Research On Image Blind Restoration Based On PSF Optimization Estimation

Satellite images produce degradation phenomeno which is influenced by hardware equipment and the environment in the process of imaging.It seriously affects the extraction of image information and application.Image restoration technology which uses the prior knowledge of the image achieves the restoration of degraded information,it is the basic problem in the field of image processing.Image restoration mainly has the following problems.The point spread function(PSF)is usually unavailable in the process of imaging,PSF is difficult to estimate;Image restoration model is difficult to construct in the coexistence of blur and noise.Image restoration is the solution of inverse problem,its own ill-posed problem is serious.The restoration model is difficult to solve,when it is non-convex.To solve above problems,the following studies were conducted:Aiming at the problem which is difficult to estimate PSF template,and analyzing the imaging features of the approximate point feature,a PSF estimation method of elliptic paraboloid model is proposed which is use the imaging characteristics of approximate point features in the image.The equation of model and the solution are given,and obtained the PSF initial template of optimal estimation.The feasibility of the method is verified by experiments.Aiming at the effect of noise in the image restoration,a ratio norm sparse constraint regularized restoration model is constructed.Which achieves effective separation of image signals and noise,and removing blur of the image while suppressing image noise.Aiming at the characteristic of non-convex restoration model,a complete optimization solution and parameter configuration are given.The experimental results show that the method can well overcome the problem of ill and noise sensitive,and it effectively improve the qualityof image.On the basis of the above research,the algorithm is developed and implemented in the Super Resolution Reconstruction System(SRR1.0).