Research On Remote Sensing Image Fusion Technology Based On Compressed Sensing

Limited by sensors,remote sensing satellites can not obtain remote sensing images with high spatial resolution and high spectral resolution.Image fusion can synthesize Panchromatic image with high spatial resolution and Multi-Spectral image with low spatial resolution into an image,which has high spatial resolution and high spectral resolution.It becomes a very critical issue that how to use image fusion to obtain high quality fused images.In response to this problem,this thesis studies the remote sensing image fusion technology:Firstly,this thesis presents the purpose and significance of the remote sensing image fusion,introduces the research status at home and aborad,analyzes the problems existing in the remote sensing image fusion and predicts the problem future research will focus on.This thesis introduces the pixel-based remote sensing image fusion method and sparse representation theoery.The subjective evaluation and objective evaluation of the fusion image quality are briefly introduced.It is suggested that the quality of the fusion image should be scientifically and reasonably evaluated based on subjective evaluation and objective evaluation.Secondly,better quality of fusion can be obtained after image denoising.Research on imge denosing is vital to the remote image fusion.Remote sensing images processed by traditional image denoising method tend to be over-smooth and lose edge structure information.For the similar block selection in the image denoising algorithm based on SVD,an improved algorithm baesd on SVD is proposed.The proposed method adopts matrix similarity to measure similarity of image block and design corresponding weight coefficients.The experiment shows that the proposed method can acquire better filtering performance than other traditional algorithms.The proposed method is beneficial to the subsequent remote sensing image fusion.Finally,the direction limitation of multi-scale transformation casues low spatial resolution of the fusion image obtained by multi-scale transformation algorithms.The spatial dtail information has influence on the quality of fusion image through analyzing the detail injection model.A new method based on guided filters and joint sparse representations is shown: Multi-scale guided filtering is used to process panchromatic images to acquire more complete spatial detail information.The joint coefficient model filters out spatial detail information to remove redundant information and improve the accuracy of spatial detail information.Experiments show that the proposed algorithm can effectively improve the spatial information accuracy of fused images which have higher spectral fidelity and improve spatial resolution.