西安交通大学大学数字图像处理大作业

数字图像处理

目录

数字图像处理...................................................................................................................................................... 1 作业一.................................................................................................................................................................. 1

一 作业要求 ............................................................................................................. 1

二 源代码 ................................................................................................................. 1

三 运行结果 ............................................................................................................. 8

作业二.................................................................................................................................................................. 9

一 作业要求 ............................................................................................................. 9

二 算法描述 ........................................................................................................... 10

三 源代码 ............................................................................................................... 13

四 运行结果 ........................................................................................................... 21

作业一

一 作业要求

在图像的空间域滤波操作中，会出现有部分掩膜矩阵在图像外面的情况，所以需要给图像先加入一个边界，执行完操作之后，再去掉这个边界，保证图像中所有的像素都参与矩阵运算。

二 源代码

byte[,] filter(byte[,]f,float[,]mask)

{

int w = f.GetLength(0);

int h = f.GetLength(1);

byte[,] g = new byte[w,h];

int M = mask.GetLength(0)/2;

int N = mask.GetLength(1)/2;

for (int y=N;y1

for (int x=M;x{

float s = 0;

for (int m=-M;m<=M;m++)

for (int n=-N;n<=N;n++)

s+=f[x+m,y+n]*mask[M+m,N+n];

g[x,y] = SByte(s);

}

return g;

}

byte SByte(double v)

{

if (v>255) return 255;

2

if (v<0) return 0;

return (byte)v;

}

float[,] averagingMask(int M,int N)

{

float[,] mask = new float[2*M+1,2*N+1];

for (int m=-M;m<=M;m++)

for (int n=-N;n<=N;n++)

mask[M+m,N+n] = 1.0f/((2*M+1)*(2*N+1));

return mask;

}

byte[,] addboard(byte[,] f,int M,int N)

{

3

int w=f.GetLength(0);

int h=f.GetLength(1);

int gw=w+2*M;

int gh=h+2*N;

byte[,] g=new byte[gw,gh];

//add top board and bottom board

for(int i=0;ifor(int j=0;jg[M+j,i]=f[j,0];

for(int i=0;ifor(int j=0;jg[M+j,i+h+N]=f[j,h-1];

//copy the image

4

for(int i=0;ifor (int j=0;jg[i+M,j+N]=f[i,j];

//add left and right board

for(int i=0;ifor (int j=0;jg[i,j]=g[M,j];

for(int i=0;ifor (int j=0;jg[w+M+i,j]=g[gw-1-M,j];

return g;

}

byte[,] removeboard(byte[,]f,int M,int N)

5

{

int w=f.GetLength(0);

int h=f.GetLength(1);

int gw=w-2*M;

int gh=h-2*N;

byte[,] g=new byte[gw,gh]; for(int i=0;ifor(int j=0;jg[i,j]=f[i+M,j+N];

return g;

}

void main()

{

6

byte[,] f = LoadImg();

ShowImg(\"f\

int w=f.GetLength(0);

int h=f.GetLength(1);

int M=10,N=20;

int gw=w-2*M;

int gh=h-2*N;

byte[,] boardimage=new byte[gw,gh];

byte[,] filterimage=new byte[gw,gh];

boardimage=addboard(f,M,N);

ShowImg(\"boardimage\

filterimage=filter(boardimage,averagingMask(M,N));

ShowImg(\"result\

7

}

运行结果

原图像：

加边界之后的图像：

8

三

均值滤波并且去除边界的图像：

作业二

一 作业要求

9

给定图像与处理结果，如图一所示，思考算法并编程实现。

图一 左边为原图，右边为处理结果

二 算法描述

Compare the result and the original image, we can see that the program change the contrast of image and enhance the edge of Lena. Consider using intensity transform and Sobel edge detection to achieve the goal. After many experiments, I choose following intensity transform function.

y0.25x110

The graph of this function is Fig 1.

10

Fig 1 Graph of intensity transform function

Use Sobel to detect the edge of the original image.

Sobel in x direction can be written as the following transform matrix:

101202101 SobleX=Similar, Sobel in y direction can be written as the following transform matrix:

121000121 SobleY=11

We can use SobelX and SobelY to compute the gradient image.

fgx2gy2 f is the value of pixels in the gradient image, gx is the value of pixels in image

processed by SobelX, and

gy is the value of pixels in image processed by SobelY.

In order to decrease noise, I use the following function to process gradient image:

fafterfbefore70

After processing, I use 0 to replace these negative values.

Because the gradient image has a very high contrast, so I use intensity transform again.

fafter0.22fbefore

After that, the final result can be calculated:

resultyf

All the function is chosen through experiment, maybe the result isn’t the same with the teacher’s result, but it’s very similar.

12

源代码

byte[,] filter(byte[,]f,float[,]mask)

{

int w = f.GetLength(0);

int h = f.GetLength(1);

int M = mask.GetLength(0)/2;

int N = mask.GetLength(1)/2;

byte[,] g = new byte[w,h];

for (int y=N;yfor (int x=M;x{

float r = 0;

for (int m=-M;m<=M;m++)

三 13

for (int n=-N;n<=N;n++)

{

r+=f[x+m,y+n]*mask[M+m,N+n];

}

g[x,y]=S(r);

}

return g;

}

byte[,] SobelX(byte[,]f)

{

float[,] mask = new float[3,3];

mask[0,0]=-1; mask[1,0] = 0; mask[0,1]=-2; mask[1,1] = 0; 14

mask[2,0] = 1;

mask[2,1] = 2;

mask[0,2]=-1; mask[1,2] = 0; mask[2,2] = 1;

return filter(f,mask);

}

byte[,] SobelY(byte[,]f)

{

float[,] mask = new float[3,3];

mask[0,0]=-1; mask[1,0] = -2; mask[0,1]= 0; mask[1,1] = 0; mask[0,2]= 1; mask[1,2] = 2; return filter(f,mask);

}

byte S(double f)

{

mask[2,1] = 0; mask[2,2] = 1;

15

mask[2,0] =-1;

if(f>255) return 255;

if(f<0) return 0;

return (byte)f;

}

byte[,] Scale(byte[,]f,double a)

{

int w = f.GetLength(0);

int h = f.GetLength(1);

for(int i=0;ifor(int j=0;jf[i,j]=(byte)(f[i,j]*a);

return f;

}

16 byte[,] Subnum(byte[,]f,int a)

{

int w = f.GetLength(0);

int h = f.GetLength(1);

for(int i=0;ifor(int j=0;jf[i,j]=S((f[i,j]-a));

return f;

}

byte[,] Sub(byte[,]f,byte[,]g)

{

int w = f.GetLength(0);

int h = f.GetLength(1);

17 byte[,] p = new byte[w,h];

for(int x=0;xfor(int y=0;yp[x,y]=S(f[x,y]-g[x,y]);

return p;

}

byte[,] Mix(byte[,]fx,byte[,]fy)

{

int w = fx.GetLength(0);

int h = fx.GetLength(1);

byte[,] g = new byte[w,h];

for (int y=0;y18

for (int x=0;x{

float px =(fx[x,y]);

float py =(fy[x,y]);

g[x,y] = S(Sqrt(px*px+py*py));

}

return g;

}

byte[,] IntensityTransform(byte[,]f)

{

int w = f.GetLength(0);

int h = f.GetLength(1);

byte[,] g = new byte[w,h];

19

for(int i=0;ifor(int j=0;jg[i,j] = S(0.25*f[i,j]+110);

return g;

}

void main()

{

byte [,] I = LoadImg();

ShowImg(\"Original\

byte [,] I1 = IntensityTransform(I);

ShowImg(\"IntensityTransform\

byte [,] I2 = SobelX(I);

byte [,] I3 = SobelY(I);

20

byte [,] I4 = Mix(I2,I3);

ShowImg(\"Gradient\

byte [,] I5=Subnum(I4,70);

ShowImg(\"DecreaseNoise\

byte [,] I6=Scale(I5,0.22);

ShowImg(\"Scale\

byte [,] I7 = Sub(I1,I6);

ShowImg(\"FinalResult\

}

四 运行结果

原图像：

21

灰度变换之后的图像：

Sobel算子提取到的边缘图像：

22

降低噪声之后的边缘图像：

改变提取到的边缘的亮度之后得到的图像：

23

最终结果：

24