Variational Based Image Inpainting Methods by using Cellular Neural Networks

摘要

Image inpainting is an interpolation problem where an image with missing or damaged parts is restored.rnThe most often used image inpainting applications are for pictures or films known or damaged partially. Discardingrnsome unwanted parts, text or objects from the whole image space, special effects can be carried out using imagernrestoration.rnComplex mathematical models based on partial differential equations (PDE) or variational computing were proposedrnas techniques for restoring damaged or partially known images. Those methods are computational expensive andrndifficult to implement, even when a large serial processing computing power is available.rnThe Cellular Neural Networks (CNN) based parallel processing ensures computing-time reduction if the processingrnalgorithm can be implemented on a continuous-time analogue CNN-UM (Cellular Neural/Nonlinear NetworksrnUniversal Machine) or using FPGA implemented emulated digital CNN-UM. Even if variational computing methodsrnare used, the design of CNN templates ensuring the desired processing of the gray-scale image remains anrnimportant step.rnIn the present paper, some variational based CNN methods are presented and analyzed that can be used for thernreconstruction of damaged or partially known images. Efficiency of these impanting methods can be enhanced byrncombining them with nonlinear template that ensures the growth of the local properties spreading area along withrnregional ones.