Fast restoration of star image under dynamic conditions via l_p regularized intensity prior

This paper is an in-depth look at the problem of removing the blur from a complex motion-blurred star image. Accordingly, a simple yet effective l_p (0<p≤1)-regularized deblurring method based on stars image intensity is proposed. The model builds on the principle that the intensity of clear star image is in accordance with Laplacian distribution or generalized p Gaussian distribution. Further, two algorithms are introduced to solve the ensuing non-smooth (p=1) or non-convex (p<1) constrained optimization problem. Simulations and actual star image restoration experiment are implemented to demonstrate that the centroids extraction accuracy of the proposed method is higher than 0.1 pixel, the running time is 3 to 5 times better than Richardson–Lucy filter or other methods based on image gradient constraint, and the peak signal to noise ratio (PSNR) of recovered star images excel results of several other image deconvolution methods.