Refined Exponential Filter with Applications to Image Restoration and Interpolation

Abstract

Ill-posed linear equations are pervasive in computer vision. A popular way to solve an ill-posed problem is regularization. In this paper, we propose a new criterion for designing the regularizing filter. This criterion reveals the implicit assumption made by regularizing filters. Then with the help of the discrete Picard condition, we refine the exponential filter using our criterion. The effectiveness of our method is demonstrated on image restoration and interpolation.