A Super-Resolution Method with EWA%lliptical Weighted Average

This paper introduces a practical algorithm for super-resolution, theprocess of reconstructing a high-resolution image from low-resolutioninput ones. The emphasis of the work is to super-resolve frames fromreal-world image/video sequences which may contain significant objectocclusion or scene changes. As the quality of super-resolved imageshighly relies on the correctness of image alignment between consecutiveframes, the robust optical flow method is used to accurately estimatemotion between the image pair. An efficient and reliable scheme isdevised to detect and discard incorrect matchings which may degrade theoutput quality. The usage of elliptical weighted average (EWA) filter isalso introduced to model the point spread function (PSF) of acquisitionsystem in order to improve accuracy of the model. A number of complexvideo sequences are tested to demonstrate the applicability andreliability of the proposed algorithm.