A Fast Automatic and Robust Image Registration Algorithm

摘要

In this paper, a fast automatic and robust image registration algorithm is presented. First, we use the Fourier transform to calculate the relative position between the input images and sort the unordered image sequence. For two images with overlap region, we detect the feature points in the part of the region in one of the image with the robust method of SIFT. According to the coordinates of the feature points and the position relationship of two images, we can compute the space of the feature points in another image and detect the feature points in this region. Then, we use the geometric relation to reduce some mismatch points with structurally similar and achieve the robust image registration. We demonstrate illustrative results obtained by comparing and contrasting our output with other methods. This method can detect the effective feature points quickly for reducing the detection range of the feature points without human intervention and reduce the running time greatly in ensuring the feature points matching accuracy.