Research On Blind Restoration Model Of Multi-frame Solar Speckle Images And Its Parameter Selection Method

Solar speckle image reconstruction is an important research issue in the field of astronomical observation.Astronomical images acquired through ground-based optical telescopes will be affected by atmospheric turbulence and atmospheric disturbances,and will be seriously blurred or degraded,and need to be reconstructed using image restoration methods.Image restoration can effectively overcome the influence of atmospheric turbulence,realize the realistic image of the target close to the telescope.So far,the conventional image blind restoration algorithm is regularization.Most of the existing methods deal with a single image.The more image features are,the better the reconstruction effect will be,while the sparse and single features of the single solar speckle image will lead to the poor quality of the reconstructed image.In the regularization method calculation,the parameter value is too large or too small will affect the restoration time and restoration effect,the parameter plays an important role.In this paper,we based on a single frame regularized blind restoration model,combined with the complementary relationship between multi-frame solar speckle images to establish a multi-frame blind restoration model suitable for solar speckle images reconstruction,and combined L-like curve method and genetic algorithm to calculate the regularization parameters in the model,we also took the calculated results as the initial values of the model parameters in this paper to realize the blind restoration of multi-frame solar speckle image.The main work of this paper is as follows:(1)Combined with the current research background,three kinds of single frame regularized blind restoration models are described and analyzed in detail.The limitation of the single frame regularization model for the reconstruction of the solar speckle image is found through experiments,and a multi-frame blind reconstruction model is constructed by using the complementary relationship between the sequence images as a new constraint.We used the data provided by the Yunnan Observatory to experimentally verify the effectiveness of the method.(2)For the model in this paper,the commonly used regularization parameter selection methods are studied.In view of the shortcomings of the traditional regularization parameter calculation methods,the l-curve method is mainly analyzed.We adopt L-like curve and genetic algorithm to choose the parameters in this model.It improved the correctness of regularization parameter selection and sped up the convergence speed of algorithm iteration process.