px.y >> 3)        *LCU_T_CU_WIDTH];
      cu_info_t *cu_b = &lcu->cu[LCU_CU_OFFSET + (lcu_px.x >> 3) + ((lcu_px.y + offset) >> 3)*LCU_T_CU_WIDTH];
      cu_info_t *cu_c = &lcu->cu[LCU_CU_OFFSET + ((lcu_px.x + offset) >> 3) + ((lcu_px.y + offset) >> 3)*LCU_T_CU_WIDTH];
      if (cbf_is_set(cu_a->cbf.u, depth+1) || cbf_is_set(cu_b->cbf.u, depth+1) || cbf_is_set(cu_c->cbf.u, depth+1)) {
        cbf_set(&cur_cu->cbf.u, depth);
      }
      if (cbf_is_set(cu_a->cbf.v, depth+1) || cbf_is_set(cu_b->cbf.v, depth+1) || cbf_is_set(cu_c->cbf.v, depth+1)) {
        cbf_set(&cur_cu->cbf.v, depth);
      }
    }

    return;
  }

  if (!(x & 4 || y & 4)) {
    const int_fast8_t log2_width_c = kvz_g_convert_to_bit[width_c] + 2;
    const vector2d_t luma_px = { x, y };
    const vector2d_t pic_px = { state->tile->frame->width, state->tile->frame->height };

    // Intra predict U-plane and put the result in lcu buffer.
    {
      kvz_intra_references refs;
      kvz_intra_build_reference(log2_width_c, COLOR_U, &luma_px, &pic_px, lcu, &refs);

      kvz_pixel pred[32 * 32];
      kvz_intra_get_pred_new(&refs, log2_width_c, intra_mode, COLOR_U, pred);

      kvz_pixel *pu_in_lcu = &lcu->rec.u[lcu_px.x / 2 + (lcu_px.y * LCU_WIDTH) / 4];
      kvz_pixels_blit(pred, pu_in_lcu, width_c, width_c, width_c, LCU_WIDTH_C);
    }

    // Intra predict V-plane and put the result in lcu buffer.
    {
      kvz_intra_references refs;
      kvz_intra_build_reference(log2_width_c, COLOR_V, &luma_px, &pic_px, lcu, &refs);
      
      kvz_pixel pred[32 * 32];
      kvz_intra_get_pred_new(&refs, log2_width_c, intra_mode, COLOR_V, pred);

      kvz_pixel *pu_in_lcu = &lcu->rec.v[lcu_px.x / 2 + (lcu_px.y * LCU_WIDTH) / 4];
      kvz_pixels_blit(pred, pu_in_lcu, width_c, width_c, width_c, LCU_WIDTH_C);
    }

    kvz_quantize_lcu_chroma_residual(state, x, y, depth, cur_cu, lcu);
  }
}