Hybrid Directional Weight-based Demosaicking for Bayer Color Filter Array

Abstract

This paper proposes and improves gradient estimation techniques in order to suppress the common demosaicking artifacts and introduces hybrid directional weights for adaptive directional interpolation. The hybrid weights are calculated using color correlation and intensity differences, and these weights apply to the whole interpolation process. Experimental results demonstrate that the proposed algorithm outperforms recent demosaicking methods on 42 images of Kodak and McM datasets.