Remote Sensing Image Fusion Algorithm Based On Spatial Resolution Enhancement

With the development of optical remote sensing imaging technology,different types of remote sensing image data are available with different spatial scales,different spectral ranges and different time phase information.Among them,due hyperspectral sensing images(HSI)have very important applications in the fields of environmental detection,disaster warning,urban planning,precision agriculture,and geological survey because of the high spectral resolution.However,due to the limitations of imaging equipment,existing hyperspectral imaging systems often cannot obtain data with both high spectral resolution and high spatial resolution.It limits the application of hyperspectral remote sensing in various fields.There are two main ways to improve the spatial resolution of hyperspectral images: on the one hand,the spatial resolution of spectral images can be improved by upgrading the hardware equipment of the imaging system,but at the same time it will also bring a sharp increase of imaging cost,which is undesirable;on the other hand,related image fusion algorithms can also be used to improve the spatial resolution without any increase of hardware cost,which has become the most commonly used method to improve the spatial resolution of hyperspectral images.According to different situations,this paper studies the fusion methods of multi-source and homologous data.The specific work of the thesis is as follows:(1)An improved smoothing filter-based intensity modulation(SFIM)image fusion algorithm is given.This method uses the least squares estimation solution(LSE)to minimize the linear regression coefficient of the spatial information error between HSI and MSI,and then use this linear regression coefficient to update MSI in the SFIM algorithm,finally,the bilinear interpolation method is used to up-sampling to obtain the rich spatial information of HSI and MSI,and the SFIM algorithm is used for calculation,and the fusion is performed separately band by band to obtain the final fused image.(2)A multi-source remote sensing image fusion algorithm based on block Bayesian sparse representation is given.The improved method first divides the multi-source remote sensing image to be processed into blocks,Then use Bayesian dictionary learning and Bayesian sparse coding to obtain the spectral dictionary and coding matrix of each sub-block of image data.Multiply the two to obtain the hyperspectral image of the fused sub-block,and finally the hyperspectral image of each sub-block is obtained.The images are combined to obtain hyperspectral image data with both high spectral resolution and high spatial resolution.Among them,the processing of each sub-block can be processed in parallel to further improve the efficiency of the algorithm.(3)An improved projection onto convex sets(POCS)fusion algorithm is given.The algorithm first optimizes the interpolation method,replaces the traditional nearest neighbor interpolation with bicubic interpolation.Then the relaxation operator is introduced into the projection formula,and the image is adaptively correction according to the sharpness of each position in the image;finally,an iterative processing method is used with the mean square error of the reconstructed image of the previous two iterations as the stopping rule so as to avoid subjectivity set of the iteration number.This paper evaluates the fusion performance from two aspects: subjective evaluation and objective evaluation.The experiments are tested on different data sets,and the results show that the three algorithms proposed in this paper have a good fusion effect.