Blind Deconvolution Restoration Of Atmospheric Turbulence Degraded Images Based On Total Variation And Sparse Regularization

Astronomical observation is an important means for human exploration of space.Ground-based telescopes with large diameter can provide higher diffraction limit resolution,but due to the influence of atmospheric turbulence,the practical resolution of telescopes is far below the diffraction limit.The adaptive optics system can effectively correct the wavefront distortion caused by turbulence.However,owing to the fitting error caused by deformed mirror,time-bandwidth error and anisoplanatism limit,the correction is partial and incomplete and there is still residual error in the closed-loop image,hence,post image processing technology is needed to further improve image resolution.This paper is actually aimed at this demand,focusing on atmospheric turbulence-degraded images blind restoration based on blind deconvolution technology,to improve the resolution of adaptive optical closed-loop images,and provide technical support for high-resolution astronomical observation.The main research contents and innovations of this paper are as follows:Firstly,to study the influence of atmospheric turbulence on the imaging quality of telescopes,the imaging model of telescope optical images is analyzed.The Zernike polynomial is used to simulate the distortion wavefront under atmospheric turbulence,the atmospheric turbulence point spread function is generated and degraded images are obtained according to the degradation model,which provides the image basis for the subsequent image blind deconvolution algorithm.Meanwhile,traditional Wiener filtering and Richardson-Lucy based blind restoration algorithms and their extended method are studied.Simulations show that classical method perform defects in convergence,speed,noise robustness,time cost and detail lost.Secondly,to cope with the lost-of-detail problem of traditional Wiener filtering blind restoration method,this paper analyzed the edge-reserving ability of total variation based on image denoising research.And total variation blind deconvolution is introduced to the restoration of atmospheric turbulence-degraded images.Details and edges are successfully restored,which improves the quality of the restoration.Meanwhile,to obtain more accurate kernel estimation,the two-stage estimation strategy is adopted.To solve the numerical difficulty caused by total variation,the split bregman optimization algorithm is employed to ensure stable convergence,simulations show that successful restored object from degraded images.Thirdly,owing to the fact that the image prior of traditional algorithm cannot truly describe the characteristics of the natural image gradient distribution,the image gradient sparse prior originated from motion deblurring is introduced into the blind deconvolution restoration of the atmospheric turbulent optical degradation image.Simulations show that the gradient sparse prior(i.e.l_p norm of gradient)is beneficial to reduce the kernel estimation error,improve the image restoration quality and fully reconstruct the details and contours of the object.Besides,to illustrate the value of p's influence on the restoration quality,this paper proposed the principle of selection of p,that is,high quality reconstruction can be obtained when p is between 0.3 and 0.7.Fourthly,the influences of different aberrations on image restoration qualities is studied,research show that edge aberration of Zernike pyramid has less influence on image restoration,while the higher-order aberration of pyramid near-axis region has greater influence.For the recovery difficulties caused by higher-order aberrations,research of restorations of higher-order aberrations under different Zernike-RMS conditions is carried out.Simulations show that the restoration algorithm can still perform high-quality reconstruction when Zernike-RMS is below a certain threshold.