#region 二维高斯滤波
//高斯滤波器
private double[,] gaussFilter(int size,double sigma)
{
double[,] arr=new double[size,size];
double sum = 0.0;
int center = size; //以第一个点的坐标为原点,求出中心点的坐标
for (int i = 0; i < size; ++i)
for (int j = 0; j < size; ++j)
sum += arr[i,j] = Math.Exp(-((i - center)*(i - center) + (j - center)*(j - center)) / (2 * sigma*sigma));
for (int i = 0; i < size; ++i)
for (int j = 0; j < size; ++j)
arr[i,j] /= sum;
return arr;
//for (int i = 0; i < size; ++i) {
// for (int j = 0; j < size; ++j)
// Console.Write(arr[i,j]+" ");
// Console.WriteLine();
//}
}
//高斯滤波算法
private void myGaussFilter(short[,] data)
{
int w_Data = data.GetLength(0);
int h_Data = data.GetLength(1);
double[,] arr=gaussFilter(3,1.5);
for (int i = 2; i < w_Data-1; i++)
{
for (int j = 2; j < h_Data-1; j++)
{
bool judgeBool = false;
for (int a = i - 1; a <= i + 1; a++)
{
for (int b = j - 1; b <= j + 1; b++)
{
if (Math.Abs(a - i) == 1 || Math.Abs(b - j) == 1)
{
if (data[a, b] == 0)
{
judgeBool=true;
break;
}
}
}
if (judgeBool)
{
break;
}
}
if (judgeBool)
{
//if (data[i, j] > 0)
//{
// short[,] shortTmp = (short[,])data.Clone();
// data[i, j] = edgeBianTong(shortTmp, new Point(i, j), arr);
//}
continue;
}
double tmpValue = 0;
for (int x = 0; x < 3; x++)
{
for (int y = 0; y < 3; y++)
{
tmpValue += data[i + 1 - x, j + 1 - y] * arr[x, y];//高斯滤波矩阵是对称的
}
}
data[i, j] = (short)tmpValue;
}
}
}
private void myGaussFilter123(short[,] data)
{
int w_Data = data.GetLength(0);
int h_Data = data.GetLength(1);
double[,] arr = gaussFilter(3, 1.5);
for (int i = 2; i < w_Data - 1; i++)
{
for (int j = 2; j < h_Data - 1; j++)
{
bool judgeBool = false;
for (int a = i - 1; a <= i + 1; a++)
{
for (int b = j - 1; b <= j + 1; b++)
{
if (Math.Abs(a - i) == 1 || Math.Abs(b - j) == 1)
{
if (data[a, b] == 0)
{
judgeBool = true;
break;
}
}
}
if (judgeBool)
{
break;
}
}
if (judgeBool)
{
if (data[i, j] > 0)
{
short[,] shortTmp = (short[,])data.Clone();
data[i, j] = edgeBianTong(shortTmp, new Point(i, j), arr);
}
continue;
}
double tmpValue = 0;
for (int x = 0; x < 3; x++)
{
for (int y = 0; y < 3; y++)
{
tmpValue += data[i + 1 - x, j + 1 - y] * arr[x, y];//高斯滤波矩阵是对称的
}
}
data[i, j] = (short)tmpValue;
}
}
}
//边界变通
private short edgeBianTong(short[,] data, Point point, double[,] arr)
{
//扩充边界和高斯滤波同时进行
double tmpValue = 0;
for (int i = point.X - 1 ,a=0; i <= point.X + 1; i++,a++)
{
for (int j = point.Y - 1 ,b=0; j <= point.Y + 1; j++,b++)
{
if (data[i, j] == 0)
{
data[i, j] = data[point.X,point.Y];
}
//tmpValue += data[i, j] * arr[1-Math.Abs(i - point.X), 1-Math.Abs(j - point.Y)];
tmpValue += data[i, j] * arr[a,b];
}
}
return (short)tmpValue;
}
#endregion
注:
1.本例子是红外图像做差得到的人体图像,非人体图像温度数值都为0.
2.myGaussFilter123含边界点(未全包含:整幅图像的边界未包含,仅涵盖了图像(不不包括图像边界)中的人体边缘点)。
3.myGaussFilter不含边界点,经测试发现,对边界去毛边并未有区别(也许与自己扩充边界用原值复制相关,有待进一步测试),所以对于整幅图像的边界点不再进行高斯滤波处理了。
4.不过对于颗粒度有点强的图像,用高斯滤波圆润挺好的。