Image Reconstruction And Restoration Algorithm Based On L0 Norm And TV Regularization

In recent years,sparse signal processing has become a hot research topic.As a special two-dimensional signal,image is widely used in real life.In the generation and transmission of images,it is necessary to consider the transmission cost of the image and the possibility of damage.Therefore,it becomes meaningful to reconstruct and restore the image.Based on this background,the theories of image reconstruction algorithm based on L0 norm and image restoration based on total variation regularization are studied,and some new algorithms are proposed respectively.First of all,aiming at the problem that the "steepness" of the smoothing function in SL0 series of algorithms in compressed sensing can affect the performance of image reconstruction,this paper constructs a higher "steepness" function named composite inverse scale function,combining the steepest descent method,gradient projection principle and the residual mechanism in OSL0 algorithm,and a new algorithm named MSL0 algorithm is proposed.At the same time,the search path of the steepest descent method in the optimization algorithm is "zigzag",which affects the convergence speed and the calculation of Newton method is complex which is time-consuming.In this paper,a new nonlinear conjugate gradient method is used to replace the steepest descent method in MSL0 algorithm and MCGMSL0 algorithm is proposed.Using these two algorithms for image reconstruction,experimental simulations show that both the MSL0 algorithm and MCGMSL0 algorithm can greatly improve the effects of image reconstruction as well as very little reconstruction time and has good performance.Secondly,in order to solve the problem that the CTVL1 model uses the first-order TV regularization term,making the restored image may have the "ladder" effect and affect the restore effect,the second-order TV regularization term is introduced into CTVL1 model,and the HOCTVL1 model is established in this paper.The model is solved by ADMM,and the HOCTVL1 algorithm is given.Meanwhile,the fixed value of the regularization parameters in the model may affect the details of image restoration.In this paper,the SAHOCTVL1 model is established and the SAHOCTVL1 algorithm is given by combining the spatial adaptive regularization term in the HOCTVL1 model.The evperiment results show that the HOCTVL1 model greatly improves the effects of image restoration and has good performance;the SAHOCTVL1 model further improves the effect of image restoration,but it needs more time.Finally,the algorithms proposed in this paper are applied to remote sensing image reconstruction and restoration.Through a large number of experiments and simulations,as well as comparisons with other algorithms,the excellent performance of several algorithms proposed in this paper is verified,and also proves that the algorithm proposed in this paper has a certain practical application value.