LSIFT—An improved SIFT algorithm with a new matching method

In terms of the high frequency of mis-match of SIFT method under the strong disturb and complex background, this paper proposes a new matching method based on comparing modulus. According to this method, the eigenvector limit is required at first before the normalization processing, and then, accurate matching pairs between two pictures can be found according to the difference between the modulus of the relative matching pair and that of the most similar one. The result of simulation indicates that the new matching method is of high stability and its matching time is much shorter.