基于高斯差分滤波和形态学滤波的Harris角点检测算法

Harris算子及其改进算子(如Harris-Laplace算子)是当前应用广泛的角点检测算法，然而它们都存在冗余点多和抗干扰能力差的缺点。基于此，提出一种新Harris算子改进算法。将图像进行高斯差分滤波，采用多层差分平均值增强稳定性和抗干扰性; 对差分图像进行灰度形态学滤波，进一步去除小于结构元素的波谷，波峰值保持不变，这样不仅可以减少冗余点，还可以提高抗噪能力。实验表明，该方法在抗噪和去冗余点方面明显优于Harris算子和Harris-Laplace算子。  

Research on Harris Corner Detection Based on Gaussian Differential and Morphological Filter

Harris operator and improvement operator （such as Harris-Laplace operator） are currently widely used in corner detec- tion algorithms, however, they exist redundant points and poor anti-interference ability. So, a new improved algorithm of Harris operator is proposed in this paper. First, difference of Gaussian filter is used in image filtering, and average of multilayer difference is used for improving anti-interfere ability and increasing stability. And then the difference image is filtered by grayseale morphological filtering. Morphological filtering further removes the troughs which are less than structural elements and make the wave peak remain unchanged, so redundancy points are reduced and noise immunity is improved. Experiment shows that the method is superior to Harris and Harris-Laplace in noise immunity and redundant points.