Enhanced digital image inpainting Models using discrete shearlet Transform

Abstract

Digital image inpainting is the technique of filling the missing newlineregions of an image by using information from surrounding area This newlinetechnique has wider applications in image restoration disocclusion and newlineimage video compression newlineThe objective of this thesis is to implement novel approach of newlineintroducing Discrete Shearlet Transform DST in digital image inpainting newlinemodels and applying them to the problem of text removal image newlinereconstruction and image video coding In this regard this thesis addresses newlinethree significant image inpainting models with error concealment newlineapplications newlineThe inpainting model is proposed by introducing DST and p newlineLaplacian operator in Total Variation model This model with 1 p 2 can newlinereduce the staircase effect by still keeping the sharp edges effectively The p newlineLaplacian operator diffuses in two directions and hence diffusion speed newlineincreases newlineThe second inpainting model is proposed with the idea of using newlineExpectation Maximization EM algorithm in a Bayesian framework with newlineshearlets The EM algorithm iteratively reconstructs the missing data and then newlinesolves the equation for the new estimates newline newline