Research On Regularization Image Restoration Methods For Linear Spacevariant Optical Imaging Systems

In actual situations, optical imaging systems, affected by aberration, inhomogeneous transmission medium, etc, are generally space variant. For the convenience of processing, traditional image restoration methods often approximated an imaging system as a space invariant system, thus its restoration effect for imaging systems of large space variability is poor. Therefore, it is necessary to study image restoration methods for space variant systems.First, acquisition of the point spread function (PSF) of space variant system was researched in this dissertation. Principal component analysis PSF model was used to represent the known PSFs and model parameters of unknown PSFs was obtained by interpolation, thus acquiring PSF of all fields of a space variant system. Four interpolation methods, namely, Polyfit, Inverse Distance to a Weight (IDW), Modified Sheperd’s (MS) and Radial Basis Function (RBF), were applied in this thesis, and were realized through Surfer10.0. An experiment was conducted with four space variant systems designed with the Unobscured Gregorian system in ZEMAX considering four degradation factors of defocus, and decenter, tilt and deformation of the primary mirror separately. Results manifested that, PSF mean square error of the adopted method attained10-6order; and precision of MS was the highest among the four interpolation methods.Secondly, combining the PSF acquisition method supposed in this paper and Laplace regularization method, a space variant Laplace regularization image restoration method (SVLRIR) was supposed. The regularization parameter selection problem of space invariant Laplace regularization image restoration method (SILRIR) and SVLRIR was discussed, and the results suggested that the best regularization parameter of SILRIR could be used to perform SVLRIR.Finally, two simulation experiments were carried on to validate the effectiveness of the space variant restoration method. One of the two experiments used the four space variant systems discussed previously, and the other one Hubble telescope. ZEMAX and TINYTIM was used to compute PSF of the corresponding space variant systems, then space variant blurred images were generated and restored with Matlab. Subjective evaluation method and ISNR were chosen to evaluate the restored images, and the results were analyzed.