Study On The Algorithms Of Registration And Restoration For The Adaptive Optical Image Sequence

When imaging through the atmosphere, the obtained images of the objects are inevitably influenced by the atmosphere turbulence. The observed images are heavily blurred and degraded by the wave-front aberration. This problem will be encountered in the imaging process by both the ground-to-space and the space-to-ground system. The adaptive optics (AO) technology is one of the most effective and promising methods for overcoming the atmospheric turbulence. The AO system, however, cannot compensate the atmospheric turbulence completely. The high frequency information of the object is still convolved and overlapped. Consequently, in order to get the clear images of the object, the observed images which have been rectified by AO system are still need post-processed with the technology of image restoration. Mainly aiming at the imminent need of imaging post-process of ground-based AO telescope, this paper applies the technology of image processing to restore high-legible imageThis paper pays main attention on the algorithms of image registration and multi-frame restoration for the AO images. The main works are as follows:1,The imaging degradation model and the image restoration algorithms for the AO images are summarized. The degradation model, PSF (Point Spread Function) and noise model of the AO images are analyzed in detail. The conventional image restoration algorithms are classified and introduced. Some evaluation indexes of image restoration quality are introduced.2,The pre-processing techniques for the AO sequences are studied. The frame selection, denoising and the support region extraction methods for AO image are separately discussed. Firstly, the importance of the frame selection for the multi-frame restoration is analyzed. The selecting criterion of the reference image, which considering both the grades and the Shannon entropy is proposed. Combining with frame selection, the quality of the restored image is improved greatly. Secondely, the ill-posed problem induced by the noise is analyzed and the TV-based denoising algorithm is performed. At last, considing the characteristics of the AO image, the support region extration algorithm based on the segmentation is preseted, and the extration results with real observed images show the validity of the algorithm.3,The co-registration among the AO sequence is studied and performed. Registration among the multi-frames is needed for making use of the complementary information of the images.The registration algorithm based on the scale invariant feature is used for the registration of turbulence degraded images. Then the experiments with simulated and real observed images are given. Experiment results show the superiorities of this algorithm.4,The anisotropic regularization based multi-frame iterative blind deconvolution algorithm is presented. Aiming at the shortcomings of the multi-frame iterative blind deconvolution (MFIBD) algorithm proposed by Zhulina, a modified multi-frame iterative blind deconvolution algorithm is proposed. This algorithm is not only simple but also easily realized. It is the application of anisotropic regularization that a better restored image can be obtained even when the SNR of the observed images is low. The quality of the restored image is efficiently improved. The restored results with the simulated and real observed images are clearer and show more details of the object.