Research On Hyperspectral Image Denoising Algorithm Based On Static Graph Model

Hyperspectral image(HSI)is composed of tens or even hundreds of gray level images corresponding to spectral channels,which has spectral resolution of nanometer order of magnitude and abundant texture information.Based on these HSI characteristics and the characteristics of the integration of gray level image and the spectrum,it can provide favorable conditions for target feature extraction.Thus,it is widely used in food quality detection,resource exploration,natural disaster monitoring,medical disease diagnosis,precision agriculture and other fields.However,HSI is often disturbed by various types of noise in the process of acquisition and transmission.These noises bring great challenges to the subsequent analysis,especially in feature extraction and target recognition,which will greatly reduce the reliability of the processing.Therefore,how to effectively remove the noise of HSI has become a crucial problem.Currently,scholars have proposed various advanced HSI denoising algorithms and achieved excellent performance,but most of these methods cannot efficiently characterize the inherent relation between spatial pixels as well as spectral adjacent bands.Besides,the existing methods always focus on a small-scale HSI,and most of the algorithms have poor denoising performance on large-scale HSI.In recent years,the emergence of graph signal processing(GSP)has provided new ideas to solve these problems.In the GSP framework,the irregular network structure is modeled as a graph model.And the data on the network is defined as graph signal.The edge of the graph structure can describe the relationship between data.This graph-based representation is adaptive to structure,which can naturally capture data characteristics.Thus,GSP has great application potential in HSI denoising domain.Based on GSP theory,this thesis proposes two HSI denoising algorithms based on GSP theory.The main research contents are as follows:(1)For the problem that existing algorithms can not adequately characterize the underlying structure of HSI,we propose a novel HSI denoising algorithm based on GSP theory and low-rank matrix recovery model.With GSP,the piecewise smoothness property of the HSI can be efficiently characterized,leading to a new regularization for HSI denoising,termed the adaptive weight graph total variation(AWGTV)regularization.Then,the denoising problem is formulated into a constrained optimization problem that incorporates the AWGTV and the low-rank matrix recovery model,which can simultaneously characterize the piecewise smoothness and low-rank property of HSI.Finally,an augmented Lagrange multiplier method is adopted to solve the problem.Experiments show that the algorithm has excellent denoising performance.(2)In view of the lack of research on denoising algorithms for large-scale HSI,we propose a novel mixed-noise removal method for HSI with large scale,by leveraging the superpixel segmentation technology and distributed algorithm based on GSP.First,the underlying structure of the HSI is modeled by a two-layer architecture graph.The upper layer,called skeleton graph,is a rough graph constructed by using the modified k-nearest-neighborhood algorithm,whose nodes correspond to a series of superpixels formed by HSI superpixel segmentation.The skeleton graph can efficiently characterize the inter-correlations between superpixels,while preserving the boundary information and reducing the computational complexity.The lower layer,called detailed graph consisting of a series of local graphs which are constructed to characterize the similarities between pixels within the superpixel region.Second,based on the two-layer graph architecture,the HSI restoration problem is formulated as a series of optimization problems each of which resides on a subgraph.In each optimization problem,a graph Laplacian regularization is incorporated into a low-rank matrix recovery model,which can simultaneously characterize the piecewise smoothness and low-rank property of the subgraph.Finally,a novel distributed algorithm is tailored for the restoration problem,by using the information interaction between the nodes of skeleton graph and subgraphs.Experiments show that the proposed denoising algorithm has excellent denoising performance in both large-scale and small-scale HSI.