Development of superresolution-based face video enhancement

A. V. Nasonov^a, A. S. Krylov^a, O. S. Ushmaev<sup>b

a</sup> M. V. Lomonosov Moscow State University, Faculty of Computational Mathematics and Cybernetics
^b Institute for Problems of Informatics RAS

Abstract: General superresolution-based method of face image enhancement for video data has been suggested. The superresolution is modeled as inverse problem to image downsampling, i.e., it finds an image that gives the minimal value of the quadratic discrepancy with initial low-resolution images after the motion dependent downsampling. High-quality superresolution method and fast superresolution methods are considered. Special deringing method for fast superresolution is proposed. New multiscale motion estimation method has been developed.

Keywords: superresolution; deringing; facial video sequence; multiscale motion estimation; fast superresolution.