A Novel Algorithm for Visual and IR Image Registration Based on Hough Transform

In order to solve the problems of visual and IR image registration, a new registration method for visual and IR image pairs is presented in this paper. As intensity discrepancy of visual and IR images, straight lines extracted by Hough transformation are used as registration primitives because these features are common to both images. The similarity transformation is considered and the parameters are estimated orderly. The rotation angle is estimated by ID angle correlation of extracted lines in both images; the scale parameter is calculated by area ratio of triangle, which composed by the intersections of straight lines; finally the translation parameters are resolved. Clustering method is used for the computation of scale and translation parameters. The proposed registration algorithm has been tested on visual and IR image pairs of buildings, and its effectiveness has been verified by the experiments.