A GABOR FILTER-BASED FEATURE POINTS MATCHING APPROACH

In this paper we propose a Gabor filter-based feature points matching approach. Gabor filters of different orientation are adopted to perform multi-orientational filtering on feature points that are to be matched, and vector elements of N-dimensional normed linear space are constructed by sub-image standard deviation of N-dimensional orientations. The similarity between feature points is described by 'distance' of two feature point vectors, thus it achieves matching between homogeneous feature points. The experiment results show the method is not sensitive to noise added in image and can be adapted to match images with great different gray scale levels. Moreover, the method has no special requirements on feature points.