[函数名称]
Harris角点检测函数 HarrisDetect(WriteableBitmap src, int CRF)
[算法说明]
目前的角点检测算法可归纳为3类:基于灰度图像的角点检测、基于二值图像的角点检测、基于轮廓曲线的角点检测。基于灰度图像的角点检测又可分为基于梯度、基于模板和基于模板梯度组合3类方法,其中基于模板的方法主要考虑像素领域点的灰度变化,即图像亮度的变化,将与邻点亮度对比足够大的点定义为角点。本文将介绍一种改进的Harris角点检测算法,该算法是一种基于模板与梯度组合的方法。
<strong><span style="font-size:14px;">[函数代码]</span></strong>
/// <summary>
/// Harris counter-detect.
/// </summary>
/// <param name="src">The source image.</param>
/// <param name="v">The threshould to control counters number.</param>
/// <returns></returns>
public static int[,] HarrisDetect(WriteableBitmap src, int CRF)
{
int x = src.PixelWidth;
int y = src.PixelHeight;
double[,] Ix = new double[x, y];
double[,] Iy = new double[x, y];
double[,] Ixy = new double[x, y];
double[,] cim = new double[x, y];
int[,] re = new int[x, y];
double[,] srcBytes = GetImageBytes(src);
GetIV(srcBytes, Ix, Iy, Ixy,x,y);
GaussFilter(Ix, Iy, Ixy,x,y);
cim = GetCim(Ix, Iy, Ixy,x,y);
for (int j = 1; j < y - 1; j++)
{
for (int i = 1; i < x - 1; i++)
{
if ((cim[i, j] == GetMax(cim[i - 1, j - 1], cim[i, j - 1], cim[i + 1, j - 1], cim[i - 1, j], cim[i, j], cim[i + 1, j], cim[i - 1, j + 1], cim[i, j + 1], cim[i + 1, j + 1])) && (cim[i, j] > CRF))
{
re[i, j] = 1;
}
}
}
return re;
}
//获得角点图像的 原始 信息
public static double[,] GetImageBytes(WriteableBitmap src)
{
if (src != null)
{
int w = src.PixelWidth;
int h = src.PixelHeight;
double[,] imageBytes = new double[w, h];
int b = 0, g = 0, r = 0;
byte[] temp = src.PixelBuffer.ToArray();
for (int y = 0; y < h; y++)
{
for (int x = 0; x < w * 4; x += 4)
{
b = temp[x + y * w * 4];
g = temp[x + 1 + y * w * 4];
r = temp[x + 2 + y * w * 4];
imageBytes[x, y] = (b * 0.114 + g * 0.587 + r * 0.299);
}
}
return imageBytes;
}
else
{
return null;
}
}
//梯度求取函数
private static void GetIV(double[,] src, double[,] Ix, double[,] Iy, double[,] Ixy,int x,int y)
{
for (int j = 1; j < y - 1; j++)
{
for (int i = 1; i < x - 1; i++)
{
Ix[i, j] = Math.Abs(src[i + 1, j - 1] + src[i + 1, j] + src[i + 1, j + 1] - src[i - 1, j - 1] - src[i - 1, j] - src[i - 1, j + 1]);
Iy[i, j] = Math.Abs(src[i - 1, j + 1] + src[i, j + 1] + src[i + 1, j + 1] - src[i - 1, j - 1] - src[i, j - 1] - src[i + 1, j - 1]);
Ixy[i, j] = Math.Abs(Ix[i, j] * Iy[i, j]);
}
}
}
//高斯滤波函数(对梯度图像进行高斯滤波,这里采用的是3*3的高斯滤波模板)
private static void GaussFilter(double[,] Ix, double[,] Iy, double[,] Ixy,int x,int y)
{
for (int j = 1; j < y - 1; j++)
{
for (int i = 1; i < x - 1; i++)
{
Ix[i, j] = (Ix[i - 1, j - 1] + Ix[i, j - 1] * 2 + Ix[i + 1, j - 1] + 2 * Ix[i - 1, j] + 4 * Ix[i, j] + 2 * Ix[i + 1, j] + Ix[i - 1, j + 1] + 2 * Ix[i, j + 1] + Ix[i + 1, j + 1]) / 16;
Iy[i, j] = (Iy[i - 1, j - 1] + Iy[i, j - 1] * 2 + Iy[i + 1, j - 1] + 2 * Iy[i - 1, j] + 4 * Iy[i, j] + 2 * Iy[i + 1, j] + Iy[i - 1, j + 1] + 2 * Iy[i, j + 1] + Ix[i + 1, j + 1]) / 16;
Ixy[i, j] = (Ixy[i - 1, j - 1] + Ixy[i, j - 1] * 2 + Ixy[i + 1, j - 1] + 2 * Ixy[i - 1, j] + 4 * Ixy[i, j] + 2 * Ixy[i + 1, j] + Ixy[i - 1, j + 1] + 2 * Ixy[i, j + 1] + Ix[i + 1, j + 1]) / 16;
}
}
}
//图像角点求取函数
private static double[,] GetCim(double[,] Ix, double[,] Iy, double[,] Ixy,int x,int y)
{
double cim = 0;
double[,] results = new double[x, y];
for (int j = 1; j < y - 1; j++)
{
for (int i = 1; i < x - 1; i++)
{
if (Ix[i, j] != 0 || Iy[i, j] != 0)
{
cim = Math.Abs(Ix[i, j] * Iy[i, j] - Ixy[i, j] * Ixy[i, j]) / (Ix[i, j] * Ix[i, j] + Iy[i, j] * Iy[i, j]);
results[i, j] = cim;
}
}
}
return results;
}
//最大值获取函数
private static double GetMax(params double[] src)
{
double tMax = 0;
for (int i = 0; i < src.Length; i++)
{
if (tMax < src[i])
{
tMax = src[i];
}
}
return tMax;
}
<strong><span style="font-size:14px;">[图像效果]</span></strong>