Search Space Reduction in Exemplar Based Image Inpainting

Abstract

This paper aims at developing accelerated exemplary inpaint method. The feature set is considered to be the pixels along with their 8-neighbors. A Multi Phase Search Space Reduction framework namely Systematic Reduction of Information System (SRIS) is employed. SRIS, basically is a roughest based approach which imputes the missing values in an adaptive manner. In this approach the order of inpainting pixels is determined by a simple but effective priority term. The best exemplar is determined based on a similarity metric which is derived by element wise difference of informative pixels of inpaint window and the corresponding pixels of the source region window.