Reasearch On Image Restoration Of Low Orbit Space Objects

Image restoration is widely used in aerospace field, the imagingprocess inevitably bring distortion effects to the original image. As spaceobjects are with high rate of movement and the observation equipment iswith limited performance, during the observation of space objects, theobserved images are blurry, and we do not know exactly how the blurryeffects happen. The restoration process is indeed a blind deconvolutionprocess with high complexity. The restoration process of space objects issimilar to traditional deblurring methods, but it is more complicated, weshould use more accurate restoration algorithms to solve it.We illustrate the ill-poseness of the deconvolution problem throughthe basic degradation model. We analyze different types of blur imageswith their corresponding mathematical models. We investigate sometraditional restoration methods with their performance evaluation.We analyze the ROF model used in the deblurring process with itsmathematical properties. We used the FISTA and Augmented Lagrangianmethods to optimize the ROF model and investigated the detailedoptimization process. We finally compare the deblurring results by ROFmodel with traditional methods.We present a novel blind restoration algorithm. We analyzeshortcomings of traditional regularization terms. Many traditional types ofregularization terms do not have an appropriate description of the latentsharp images. We use a new type of sparse regularization term that hasbetter description of the latent clear images. In this paper we use a newtype of regularization term that has a better description of sharp imagesthan traditional ones. Additionally, this paper presents a blur kernel refinement process with a two phase kernel estimation strategy. Theprocess will iteratively eliminate the noise of the inaccurate blur kernel andkeep the support region of the kernel. It is helpful to separate the kernelestimation process and the non-blind deconvolution process. Finally thisalgorithm will produce a better deblurring result using the optimized blurkernel.