A Contour-Based Multi-scale Vision Corner Feature Recognition Using Gabor Filters

This paper proposes a corner detection algorithm based on the correlation matrices, in which the combination of edge shapes and gray variations in multiple scalars are used. First, a Canny edge detector is used to detect the edge contours of an input image. In each scale, the direction derivative of each pixel on the edge curves and its surrounding pixels are extracted by using Gabor filters with imaginary parts (IPGFs), which are further used to construct the correlation matrices. Then, the sum of normalized eigenvalues of the correlation matrices at different scales is computed to extract potential corners. Finally, non-maximum suppression and a threshold are used to extract final corners. The experimental results show that the proposed method improves the detection accuracy and is noise resistance compared with Harris algorithm.