Research On Super-resolution Reconstruction Method Of Remote Sensing Image

The super-resolution reconstruction technology can save a lot of manpower and material resources,and the resolution of the image has been greatly improved without changing the hardware.In actual research,according to the different processing methods,image reconstruction is mainly divided into two types,one is the reconstruction of a single-frame image,and the other is the reconstruction of a multi-frame image.Starting from these two types of reconstruction,the paper studies the super-resolution reconstruction of single-frame remote sensing image based on sparse representation and the super-resolution reconstruction of multi-frame remote sensing image based on total variation regularization.First,the super-resolution reconstruction of single-frame remote sensing images based on sparse representation is studied.To improve the algorithm proposed by Yang in [33],the algorithm takes the image of the natural scene as the research object and reconstructs it.In this paper,iterative back projection,sparse regularization parameters,and feature extraction operators are improved to make them more suitable for remote sensing image reconstruction.The reconstruction mainly includes three steps.One is to train the remote sensing image database to obtain high and low image block dictionaries;the other is to use the obtained low-resolution image block dictionary to find the corresponding sparse representation coefficients;the third is,according to The obtained sparse representation coefficient is operated with the high-resolution image block to obtain a reconstructed image.Comparing the experimental results,it can be seen that after improvement,the objective evaluation index of the reconstructed image is better.In order to solve the problem that the edge effect of the reconstructed image is not ideal after the traditional total variation regularization algorithm is reconstructed,this paper proposes a reconstruction algorithm that can enhance the edge information.This algorithm is mainly used to perform Butterworth low-pass filter preprocessing on the image after cubic spline interpolation when constructing the initial high-resolution image,and then perform edge sharpening processing.The regularization item selects the total variation for processing,and the data fidelity item selects the L1 norm for processing.It can be seen from the experimental results that the algorithm can further enhance the edge area of the image,and the improved objective evaluation index of the reconstructed image is better.