Image Processing Based On Higher Order Regularization

Digital image processing technology is one of the research hotspots in computer science.It is mainly engaged in the research of computer denoising,image extraction,feature restoration,enhancement and segmentation.Image processing technology is developing very fast and has a wide range of applications.It has rapidly expanded from the earliest aerospace field to biomedicine,information science and even nation-al defense and art.It has a great impact on human life.Therefore,the research of dig-ital image processing has always been the focus of research in the field of computer vision.In this paper we present a new image processing technique based on weighted high order regularization with non-quadratic fidelity.By the good property of weighted second order differential operator and L₁ norm,we first introduce a new weighted two order regularization term,which can capture more structures and details of images.By combining the new regularization term with non-quadratic fidelities,we propose a new image processing optimization model and implemented the new model for image denoising,image inpainting and image infusion.Although the computation is difficult due to the nonlinearity and non-diffe rentiability,in recent years,the augmented Lagrangian method has been proved to be an effective method for solving non-quadratic problems.So,augmented Lagrangian method is used for solving the optimization problem iteratively.For each iteration,the subproblems can be efficiently solved by linear systems or closed form solution.In addition,our algorithms have been discussed about influence of parameters and compared with several typical existing methods.Experimental results show that our method can further improve the restored results with more structures and prevent the staircase artifacts effectively.