An Eigenspace Approach to Multiple Image Registration

Multiple images can be indexed by their projections on a few \eigenfeatures". These eigenfeatures are the eigenvectors of a large covariance matrix, constructed from the images. It is known that registration is essential for computing \useful" eigenfeatures, and a preliminary step of putting the images in register is common practice. We propose to evaluate multiple image registration by the quality of their eigenfeatures. An algorithm that simultaneously registers the images and computes the eigenfeatures is proposed. The key idea is to iterate the following two steps: 1. Eigenfeatures are computed from the images. 2. New images are computed by registering the images on the eigenfeatures subspace. In the next iteration Step 1 is applied to the set of images that were most recently computed in Step 2. The implementation of this simple procedure requires a novel deenition of registration on an image subspace. It is demonstrated that the algorithm registers multiple images, and produces improved eigenfeatures.