Search

• A prior analysis method of remote sensing image characteristics is presented based on hyper-Laplacian regression. • A novel regularization model is established by taking the characteristic prior parameters and image together as the optimization target. • The solving approach is presented based on the alternate iteration of damping least square and iterative jumping thresholding algorithm. • A fine-processing method based on the adaptive hyper-Laplacian prior is put forward, combined with the idea of block processing. High-quality remote sensing images have wide application prospects in agroforestry investigation, target monitoring, disaster prevention, urban planning and military defense. However, remote sensing imaging links such as atmosphere, platform and optical system seriously affect the ability of image interpretation and analysis. The traditional regularized processing methods have a strong ability to improve the definition, but most of them may sacrifice texture details or introduce artifacts, because their fixed prior parameters cannot fully adapt to various kinds of scenes. To address this problem, we propose a novel fine-processing method based on the adaptive hyper-Laplacian prior for remote sensing imaging systems. The method is developed by automatically updating and optimizing the prior parameters and objective function in the iterative process based on the prior characteristics of different regions of remote sensing images. Experimentally, the proposed method can realize the fine-processing of remote sensing images, including the edge enhancement, texture detail preservation, and artifact suppression. [ABSTRACT FROM AUTHOR]