A method to avoid zero convergence of confidence term on exemplar based inpaintings

Abstract

A method to improve the behavior of zero convergence of confidence terms on exemplar based inpainting approaches. If the target area of inpainting is significantly large, the confidence values are conversed to zero near the center and it results a randomized selection of patches. It degrades the inpainting results. A patching error based confidence compensation method has been proposed. Error between selected patches and known area of target patch is calculated and a new confidence terms is defined based on the error to protect zero convergence. Several experiments were conducted to verify the effects of the proposed method. The experimental results show that the proposed method can protect the zero convergence and stabilize the inpainting procedures.