A semi-blind deconvolution method to estimate and remove optical aberrations

Optical aberrations (ABRs) always hinder high-quality imaging, and designers make a vast effort to correct them. We propose a computational method to estimate and remove an ABR that could blur an image. It works upon the combination of an elementary blind deconvolution process and finding the best correlation between the estimated point spread function (PSF) and a pre-defined basis set of ABR-related PSFs. The process is helpful to approximately extract a non-uniform PSF of an ABR and a sharp image from a single blurry one. We successfully tested the method for a dominant effect of defocusing, spherical, coma, astigmatism, or field curvature in different pictures. All the MATLAB codes are available in the Supplementary Material.