`timescale 1ns / 1ps

module DPC #(
    parameter reg [15:0] TOTAL_WIDTH  = 512 + 3,  // 总图像宽度
    parameter reg [15:0] TOTAL_HEIGHT = 256 + 3,  // 总图像高度
    parameter reg [ 1:0] RAW_TYPE     = 3,        // (0,0)位置算起RAW_TYPE的值
    parameter reg [ 4:0] DATA_WIDTH   = 16        // 输入/输出数据位宽
) (
    input wire clk,
    input wire reset,

    input  wire [DATA_WIDTH - 1:0] in_data [5*5],
    output reg  [DATA_WIDTH - 1:0] out_data,

    input  wire in_valid,
    output wire out_valid,

    input  wire in_ready,
    output wire out_ready
);

  localparam WINDOW_LENGTH = 5;
  localparam DATA_NUM = WINDOW_LENGTH * WINDOW_LENGTH;
  localparam EXPAND_BITS = 5;
  localparam PIPILINE = 9;

  reg [PIPILINE-1:0] pipeline_valid;
  wire pipeline_running;
  assign pipeline_running = in_ready | ~pipeline_valid[PIPILINE-1];

  //out_ready ：只要本模块可以接收数据就一直拉高
  assign out_ready = pipeline_running;
  //out_valid ：只要本模块可以发出数据就一直拉高
  assign out_valid = pipeline_valid[PIPILINE-1];

  reg [DATA_WIDTH-1:0] data_cache[DATA_NUM];  // 缓存颜色数据，行列nxn
  reg [DATA_WIDTH-1:0] data_cache0[DATA_NUM];  // 缓存颜色数据，行列nxn
  reg [DATA_WIDTH-1:0] channel_cache[9];  // 缓存颜色通道数据，channel_cache[4]就是中心像素点
  reg [DATA_WIDTH-1:0]
      channel_cache0,
      channel_cache1,
      channel_cache2,
      channel_cache3,
      channel_cache4;  // 缓存中心像素点的颜色数据
  reg signed [DATA_WIDTH-1+EXPAND_BITS:0]
      grad_h_cache[3], grad_v_cache[3], grad_45_cache[3], grad_135_cache[3];
  reg [DATA_WIDTH-1+EXPAND_BITS:0]
      grad_h_cache0[3], grad_v_cache0[3], grad_45_cache0[3], grad_135_cache0[3];
  reg [DATA_WIDTH-1+EXPAND_BITS:0]
      grad_h_cache1[3], grad_v_cache1[3], grad_45_cache1[3], grad_135_cache1[3];
  reg [DATA_WIDTH-1+EXPAND_BITS:0]
      grad_h_cache2[3], grad_v_cache2[3], grad_45_cache2[3], grad_135_cache2[3];
  reg [DATA_WIDTH-1+EXPAND_BITS:0] grad_median_cache[4];
  reg [1:0] flag_which_dict;
  reg [DATA_WIDTH-1:0]
      channel_cache_correct[4],
      channel_cache_correct0[4],
      channel_cache_correct1[4],
      channel_cache_correct2[4];
  reg [DATA_WIDTH-1:0] channel_cache_correct_final;
  reg flag_if_need_corection;
  reg [15:0] pos_x;
  reg pos_y_bit;
  reg [1:0] raw_type;

  integer i;
  always @(posedge clk) begin
    if (reset) begin
      for (i = 0; i < DATA_NUM; i = i + 1) data_cache[i] <= 0;
      for (i = 0; i < DATA_NUM; i = i + 1) data_cache0[i] <= 0;
      for (i = 0; i < 9; i = i + 1) channel_cache[i] <= 0;
      channel_cache0 <= 0;
      channel_cache1 <= 0;
      channel_cache2 <= 0;
      channel_cache3 <= 0;
      channel_cache4 <= 0;
      channel_cache_correct_final <= 0;
      for (i = 0; i < 3; i = i + 1) grad_h_cache[i] <= 0;
      for (i = 0; i < 3; i = i + 1) grad_h_cache1[i] <= 0;
      for (i = 0; i < 3; i = i + 1) grad_h_cache2[i] <= 0;
      for (i = 0; i < 3; i = i + 1) grad_v_cache[i] <= 0;
      for (i = 0; i < 3; i = i + 1) grad_v_cache1[i] <= 0;
      for (i = 0; i < 3; i = i + 1) grad_v_cache2[i] <= 0;
      for (i = 0; i < 3; i = i + 1) grad_45_cache[i] <= 0;
      for (i = 0; i < 3; i = i + 1) grad_45_cache1[i] <= 0;
      for (i = 0; i < 3; i = i + 1) grad_45_cache2[i] <= 0;
      for (i = 0; i < 3; i = i + 1) grad_135_cache[i] <= 0;
      for (i = 0; i < 3; i = i + 1) grad_135_cache1[i] <= 0;
      for (i = 0; i < 3; i = i + 1) grad_135_cache2[i] <= 0;
      for (i = 0; i < 3; i = i + 1) grad_median_cache[i] <= 0;
      flag_which_dict <= 0;
      flag_if_need_corection <= 0;
      for (i = 0; i < 4; i = i + 1) channel_cache_correct[i] <= 0;
      for (i = 0; i < 4; i = i + 1) channel_cache_correct1[i] <= 0;
      for (i = 0; i < 4; i = i + 1) channel_cache_correct0[i] <= 0;
      for (i = 0; i < 4; i = i + 1) channel_cache_correct2[i] <= 0;

      pipeline_valid <= 0;
      out_data <= 0;
      pos_x <= ~0;
      pos_y_bit <= ~0;
      raw_type <= RAW_TYPE;
    end else if (pipeline_running) begin

      pipeline_valid <= {pipeline_valid[PIPILINE-2:0], in_valid};

      if (in_valid) begin
        for (i = 0; i < DATA_NUM; i = i + 1) data_cache0[i] <= in_data[i];
        pos_x <= (pos_x >= TOTAL_WIDTH - 1) ? (0) : (pos_x + 1);
        pos_y_bit <= (pos_x >= TOTAL_WIDTH - 1) ? (~pos_y_bit) : (pos_y_bit);
      end

      if (pipeline_valid[0]) begin
        for (i = 0; i < DATA_NUM; i = i + 1) data_cache[i] <= data_cache0[i];
        case (RAW_TYPE)
          2'b00: raw_type <= {pos_y_bit, pos_x[0]};
          2'b01: raw_type <= {pos_y_bit, ~pos_x[0]};
          2'b10: raw_type <= {~pos_y_bit, pos_x[0]};
          2'b11: raw_type <= {~pos_y_bit, ~pos_x[0]};
        endcase
      end

      if (pipeline_valid[1]) begin
        case (raw_type)
          1, 2: begin
            channel_cache[0] <= data_cache[00];
            channel_cache[1] <= data_cache[10];
            channel_cache[2] <= data_cache[20];
            channel_cache[3] <= data_cache[02];
            channel_cache[4] <= data_cache[12];
            channel_cache[5] <= data_cache[22];
            channel_cache[6] <= data_cache[04];
            channel_cache[7] <= data_cache[14];
            channel_cache[8] <= data_cache[24];
          end
          0, 3: begin
            channel_cache[0] <= data_cache[02];
            channel_cache[1] <= data_cache[06];
            channel_cache[2] <= data_cache[10];
            channel_cache[3] <= data_cache[08];
            channel_cache[4] <= data_cache[12];
            channel_cache[5] <= data_cache[16];
            channel_cache[6] <= data_cache[14];
            channel_cache[7] <= data_cache[18];
            channel_cache[8] <= data_cache[22];
          end
        endcase
      end

      if (pipeline_valid[2]) begin  //计算梯度，同时开始校正后数据的部分计算
        channel_cache0 <= channel_cache[4];

        grad_h_cache[0] <= channel_cache[0] + channel_cache[2] - 2 * channel_cache[1];
        grad_h_cache[1] <= channel_cache[3] + channel_cache[5] - 2 * channel_cache[4];
        grad_h_cache[2] <= channel_cache[6] + channel_cache[8] - 2 * channel_cache[7];
        grad_v_cache[0] <= channel_cache[0] + channel_cache[6] - 2 * channel_cache[3];
        grad_v_cache[1] <= channel_cache[1] + channel_cache[7] - 2 * channel_cache[4];
        grad_v_cache[2] <= channel_cache[2] + channel_cache[8] - 2 * channel_cache[5];
        grad_45_cache[0] <= 2 * (channel_cache[1] - channel_cache[3]);
        grad_45_cache[1] <= channel_cache[6] + channel_cache[2] - 2 * channel_cache[4];
        grad_45_cache[2] <= 2 * (channel_cache[7] - channel_cache[5]);
        grad_135_cache[0] <= 2 * (channel_cache[1] - channel_cache[5]);
        grad_135_cache[1] <= channel_cache[0] + channel_cache[8] - 2 * channel_cache[4];
        grad_135_cache[2] <= 2 * (channel_cache[3] - channel_cache[7]);

        channel_cache_correct[0] <= channel_cache[3] / 2 + channel_cache[5] / 2;
        channel_cache_correct[1] <= channel_cache[1] / 2 + channel_cache[7] / 2;
        channel_cache_correct[2] <= channel_cache[2] / 2 + channel_cache[6] / 2;
        channel_cache_correct[3] <= channel_cache[0] / 2 + channel_cache[8] / 2;
      end

      if(pipeline_valid[3]) begin //计算绝对值，同时完成校正后数据的计算,注意grad_h_cache等是singed，可能为负数
        channel_cache1 <= channel_cache0;

        grad_h_cache0  [0] <= grad_h_cache  [0][DATA_WIDTH-1+EXPAND_BITS] ? (~grad_h_cache  [0] + 1) : (grad_h_cache  [0]);
        grad_h_cache0  [1] <= grad_h_cache  [1][DATA_WIDTH-1+EXPAND_BITS] ? (~grad_h_cache  [1] + 1) : (grad_h_cache  [1]);
        grad_h_cache0  [2] <= grad_h_cache  [2][DATA_WIDTH-1+EXPAND_BITS] ? (~grad_h_cache  [2] + 1) : (grad_h_cache  [2]);
        grad_v_cache0  [0] <= grad_v_cache  [0][DATA_WIDTH-1+EXPAND_BITS] ? (~grad_v_cache  [0] + 1) : (grad_v_cache  [0]);
        grad_v_cache0  [1] <= grad_v_cache  [1][DATA_WIDTH-1+EXPAND_BITS] ? (~grad_v_cache  [1] + 1) : (grad_v_cache  [1]);
        grad_v_cache0  [2] <= grad_v_cache  [2][DATA_WIDTH-1+EXPAND_BITS] ? (~grad_v_cache  [2] + 1) : (grad_v_cache  [2]);
        grad_45_cache0 [0] <= grad_45_cache [0][DATA_WIDTH-1+EXPAND_BITS] ? (~grad_45_cache [0] + 1) : (grad_45_cache [0]);
        grad_45_cache0 [1] <= grad_45_cache [1][DATA_WIDTH-1+EXPAND_BITS] ? (~grad_45_cache [1] + 1) : (grad_45_cache [1]);
        grad_45_cache0 [2] <= grad_45_cache [2][DATA_WIDTH-1+EXPAND_BITS] ? (~grad_45_cache [2] + 1) : (grad_45_cache [2]);
        grad_135_cache0[0] <= grad_135_cache[0][DATA_WIDTH-1+EXPAND_BITS] ? (~grad_135_cache[0] + 1) : (grad_135_cache[0]);
        grad_135_cache0[1] <= grad_135_cache[1][DATA_WIDTH-1+EXPAND_BITS] ? (~grad_135_cache[1] + 1) : (grad_135_cache[1]);
        grad_135_cache0[2] <= grad_135_cache[2][DATA_WIDTH-1+EXPAND_BITS] ? (~grad_135_cache[2] + 1) : (grad_135_cache[2]);
        channel_cache_correct0[0] <= channel_cache_correct[0] + grad_h_cache[0]/4 + grad_h_cache[2]/4;
        channel_cache_correct0[1] <= channel_cache_correct[1] + grad_v_cache[0]/4 + grad_v_cache[2]/4;
        channel_cache_correct0[2] <= channel_cache_correct[2] + grad_45_cache[0]/4 + grad_45_cache[2]/4;
        channel_cache_correct0[3] <= channel_cache_correct[3] + grad_135_cache[0]/4 + grad_135_cache[2]/4;
      end

      if (pipeline_valid[4]) begin  //计算中位数
        channel_cache2 <= channel_cache1;
        for (i = 0; i < 4; i = i + 1) channel_cache_correct1[i] <= channel_cache_correct0[i];
        for (i = 0; i < 3 + EXPAND_BITS; i = i + 1) grad_h_cache1[i] <= grad_h_cache0[i];
        for (i = 0; i < 3 + EXPAND_BITS; i = i + 1) grad_v_cache1[i] <= grad_v_cache0[i];
        for (i = 0; i < 3 + EXPAND_BITS; i = i + 1) grad_45_cache1[i] <= grad_45_cache0[i];
        for (i = 0; i < 3 + EXPAND_BITS; i = i + 1) grad_135_cache1[i] <= grad_135_cache0[i];

        grad_median_cache[0] <= MEDIAN(grad_h_cache0);
        grad_median_cache[1] <= MEDIAN(grad_v_cache0);
        grad_median_cache[2] <= MEDIAN(grad_45_cache0);
        grad_median_cache[3] <= MEDIAN(grad_135_cache0);
      end

      if (pipeline_valid[5]) begin  //计算最小值，判断最小梯度方向
        channel_cache3 <= channel_cache2;
        for (i = 0; i < 4; i = i + 1) channel_cache_correct2[i] <= channel_cache_correct1[i];
        for (i = 0; i < 3 + EXPAND_BITS; i = i + 1) grad_h_cache2[i] <= grad_h_cache1[i];
        for (i = 0; i < 3 + EXPAND_BITS; i = i + 1) grad_v_cache2[i] <= grad_v_cache1[i];
        for (i = 0; i < 3 + EXPAND_BITS; i = i + 1) grad_45_cache2[i] <= grad_45_cache1[i];
        for (i = 0; i < 3 + EXPAND_BITS; i = i + 1) grad_135_cache2[i] <= grad_135_cache1[i];

        flag_which_dict <= MIN(grad_median_cache);
      end

      if (pipeline_valid[6]) begin  //在最小梯度方向上判断中心点是否是坏点
        channel_cache4 <= channel_cache3;
        channel_cache_correct_final <= channel_cache_correct2[flag_which_dict];
        case (flag_which_dict)
          2'b00:
          flag_if_need_corection <= grad_h_cache2[1] > 4 * (grad_h_cache2[0] + grad_h_cache2[2]);
          2'b01:
          flag_if_need_corection <= grad_v_cache2[1] > 4 * (grad_v_cache2[0] + grad_v_cache2[2]);
          2'b10:
          flag_if_need_corection <= grad_45_cache2[1] > 3 * (grad_45_cache2[0] + grad_45_cache2[2]);
          2'b11:
          flag_if_need_corection <= grad_135_cache2[1] > 3*(grad_135_cache2[0] + grad_135_cache2[2]);
        endcase
      end

      if(pipeline_valid[7]) begin //如果是坏点，输出计算后的值；如果不是坏点，输出原值
        out_data <= (flag_if_need_corection) ? (channel_cache_correct_final) : (channel_cache4);
      end
    end
  end

  function [DATA_WIDTH-1+EXPAND_BITS:0] MEDIAN;
    input [DATA_WIDTH-1+EXPAND_BITS:0] inx[3];
    begin
      if ((inx[0] >= inx[1] && inx[1] >= inx[2]) || (inx[2] >= inx[1] && inx[1] >= inx[0]))
        MEDIAN = inx[1];
      else if ((inx[1] >= inx[0]) || (inx[0] >= inx[1])) MEDIAN = inx[0];
      else MEDIAN = inx[2];
    end
  endfunction

  function [1:0] MIN;
    input [DATA_WIDTH-1+EXPAND_BITS:0] inx[4];
    begin
      if (inx[0] >= inx[1] && inx[0] >= inx[2] && inx[0] >= inx[3]) MIN = 2'b00;
      else if (inx[1] >= inx[0] && inx[1] >= inx[2] && inx[1] >= inx[3]) MIN = 2'b01;
      else if (inx[2] >= inx[0] && inx[2] >= inx[1] && inx[2] >= inx[3]) MIN = 2'b10;
      else MIN = 2'b11;
    end
  endfunction

  /*
    00 05 10 15 20
    01 06 11 16 21      0 1 2
    02 07 12 17 22  ->  3 4 5
    03 08 13 18 23      6 7 8
    04 09 14 19 24

    rawtype==0: center is GREEN
    g r g r g      / / g / /
    b g b g b      / g / g /
    g r g r g  ->  g / G / g
    b g b g b      / g / g /
    g r g r g      / / g / /
    
    rawtype==1: center is RED
    r g r g r      r / r / r
    g b g b g      / / / / /
    r g r g r  ->  r / R / r
    g b g b g      / / / / /
    r g r g r      r / r / r
    
    rawtype==2: center is BLUE
    b g b g b      b / b / b
    g r g r g      / / / / /
    b g b g b  ->  b / B / b
    g r g r g      / / / / /
    b g b g b      b / b / b
    
    rawtype==3: center is GREEN
    g b g b g      / / g / /
    r g r g r      / g / g /
    g b g b g  ->  g / G / g
    r g r g r      / g / g /
    r g r g r      / / g / /


  */

endmodule