Blind Restoration Of Defocuesd Images With Low-snr

Target image is a kind of image with obvious edge features,which is often used in visual measurement to obtain the appearance shape,spatial position,rotation angle and other information of large components,so as to realize the precision measurement of 3D geometric parameters of large components.However due to the complexity of industrial environment and the optical imaging system performance constraints,image is prone to be defocused with low SNR by noise and out of defocus in image acquisition process,which floods its morphological characteristics,influences subsequent segmentation of target recognition and pose estimation results,confines the enhancing of visual measurement precision.The existing image blind restoration methods rely on mathematical deconvolution or iterative method,and the processes are sensitive to noise and initial iterative value,resulting the restored image edge and details blurry,or even ringing effect,which are not conducive to the effective extraction of target image information.Therefore,the blind restoration method of defocused image with low SNR is studied in this paper and applied to the blind restoration of target image to reduce the impact of noise on the blind restoration of target image,meanwhile enhance the edge features of target image,and realize the high-quality restoration of target image,which is used in the high-precision positioning and tracking of large components.In this paper,on the basis of the analysis of image degradation model and the influence of noise on image blind restoration,an improved blind restoration method for defocused image based on blur map is proposed.First,the anisotropic partial differential equation is used to preprocess the image to reduce the influence of noise on the blind restoration while keeping the image edge clear.Secondly,the degenerated image is produced by re-blur processing with re-blur amount estimated automatically in the scale space,and the blur map is obtained by image difference ratio and guided filtering.Then,the latent image is restored preliminarily based on the blur map and the defocusing model.Finally,based on the maximum posterior probability criterion,a regular constraint equation is established to optimize the blind restoration image,and the numerical iterative method is used to solve the problem.Experimental results show that the proposed blind restoration method for defocused images can effectively suppress the ringing effect,enhance the image edge features,obtain clear target images,and achieve high quality restoration for defocused images with low SNR.