Example-based learning using heuristic orthogonal matching pursuit teaching mechanism with auxiliary coefficient representation for the problem of de-fencing and its affiliated applications

Abstract

Orthogonal Matching Pursuit (OMP) is a good candidate for solving energy function optimization problems. In this paper, we propose a novel auxiliary coefficient representation for the problem of image de-fencing. To improve the optimization efficiency of the OMP algorithm, we propose a heuristic form of the OMP (named h-OMP) approximation based on auxiliary coefficient representation. A frequency-domain optimization approach is derived by selecting an over-complete example set for the image signal, the h-OMP algorithm is used to simultaneously remove the fences on the image matrix and find the auxiliary coefficient basis to form the image segment. Experiments show that the proposed h-OMP algorithm generates better output image, whose performance is superior in terms of both subjective and objective evaluation criteria.