FPGA學(xué)習(xí)筆記

圖像處理算法

1. 中值濾波算法1.1 原理1.2 FPGA實(shí)現(xiàn)中值濾波算法 2. 應(yīng)用MATLAB給圖片加椒鹽噪聲 3. 項(xiàng)目結(jié)果

1. 中值濾波算法

1.1 原理

• 將每一像素點(diǎn)的灰度值設(shè)置為該點(diǎn)的某領(lǐng)域窗口內(nèi)的所有像素點(diǎn)灰度值的中值。
• 特點(diǎn)：有效抑制噪聲，保護(hù)圖像邊緣信息，是經(jīng)典的平滑噪聲方法，可用作處理RGB圖像格式。

方法：

將數(shù)據(jù)按大小排序，然后根據(jù)有序的數(shù)字序列找中值，排序算法：冒泡排序，二等分排序等軟件算法，適合硬件的排序算法比較少。

• 分別對三行像素進(jìn)行排序
• 分別對三行像素中的3個最大，3個中間，3個最小分別進(jìn)行排序
• 對最大的最小，中間的中間，最小的最大進(jìn)行排序得到中值

1.2 FPGA實(shí)現(xiàn)中值濾波算法

• 項(xiàng)目目標(biāo)：
  將椒鹽噪聲圖片，通過中值濾波算法，進(jìn)行去噪處理，并通過FPGA顯示到TFT顯示屏上。
  

• 項(xiàng)目工具：
  ① 硬件：Intel Cyclone IV E系列FPGA開發(fā)板，5寸（800*480）TFT電容觸摸顯示屏；
  ② 軟件：Quartus軟件，Picture2Hex軟件，MATLAB；

• 項(xiàng)目組成模塊：
  ① pll: 產(chǎn)生項(xiàng)目所需時鐘：1. SDRAM控制器時鐘；2. SDRAM時鐘信號；3. TFT屏控制器時鐘
  ② uart串口協(xié)議（uart_rx, uart_tx）
  ③ 讀FIFO
  ④ 寫FIFO
  ⑤ SDRAM控制模塊
  ⑥ TFT屏控制模塊
  ⑦ 中值濾波模塊

• Verilog代碼

/* 1. 分別對三行像素進(jìn)行排序 2. 分別對三行像素中：3個最大值，3個中間值，3個最小值，進(jìn)行排序 3. 對：3最大值中的最小值，3個中間值中的中間值，3個最小值中的最大值，進(jìn)行排序，得到中值 */module median_filter_r0(input clk, //33MHZinput rst_n,input [15:0] data_in, //灰度像素輸入input data_in_en,//lcd顯示有效區(qū)使能信號input hs_in,//行同步信號輸入input vs_in,//場同步信號輸入output wire [15:0] data_out,//中值處理后灰度像素輸出output wire data_out_en,output wire hs_out,//行同步信號輸出output wire vs_out//場同步信號輸出 ); //----timing declaration----reg hs_reg0;reg hs_reg1;reg hs_reg2;reg vs_reg0;reg vs_reg1;reg vs_reg2;reg de_reg0;reg de_reg1;reg de_reg2;//----pipeline declaration---- wire [15:0] row0;wire [15:0] row1;wire [15:0] row2;reg [15:0] r0_c0;reg [15:0] r0_c1;reg [15:0] r0_c2;reg [15:0] r1_c0;reg [15:0] r1_c1;reg [15:0] r1_c2;reg [15:0] r2_c0;reg [15:0] r2_c1;reg [15:0] r2_c2;reg [15:0] r0_max;reg [15:0] r0_mid;reg [15:0] r0_min;reg [15:0] r1_max;reg [15:0] r1_mid;reg [15:0] r1_min;reg [15:0] r2_max;reg [15:0] r2_mid;reg [15:0] r2_min;reg [15:0] max_max;reg [15:0] max_mid;reg [15:0] max_min;reg [15:0] mid_max;reg [15:0] mid_mid;reg [15:0] mid_min;reg [15:0] min_max;reg [15:0] min_mid;reg [15:0] min_min;reg [15:0] mid;//-------------------------------------------------- //----3行像素緩存--------------------------//shifter3_3 shifter3_3(.clken(data_in_en),.clock(clk),.shiftin(data_in),.shiftout(),.taps0x(row0),.taps1x(row1),.taps2x(row2));//----timing control-----------------------//always@(posedge clk or negedge rst_n)beginif(!rst_n)beginhs_reg0 <= 1'd0;hs_reg1 <= 1'd0;hs_reg2 <= 1'd0;vs_reg0 <= 1'd0;vs_reg1 <= 1'd0;vs_reg2 <= 1'd0;de_reg0 <= 1'd0;de_reg1 <= 1'd0;de_reg2 <= 1'd0;endelse if(data_in_en)beginhs_reg0 <= hs_in;hs_reg1 <= hs_reg0;hs_reg2 <= hs_reg1;vs_reg0 <= vs_in;vs_reg1 <= vs_reg0;vs_reg2 <= vs_reg1;de_reg0 <= data_in_en;de_reg1 <= de_reg0;de_reg2 <= de_reg1;endend //-----------------------------------------////----pipeline-----------------------------// //----3*3 matix from imagealways@(posedge clk or negedge rst_n)beginif(!rst_n)beginr0_c0 <= 16'd0;r0_c1 <= 16'd0;r0_c2 <= 16'd0;r1_c0 <= 16'd0;r1_c1 <= 16'd0;r1_c2 <= 16'd0;r2_c0 <= 16'd0;r2_c1 <= 16'd0;r2_c2 <= 16'd0; end else if(data_in_en)beginr0_c0 <= row0;r0_c1 <= r0_c0;r0_c2 <= r0_c1;r1_c0 <= row1;r1_c1 <= r1_c0;r1_c2 <= r1_c1;r2_c0 <= row2;r2_c1 <= r2_c0;r2_c2 <= r2_c1;end end //------------------------------------------------------------------ //----1.分別對三行像素進(jìn)行排序r0_max, r10_max,r2_max //----r0------------------------------------------always@(posedge clk or negedge rst_n)beginif(!rst_n)beginr0_max <= 16'd0;r0_mid <= 16'd0;r0_min <= 16'd0; end else if(data_in_en)beginif((r0_c0>=r0_c1)&&(r0_c0>=r0_c2))begin//r0_max <= r0_c0r0_max <= r0_c0;if(r0_c1>=r0_c2)beginr0_mid <= r0_c1;r0_min <= r0_c2; endelse beginr0_mid <= r0_c2;r0_min <= r0_c1; endendelse if((r0_c1>=r0_c0)&&(r0_c1>=r0_c2))begin//r0_max <= r0_c1;r0_max <= r0_c1;if(r0_c0>=r0_c2)beginr0_mid <= r0_c0;r0_min <= r0_c2; endelse beginr0_mid <= r0_c2;r0_min <= r0_c0; endendelse if((r0_c2>=r0_c0)&&(r0_c2>=r0_c1))begin//r0_max <= r0_c2;r0_max <= r0_c2;if(r0_c0>=r0_c1)beginr0_mid <= r0_c0;r0_min <= r0_c1; endelse beginr0_mid <= r0_c1;r0_min <= r0_c0; endendendend //----r1--------------------------------------always@(posedge clk or negedge rst_n)beginif(!rst_n)beginr1_max <= 16'd0;r1_mid <= 16'd0;r1_min <= 16'd0; end else if(data_in_en)beginif((r1_c0>=r1_c1)&&(r1_c0>=r1_c2))begin//r0_max <= r0_c0r1_max <= r1_c0;if(r1_c1>=r1_c2)beginr1_mid <= r1_c1;r1_min <= r1_c2; endelse beginr1_mid <= r1_c2;r1_min <= r1_c1; endendelse if((r1_c1>=r1_c0)&&(r1_c1>=r1_c2))begin//r0_max <= r0_c1;r1_max <= r1_c1;if(r1_c0>=r1_c2)beginr1_mid <= r1_c0;r1_min <= r1_c2; endelse beginr1_mid <= r1_c2;r1_min <= r1_c0; endendelse if((r1_c2>=r1_c0)&&(r1_c2>=r1_c1))begin//r0_max <= r0_c2;r1_max <= r1_c2;if(r1_c0>=r1_c1)beginr1_mid <= r1_c0;r1_min <= r1_c1; endelse beginr1_mid <= r1_c1;r1_min <= r1_c0; endendendend //----r2--------------------------------------always@(posedge clk or negedge rst_n)beginif(!rst_n)beginr2_max <= 16'd0;r2_mid <= 16'd0;r2_min <= 16'd0; end else if(data_in_en)beginif((r2_c0>=r2_c1)&&(r2_c0>=r2_c2))begin//r0_max <= r0_c0r2_max <= r2_c0;if(r2_c1>=r2_c2)beginr2_mid <= r2_c1;r2_min <= r2_c2; endelse beginr2_mid <= r2_c2;r2_min <= r2_c1; endendelse if((r2_c1>=r2_c0)&&(r2_c1>=r2_c2))begin//r0_max <= r0_c1;r2_max <= r2_c1;if(r2_c0>=r2_c2)beginr2_mid <= r2_c0;r2_min <= r2_c2; endelse beginr2_mid <= r2_c2;r2_min <= r2_c0; endendelse if((r2_c2>=r2_c0)&&(r2_c2>=r2_c1))begin//r0_max <= r0_c2;r2_max <= r2_c2;if(r2_c0>=r2_c1)beginr2_mid <= r2_c0;r2_min <= r2_c1; endelse beginr2_mid <= r2_c1;r2_min <= r2_c0; endendendend //-------------------------------------------------------------------- //----2. 分別對三行像素中：3個最大值，3個中間值，3個最小值，進(jìn)行排序 //----3個最大值------------------------------------------------always@(posedge clk or negedge rst_n)beginif(!rst_n)beginmax_max <= 16'd0;max_mid <= 16'd0;max_min <= 16'd0; endelse if(data_in_en)beginif((r0_max >= r1_max)&&(r0_max >= r2_max))begin//max_max <= r0_maxmax_max <= r0_max;if((r1_max >= r2_max))beginmax_mid <= r1_max;max_min <= r2_max; end else beginmax_mid <= r2_max;max_min <= r1_max; end endelse if((r1_max >= r0_max)&&(r1_max >= r2_max))begin//max_max <= r1_max;max_max <= r1_max;if((r0_max >= r2_max))beginmax_mid <= r0_max;max_min <= r2_max; end else beginmax_mid <= r2_max;max_min <= r0_max; end endelse if((r2_max >= r0_max)&&(r2_max >= r1_max))begin//max_max <= r2_max;max_max <= r2_max;if((r0_max >= r1_max))beginmax_mid <= r0_max;max_min <= r1_max; end else beginmax_mid <= r1_max;max_min <= r0_max; end endendend //----3個中間值------------------------------------------------always@(posedge clk or negedge rst_n)beginif(!rst_n)beginmid_max <= 16'd0;mid_mid <= 16'd0;mid_min <= 16'd0; endelse if(data_in_en)beginif((r0_mid >= r1_mid)&&(r0_mid >= r2_mid))begin//mid_max <= r0_midmid_max <= r0_mid;if((r1_mid >= r2_mid))beginmid_mid <= r1_mid;mid_min <= r2_mid; end else beginmid_mid <= r2_mid;mid_min <= r1_mid; end endelse if((r1_mid >= r0_mid)&&(r1_mid >= r2_mid))begin//mid_max <= r1_mid;mid_max <= r1_mid;if((r0_mid >= r2_mid))beginmid_mid <= r0_mid;mid_min <= r2_mid; end else beginmid_mid <= r2_mid;mid_min <= r0_mid; end endelse if((r2_mid >= r0_mid)&&(r2_mid >= r1_mid))begin//mid_max <= r2_mid;mid_max <= r2_mid;if((r0_mid >= r1_mid))beginmid_mid <= r0_mid;mid_min <= r1_mid; end else beginmid_mid <= r1_mid;mid_min <= r0_mid; end endendend//----3個最小值------------------------------------------------always@(posedge clk or negedge rst_n)beginif(!rst_n)beginmin_max <= 16'd0;min_mid <= 16'd0;min_min <= 16'd0; endelse if(data_in_en)beginif((r0_min >= r1_min)&&(r0_min >= r2_min))begin//min_max <= r0_minmin_max <= r0_min;if((r1_min >= r2_min))beginmin_mid <= r1_min;min_min <= r2_min; end else beginmin_mid <= r2_min;min_min <= r1_min; end endelse if((r1_min >= r0_min)&&(r1_min >= r2_min))begin//min_max <= r1_min;min_max <= r1_min;if((r0_min >= r2_min))beginmin_mid <= r0_min;min_min <= r2_min; end else beginmin_mid <= r2_min;min_min <= r0_min; end endelse if((r2_min >= r0_min)&&(r2_min >= r1_min))begin//min_max <= r2_min;min_max <= r2_min;if((r0_min >= r1_min))beginmin_mid <= r0_min;min_min <= r1_min; end else beginmin_mid <= r1_min;min_min <= r0_min; end endendend //--------------------------------------------------------------------------------- //----3. 對：3個最大值中的最小值，3個中間值中的中間值，3個最小值中的最大值，進(jìn)行排序，得到中值always@(posedge clk or negedge rst_n)beginif(!rst_n)mid <= 16'd0; else if(data_in_en)beginif(((max_min >= mid_mid)&&(max_min < min_max))||((max_min < mid_mid)&&(max_min >= min_max)) )//mid <= max_min;mid <= max_min;else if(((mid_mid >= max_min)&&(mid_mid < min_max))||((mid_mid < max_min)&&(mid_mid >= min_max)) )//mid <= mid_mid;mid <= mid_mid;else if(((min_max >= max_min)&&(min_max < mid_mid))||((min_max < max_min)&&(min_max >= mid_mid)) )//mid <= min_max;mid <= min_max;else;end else;end //----result-------------------------------------------------------------------------assign data_out = mid;assign data_out_en = de_reg2;assign hs_out = hs_reg2;assign vs_out = vs_reg2; //----------------------------------------------------------------------------------- endmodule

---

2. 應(yīng)用MATLAB給圖片加椒鹽噪聲：

• MATLAB代碼

close all;clear all;clc;%清除L = imread('C:\Users\GloriaHuo\Desktop\0001.jpg'); % 在指定路徑讀取圖像 J = imnoise(L,'salt & pepper', 0.1); %Salt&pepper 噪聲密度=0.1figure(1); imshow(L); % 顯示原圖 figure(2); imshow(J); %立即彈出窗口，顯示加了加椒鹽噪聲后的圖片 hold on;imwrite(J,'C:\Users\GloriaHuo\Desktop\0001_saltpepper.jpg');%把加入噪聲的圖像保存起來

• 1. 原圖像
  
• 2. 加入椒鹽噪聲圖像
  

---

3. 項(xiàng)目結(jié)果

5寸TFT電容觸摸顯示屏
圖像：800*480像素

(1). （易烊千璽^^）網(wǎng)絡(luò)原圖：

(2). FPGA顯示原圖：

(3). 加入椒鹽噪聲圖：

• Salt&pepper 噪聲密度=0.1
  
  (4). FPGA中值濾波處理圖：

• Salt&pepper 噪聲密度=0.1
  
  項(xiàng)目結(jié)果分析：

• 對比圖3與圖4可以發(fā)現(xiàn)，中值濾波算法可以有效地去除椒鹽噪聲；

---

參考資料：《FPGA系統(tǒng)設(shè)計(jì)與驗(yàn)證實(shí)戰(zhàn)指南》
【注】：個人學(xué)習(xí)筆記，如有錯誤，望不吝賜教，這廂有禮了~~~

---

總結(jié)

以上是生活随笔為你收集整理的FPGA学习笔记_图像处理3_FPGA实现中值滤波算法的全部內(nèi)容，希望文章能夠幫你解決所遇到的問題。

如果覺得生活随笔網(wǎng)站內(nèi)容還不錯，歡迎將生活随笔推薦給好友。