A Novel Method for Exemplar-based Image Inpainting

摘要

A novel exemplar-based image inpainting algorithm is presented in this paper, in which optimal exemplar is determined by Gaussian weight Euclidean distance in view of not only the differences between patches but also space distance. The new method of computing confidence term when a damaged block was repaired is introduced, which improves the priori computation precision. Furthermore, a damaged front searching method within the adjacent region of repaired block is used to narrow the search area and reducing the running time. Experimental results show that the proposed method achieves better inpainting effect and saves the necessity of the damaged front selection and improve inpainting efficiency.