Research On Blind Deconvolution Of Space Object Images

Space object images obtained through ground-based telescopes are often heavilydegraded by atmospheric turbulence, which affects the identification, tracking and posejudgement of the space objects. Adaptive Optics (AO) is the main method to reduce theinfluence of atmospheric turbulence, but its correction is usually partial due to hardwarelimitations of the AO system. In order to obtain high resolution images of space objects,AO-corrected images must be further restored by post processing techniques. To meetthe need of post processing interested space object images, this paper focuses on theblind deconvolution techniques, modifying two image blind deconvolution algorithmsas follows:1) Considering space objects are commonly extended targets, asymmetriciteration is introduced into standard Richardson-Lucy (RL) deconvolution algorithmunder the framework of iterative blind deconvolution (IBD), and correspondingasymmetric factor (AF) is deduced to be updated automatically. Based on imagingprocess of space object images, a new method is proposed to estimate the initial pointspread function (PSF). In order to improve the algorithm’s convergence rate, vectorextrapolation method is adopted. Finally, the algorithm is extended from single framedeconvolution to multi-frame case.2) For image blind deconvolution algorithm based on maximum likelihood (ML)strategy, strict non-negative constraints are imposed on images and PSFs, and thecorresponding AF is also deduced to be updated automatically when image noise obeysPoisson distribution. The new method for estimating initial PSF is also adopted, and themethod for estimating PSF support is further modified to improve its adaptability.Finally, the algorithm is extended to multi-frame case as well.The results of blind restoration of simulated and real images show that the twomodified algorithms can effectively restore space object images, and greatly improveimages sharpness.