A novel image matching algorithm using local description

As we know, the problem of image matching is difficult and important in the field of computer vision. In this paper we present a novel matching algorithm based on local invariant feature description. Firstly, feature points are detected by difference of Gaussian. Secondly, the Haar-wavelet responses within a feature point neighborhood are projected into four directions, and then a 64-dimensional vector is generated for describing the feature point. Finally, matching pairs are determined by using the nearest neighbor distance ratio. Experimental results show that the proposed algorithm is not only rapid and robust, but the matching rate is higher than PCA-SIFT and SURF algorithms.