The Research On Denoising Model And Algorithm Of Hyperspectral Image Based On Total Variation And Low-Rank Representation

When hyperspectral images are imaged,due to the influence of resolution,the complexity of ground objects and imaging weather,a single pixel is usually a mixture of multiple materials,also known as mixed pixel.Hyperspectral unmixing is the process of decomposing mixed pixels into some pure substances called endmembers,and their corresponding ratios called abundances,it is an important task in analyzing and processing hyperspectral images.With the availability of large-scale libraries,a plethora of sparse unmixing algorithms exploit spectral and spatial information in HSIs to enhance abundance estimation results.However,many algorithms treat the abundance matrices corresponding to active and nonactive endmembers in the scene equivalently.In this thesis,a universal framework named mixing support detection（MSD）is proposed for the spectral unmixing problem.Its specific method is as follows: First of all,active and nonactive endmembers are detected in each iteration,and then the corresponding abundance matrices are treated differently.The computational cost is reduced as the size of the abundance matrix and spectral library matrix is gradually reduced in each iteration.Furthermore,the interference of the abundance matrix of nonactive endmembers is attenuated by avoiding a large number of nonactive endmembers participating in the unmixing process.Moreover,the sparsity of abundance matrix is also enhanced by assuming that the abundance matrix f nonactive endmembers is zero.To illustrate the effectiveness and applicability of the mixing support detection framework,this thesis embeds this framework into five state-of-the-art sparse unmixing algorithms,and obtains five new unmixing models based on the mixing support detection framework.The Alternating Direction Method of Multiplier solves models and obtains new unmixing algorithms.Finally,the experimental results of simulated data and real-data experiments demonstrate the effectiveness of the algorithm based on the mixing support detection framework,and also show the effectiveness and universality of the mixing support detection framework.