Motion estimation in high resolution image reconstruction from compressed video sequences

In order to obtain a high resolution image from a compressed video sequence it is essential to correctly estimate the motion vectors in the sequence. Most of the approaches reported in the literature address this problem using standard motion estimation techniques. In this paper we tackle the correct estimation of the motion vectors by consistently estimating the optical flow across multiple images. Consistency is achieved by adding a regularization term to the classical Lucas-Kanade approach to motion estimation. The proposed algorithm is tested on real video sequences.