Multi-modality image registration using the Hough transform

Abstract

We present a method for finding relative translation and rotation between images of different modalities, for cases in which external contours can be approximated by ellipses. The Hough transform is used to map the feature space to the parameter space. The feature space consists of the pixel coordinates and gradient direction of all possible edge points, and the parameter space defines all acceptable ellipses. The difference between ellipse parameters, calculated separately for each image, is used to determine the relative translation and rotation. The images are assumed to be corrected for scale. The method is tested on simulated images and on pairs of actual PET/MRI images of the brain.