首先, 计算每个像素点的兴趣值, 即以该像素点为中心, 取一个w×w(如:5×5)的方形窗口, 计算0度、45度、90度、135度四个方向灰度差的平方和, 取其中的最小值作为该像素点的兴趣值。

其次，根据实际图像设定一个阈值，遍历图像以兴趣值大于该阈值的点为候选点。

最后，选一个一定大小的窗口，让该窗口遍历灰度图象，在此过程中取窗口中兴趣值最大的候选点为特征点，算法结束。

-First, calculate the value of each pixel of interest, that is the center of the pixel, take a w × w (eg: 5 × 5) square window, computing 0 degrees, 45 degrees, 90 degrees, 135 degrees in four directions gray squares difference, whichever is the minimum value as the value of the pixel of interest. Secondly, based on the actual images to set a threshold value of traversing the image in order to interest more than the threshold point for the candidate point. Finally, the selection of a certain size of the window, so that the window traversal gray image, taken in the process window, the candidate interested in the value of the maximum points for the feature points, the algorithm end.