Super-resolution using the Walsh functions: a new algorithm for image reconstruction

Super-resolution techniques attempt to produce a high-resolution image from a sequence of shifted low-resolution images. The super-resolution problem has received recently a considerable attention within the image community. Spatial methods have proven to be more flexible and efficient comparing to the frequency methods. This paper introduces a new spatial approach for achieving image reconstruction from a sequence of images. The new approach we propose herein is based upon the representation of a signal by the superposition of members of a set of the orthogonal functions Walsh. We shall show how is it possible to retrieve the missing information in an image of resolution 2^M from another image shifted by a subpixel shift in order to reconstruct a new image with resolution 2^M+1. The shift is estimated using a new measure of sharpness. Some experimental results using synthetic sequences are presented in this paper to demonstrate the performance of this new method. We assume here only translation motion and no noise was considered.