Image Inpainting Study Based On Partial Differential Equation

Image inpainting is an important research field in digital image processing, which using the available surrounding information to fill in the damaged portions. It serves a wide range of applications, such as recovering the damaged regions, removing text and hiding some objects. The technique of image inpainting is used widely in many fields including repair of damaged medical image, ancient artifact and removing of scratches and spots in film, etc.This dissertation is mainly focused on the image inpainting methods that based on Partial Differential Equation. At first, the background and the development conditions of the image inpainting and its search conditions are introduced. Then the basic knowledge of the technique, classification of image inpainting and several typical image inpainting models are introduced, which including the Bertalmio-Sapiro-Caselles-Ballester model and Curvature Driven Diffusions model based on PDE, and total variation model based on the Bayesian and variational principles, different inpainting effect of each model are provided as well.The model that based on Partial Differential Equations can repair the image using the available surrounding information and diffusing. This algorithm can maintain the image edge as well as noise smoothing, but it is very complex to be used and the computing speed is slow. According to this issue, an improved method is adopted to simplify the model, and the anisotropic information diffusion is directly used to the image inpainting. Good experimental results are draw out through this method.