The Application Of Augmented Lagrangian Method Based On Euler's Elasticity Model In Image Restoration

In actual production and life,the images which people obtain are often contaminated and require image restoration.For image restoration,it is necessary to establish a degenerate model that causes image quality degradation,and then the reverse process is used to restore the original image.Some criteria is applied to determine the effect of image processing.Now,Euler's elastic model has been widely applied to image restoration and good results are achieved.In this paper,an Euler's elastic model that combines the L1 and L2 fidelity terms is imposed to remove the Salt&Pepper-Gaussian mixed noise.The genetic algorithm is used to select the parameters in the new image restoration model.The numerical experiments show the superiority of the new model.By comparing with the advantages and disadvantages of the ROF model,the harmonic model and the Euler's elastic model,and comprehensively and deeply analyzing the characteristics of the L' and L2 fidelity terms,the model of fusing L1 and L2 fidelity terms is established for denoising of the common Salt&Pepper noise and Gaussian noise of the image.The new Euler's elastic model is solved by the augmented Lagrangian algorithm.Numerical experiments exhibit that the new model not only can remove Salt&Pepper noise and Gaussian noise,but also avoid the staircase effect to a certain extent,maintain the edge detail information,and show great superiority.For the new image restoration model,many artifacts need to be determined during the solution by the augmented Lagrangian method.Aiming at this kind of problem,an algorithm of parameter selection based on genetic algorithm is put forward.For the Euler's elastic model with fusing L1 and L2 fidelity terms,the corresponding parameters which are selected by genetic algorithm are compared with the traditional empirical parameters.The experiment results show the method of the parameter selection is better.