Research On Remote Sensing Image Reconstruction Based On Edge Method

The remote sensing image has become the important data resource of spatial geographic information. But due to space radiation of the universe, the temperature gap between day and night, the complicated atmospheric environment, and many other factors, the remote sensing image is distorted in the process of acquisition, transmission and storage. The degradation of remote sensing image seriously affects the application of remote sensing data. Therefore, it is particularly important to do research on remote sensing image reconstructionAt present, most image reconstruction algorithms need to know certain prior knowledge of imaging system and even some of them need to know the parameters of satellite sensor platform and optical systems, etc. But it is difficult to obtain the prior knowledge in some cases, it has limited the serviceable range of these reconstruction algorithms and the application of remote sensing data. As a consequence, it becomes an urgent event to study reconstruct image with less or no prior parameters. This article has carried out the research work of the following aspects in order to solve these problems that exist in the process of remote sensing image reconstruction.(1) Summarize the existing measuring methods of point spread function of imaging system. Put forward a new kind of improved method to select the edge on the basis of traditional edge measuring methods, which can better fit the point spread function. The method combines the Canny operator and the Hough operator, barely contains artificial factors, and extracts adaptively the edge from image to fit the point spread function to a certain extent. It has single response for the edges in image, high detecting accuracy, and is in favor of fitting the point spread function precisely.(2) Fit the edge in remote sensing image by using the least squares fitting and calculate the distance of pixels to the edge. Fit the curved of edge spread function by using Fermi function with the distance, and derive the Fermi function to obtain the line spread function. Calculate and fit the point spread function of imaging system with its property of separable variables.(3) Establish the imaging model and reconstruction model of remote sensing image, and analyze the factors of remote sensing image quality degradation through whole link. Propose a reconstruction algorithm based on total variation regularization, which is applicable to not only the anisotropic but also isotropic forms of total variation discretization. The algorithm has convergence properties with finite convergence for some variables, and a continuation scheme to accelerate the practical convergence of it.(4) Put out a novel algorithm based on Hough transform to judge whether the fuzzy type of image is defocus blur or motion blur. With the method of single-edge fitting point spread function, further, double-edge, multi-edge and curved-edge are proposed and developed to improve the fitting precision.(5) Simulate the reconstruction of different remote sensing images with defocus blur and motion blur. Determine the fuzzy type of the image, fit the point spread function with single-edge, double-edge, multi-edge or curved-edge, and reconstruct the remote sensing images via total variation regularization algorithm.(6) Summarize all the work the article has done, put forward the deficiencies, and give a prospect to the further research.The article proposed an improved reconstruction algorithm with information of image itself and no parameters of remote sensing imaging system, based on knife-edge method, Canny operator, Hough transform and total variation regularization method, from the idea of measurement of point spread function and model establishment of imaging and reconstruction. It has verified the feasibility of the algorithm theoretically and demonstrated the validity of the algorithm through simulations.