Super-resolution from Multiple Images Having Arbitrary Mutual Motion

1. INTRODUCTION Video sequences usually contain a large overlap between successive frames, and regions in the scene are sampled in several images. This multiple sampling can sometime be used to achieve images with a higher spatial resolution. The process of reconstructing a high resolution image from several images covering the same region in the world is called Super Resolution. Additional tasks may include the reconstruction of a high resolution video sequence [Elad and Feuer, 1996], or a high resolution 3D model of the scene [Smelyanskiy et al., 2000]. A common model for super resolution presents it in the following way: The low resolution input images are the result of projection of a high resolution image onto the image plane, followed by sampling. The goal is to find the high resolution image which fits this model. Formulating it in mathematical