機(jī)器視覺實驗報告

1、-大學(xué)實 驗 報 告作者所在學(xué)院： 作者所在專業(yè)： 作者所在班級： 作 者 姓 名 ： 實驗報告1.1.1實驗內(nèi)容Averaging kernel(3*3 and 5*5)1.1.2算法實現(xiàn)clear all;clc;I=imread('.PA1House1.jpg');n=input('請輸入濾波器模板大小n');K1=avefilt(I,n);figure;subplot(1,2,1);imshow(I);title('原始圖像');subplot(1,2,2);imshow(K1);title('均值濾波后圖像');func

2、tion d=avefilt(x,n) a=ones(n,n); p=size(x); x1=double(x);x2=x1;for i=1:p(1)-n+1 for j=1:p(2)-n+1 c=x1(i:i+(n-1),j:j+(n-1).*a; s=sum(sum(c); x2(i+(n-1)/2,j+(n-1)/2)=s/(n*n); endendd=uint8(x2);1.1.3實驗結(jié)果1.2.1實驗內(nèi)容Gaussian Kernel ( =1,2,3 ) Use (2 +1)×(2 +1) as size of Kernel (You may write a separa

3、te function to generate Gaussian Kernels for different values of .)1.2.2算法實現(xiàn)clear all;clc;I=imread('.PA1House1.jpg');sigma=input('請輸入方差的大小n');n=2*sigma+1;K1=gaussfilt(sigma,n,I);figure;subplot(1,2,1);imshow(I);title('原始圖像');subplot(1,2,2);imshow(K1);title('高斯濾波后圖像');f

4、unction d=gaussfilt(k,n,s) Img = double(s); n1=floor(n+1)/2);%計算圖象中心 for i=1:n for j=1:n b(i,j) =exp(-(i-n1)2+(j-n1)2)/(4*k)/(4*pi*k); end end Img1=conv2(Img,b,'same'); d=uint8(Img1);1.2.3實驗結(jié)果1.3.1實驗內(nèi)容Sobel Edge Operators and Prewitt Edge Operators1.3.2算法實現(xiàn)clcclear allI=imread('.PA1House

5、1.jpg');J=imread('.PA1House2.jpg');K1= sobeledge(I);%調(diào)用sobel水平算子函數(shù)K2= sobeledgev(I);%調(diào)用sobel垂直算子函數(shù)K3= Prewittedge(I);%調(diào)用prewitt水平算子函數(shù)K4= Prewittedgev(I);%調(diào)用prewitt垂直算子函數(shù)K5= sobeledge(J);K6= sobeledgev(J);K7= Prewittedge(J); K8= Prewittedgev(J);figure;subplot(2,2,1);imshow(K1);title('

6、sobel算子水平方向邊緣檢測');subplot(2,2,2);imshow(K2);title('sobel算子垂直方向邊緣檢測');subplot(2,2,3);imshow(K3);title('prewitt算子水平方向邊緣檢測');subplot(2,2,4);imshow(K4);title('prewitt算子垂直方向邊緣檢測');figure;subplot(2,2,1);imshow(K5);title('sobel算子水平方向邊緣檢測');subplot(2,2,2);imshow(K6);title

7、('sobel算子垂直方向邊緣檢測');subplot(2,2,3);imshow(K7);title('prewitt算子水平方向邊緣檢測');subplot(2,2,4);imshow(K8);title('prewitt算子垂直方向邊緣檢測');sobel算子水平方向函數(shù)function d=sobleedge(I) I1=double(I);m,n=size(I1);u=zeros(1,9);for i=2:m-1 for j=2:n-1 u(1)=-1*I(i-1,j-1);u(2)=-2*I(i-1,j);u(3)=-1*I(i-1,

8、j+1); u(4)=0*I(i,j-1);u(5)=0*I(i,j);u(6)=0*I(i,j+1); u(7)=1*I(i+1,j-1);u(8)=2*I(i+1,j);u(9)=1*I(i+1,j+1); image1(i,j)=abs(sum(u); endendd=uint8(image1);sobel算子垂直方向函數(shù)function d=sobleedgev(I) I1=double(I);m,n=size(I1);k=zeros(1,9);for i=2:m-1 for j=2:n-1 k(1)=-1*I(i-1,j-1);k(2)=0*I(i-1,j);k(3)=1*I(i-1

9、,j+1); k(4)=-2*I(i,j-1);k(5)=0*I(i,j);k(6)=2*I(i,j+1); k(7)=-1*I(i+1,j-1);k(8)=0*I(i+1,j);k(9)=1*I(i+1,j+1); image(i,j)=abs(sum(k); endendd=uint8(image);prewitt算子水平方向函數(shù)function d=Prewittedge(I) I1=double(I);m,n=size(I1);u=zeros(1,9);for i=2:m-1 for j=2:n-1 u(1)=1*I(i-1,j-1);u(2)=1*I(i-1,j);u(3)=1*I(

10、i-1,j+1); u(4)=0*I(i,j-1);u(5)=0*I(i,j);u(6)=0*I(i,j+1); u(7)=-1*I(i+1,j-1);u(8)=-1*I(i+1,j);u(9)=-1*I(i+1,j+1); image(i,j)=abs(sum(u); endendd=uint8(image);prewitt算子垂直方向函數(shù)function d=Prewittedgev(I) I1=double(I);m,n=size(I1);u=zeros(1,9);for i=2:m-1 for j=2:n-1 u(1)=-1*I(i-1,j-1);u(2)=0*I(i-1,j);u(3

11、)=1*I(i-1,j+1); u(4)=-1*I(i,j-1);u(5)=0*I(i,j);u(6)=1*I(i,j+1); u(7)=-1*I(i+1,j-1);u(8)=0*I(i+1,j);u(9)=1*I(i+1,j+1); image(i,j)=abs(sum(u); endendd=uint8(image);1.3.3實驗結(jié)果2.1實驗內(nèi)容Attached Noisy image1 and Noisy image2 are corrupted by salt and paper noise. Apply 5by 5 Averaging and Median filter and

12、show your outputs. Why Median filter works better thanaveraging filter?2.2算法實現(xiàn)clear all;clc;I=imread('.PA1Noisyimage1.jpg');figure;subplot(1,3,1);imshow(I);title('噪聲圖像');K1=imfilter(I,fspecial('average',5);subplot(1,3,2);imshow(K1);title('5*5的均值濾波');K2=medfilt2(I,5 5)

13、;subplot(1,3,3);imshow(K2);title('5*5的中值濾波');I=imread('.PA1Noisyimage2.jpg');figure;subplot(1,3,1);imshow(I);title('噪聲圖像');K3=imfilter(I,fspecial('average',5);subplot(1,3,2);imshow(K3);title('5*5的均值濾波');K4=medfilt2(I,5 5);subplot(1,3,3);imshow(K4);title('5

14、*5的中值濾波');2.3實驗結(jié)果均值濾波：是把每個像素都用周圍的8個像素來做均值操作，幅值近似相等且隨機(jī)分布在不同位置上，這樣可以平滑圖像，速度較快，算法簡單。 但是無法去掉噪聲，只能微弱的減弱它。中值濾波：常用的非線性濾波方法也是圖像處理技術(shù)中最常用的預(yù)處理技術(shù)。它在平滑脈沖噪聲方面非常有效,同時它可以保護(hù)圖像尖銳的邊緣，選擇適當(dāng)?shù)狞c來替代污染點的值，所以處理效果好。其中加權(quán)中值濾波能夠改進(jìn)中值濾波的邊緣信號，使其良好保持效果。3.1實驗內(nèi)容Compute gradient magnitude for attached image Q3_Imagei. Stretch the re

15、sulting magnitude (between 0 to 255) for better visualizationii. Compute the histogram of gradient magnitudeiii. Compute gradient orientation ( the angle of gradient vector)iv. Compute histogram of gradient orientation (angle between 0 and 2*pi)3.2算法實現(xiàn)clear all;clc;I=imread('.PA1Q_3.jpg');Gx

16、, Gy = gradient(I);G=sqrt(Gx.*Gx+Gy.*Gy);G1 =(G-min(min(G)./(max(max(G)-min(min(G)*255;theta=atan2(Gy,Gx);%求圖像梯度方向角，其中theta的范圍是-pipithta = theta;n = find(thta<0);%找出htat中小于零的值thta(n)=thta(n)+2*pi;%將thta的范圍變?yōu)?2*pitheta1 = thta;theta1 = rad2deg(theta1);%將theta1由弧度轉(zhuǎn)換為角度theta1 = theta1(:);figure;subp

17、lot(2,2,1);imshow(I);title('原始圖像');subplot(2,2,2);imshow(G1,);title('歸一化梯度幅度圖像');subplot(2,2,3),imshow(I);title('原始圖像');subplot(2,2,4);imshow(theta);title('梯度方向圖');figure,imhist(uint8(G1);title('梯度幅度直方圖');figure,hist(theta1,12);%統(tǒng)計梯度方向直方圖set(gca,'XTick'

18、;,0:30:360);%X軸的數(shù)據(jù)顯示范圍title('梯度方向直方圖');3.3實驗結(jié)果4.1實驗內(nèi)容Load walk_1.jpg and walk_2.jpg images in Matlab. Convert them to gray scale and subtractwalk_2.jpg from walk_1.jpg. What is the result? Why?4.2算法實現(xiàn)clear all;clc;I=imread('.PA1walk_1.jpg');J=imread('.PA1walk_2.jpg');figure;subplot(1,3,1);imshow(I);subplot(1,3,2);imshow(J);K=imsubtract(I,J);subplot(1,3,3);imshow(K);4.3實驗結(jié)果圖像減法是一種用于檢測圖像變化的處理方法；可以使用圖像減法來檢測圖像相同場景圖像的差異。所以相