Image inpainting based on hidden Markov random field

Abstract

In order to solve the problem of mismatching and structure disconnecting for exemplar-based image inpainting, we study the problem from the exemplar of known and unknown patches two part of the neighborhood, the introduction of the corresponding information consistency criteria from the matching patch and the unknown region to be repaired patch. Take full account of the degree of similarity to be repaired patches unknown information about the neighborhood and the matching patch. established hidden Markov random field image inpainting model, Be repaired area of the neighborhood be regarded as hidden random field, matching patches are determined by similarity and consistency, Effective solution to the mismatch of the exemplar inpainting method and structure of discontinuous problems, Experimental results show that our method is effective and practicable.