Research On Algorithm Of Boosting The Qulity Of Pansharpened Image

Pan-sharpening means integrate the information of low spatial resolution multi-spectral image with the high spatial resolution panchromatic image to obtain high spatial resolution multi-spectral image.The fused image has good spectral properties and spatial resolution which is the data resource of other remote sensing image application.Traditional pan-sharpening methods are mostly based on a certain prior which lead some problems such as loss of detail or spectral distortion in the fused image.For example,methods based on component substitution are suffer from spectral distortion and color deformation;methods based on multi-resolution analysis may loss some of detail and cause some false contour around the edge.In this paper,we proposed a quality boosting framework for pan-sharpened image.Based on the result of traditional pan-sharpening method,the residual image is obtained by learning the difference between the fused image and the reference image and utilized to compensate and enrich the information of the fused image to enhance the quality.The specific research contents are as follows:1.We proposed a framework for boosting the quality of pan-sharpened image.The framework using the learning-based algorithm to learn the difference between the pan-sharpened image and the reference image to generate residual image and using the residual image to repair the loss of detail and spectral distortion of the pan-sharpened image which can effectively improving the spectral quality and maintaining the spatial information of the pan-sharpened image.2.Aiming at the problems of traditional pan-sharpening methods and integrate with framework for boosting the quality of pan-sharpened image,a pan-sharpened image quality boosting algorithm based on sparse representation is proposed.The algorithm based on redundant dictionary learning and neighbor embedding is used to learn the difference between the fused image and the reference image,mapping the fused image to residual image and fixing the spectral distortion and loss of detail in the fusion result by adding the fused image with the residual image to effectively maintain the spectral and spatial information.The experiment result show that the proposed method can effectively improve the spectral quality and preserve the spatial information of the fused image3.Aiming at the problems of traditional pan-sharpening methods and integrate with framework for boosting the quality of pan-sharpened image,a pan-sharpened image quality boosting algorithm based on deep residual denoising network is proposed.By considering the difference between the pan-sharpened image and the ideal high-resolution multi-spectral image as generalized noise,taking the advantage of the deep residual denoising network in generalized denoising,learning the pattern of detail loss or spectral distortion of traditional pan-sharpening methods,generating the residual images,using residual images to complement and repair information loss and defects in the pan-sharpened image.The experiment result show that the proposed method can effectively improve the spectral quality and preserve the spatial information of the fused image.The experimental results demonstrate that the quality of the fused images are improved by the proposed method,and that the best boosting result is given by the proposed boosting method performed on the support vector decomposition-based pan-sharpening method,and its performances are better than the latest methods.