High‐resolution images from compressed low‐resolution video: Motion estimation and observable pixels

In this article, we address the problem of obtaining a high‐resolution (HR) image from a compressed low‐resolution (LR) video sequence. Motion information plays a critical role in solving this problem, and we determine which pixels in the sequence provide useful information for calculating the high‐resolution image. The bit stream of hybrid motion compensated video compression methods includes low‐resolution motion‐compensated images; we therefore also study which pixels in these images should be used to increase the quality of the reconstructed image. Once the useful (observable) pixels in the low‐resolution and motion‐compensated sequences have been detected, we modify the acquisition model to only account for these observations. The proposed approach is tested on real compressed video sequences and the improved performance is reported.