Corner Detection Using Multi-directional Gabor Filters

Exploiting the characteristics of corner, an information entropy-based corner detection algorithm using the multi-directional Gabor filters is proposed in this paper. Different from the methods which detect corners by analyzing plane curve contour shape or finding for local maxima of curvature, the proposed method combines with the gray level changing information at edge contour pixels and pixels around the contour pixels to find corner . Firstly, the Canny edge operator is used to extract edge map and the gaps in the edge map are filled. Secondly, the imaginary parts of Gabor filters are used to smooth the edge pixels and their surrounding pixels along multi-directions. Finally, the gradient direction information entropy at the edge pixels is used to detect corners. Experimental results show that the proposed algorithm attains better detection performance, higher localization accuracy and noise robustness than the existing several algorithms.