A PDE Constraint Adaptive-Exemplar Based Image Inpainting Model

A novel image inpainting model based on large target region exemplar is proposed.The model can preserve the linear structure and texture information of the image，and adaptively select exemplar based on local texture information.The model uses cross isophotes data term to constrain inpainting order.Cross isophotes data term is the result of partial differential equation （PDE） under the control of image geometry property.By taking into consideration of the extent of edge，the model shows a good linear structure preserving property.Diffusion of cross isophotes is morphologically invariant in the method，and the target region of natural scene could be more plausibly filled up.According to the feature of exemplar-based inpainting，Euclidean distance is used as similarity measure.To reduce seams caused by this model，a gradient constrained total variation （TV） interpolation is used in the inpainting result to obtain a seamless image.Both theoretical analysis and experiments verified the validity of the new inpainting model.