Image registration in hough space using gradient of images

Abstract

We present an accurate and fast method for rigid registration of images with large non-overlapping areas using a Hough transformation of image gradients. The Hough space representation of gradients can be used to separate estimation of the rotation parameter from the translation. It also allows us to estimate transformation parameters for 2D images over a 1D space, hence reducing the computational complexity. The cost functions in the Hough domain have larger capture ranges compared to the cost functions in the intensity domain. This allows the optimization to converge better in the presence of large misalignments. We show that the combination of estimating registration parameters in the Hough domain and fine tuning the results in the intensity domain significantly improves performance of the application compared to the conventional intensity-based multi-resolution methods.