Hyperspectral collaborative sparse unmixing based on fusion of stacked edge-preserving filterings

Abstract

The spatial resolution of the hyperspectral image is usually low, the endmembers are uniformly mixed, and a single pixel usually contains multiple mixed endmembers. In order to extract a single endmember and its corresponding abundance coefficient in the hyperspectral images, a hyperspectral linear collaborative sparse unmixing method based on edge-preserving filtering is proposed. Firstly, the multiparameter EPF extraction and the PCA-based fusion of stacked EPFs are used to preprocess the hyperspectral image, the b,o mixed norm is used as the measure of the abundance vector to realize the collaborative sparse constraint, and the row hard threshold function is used to solve the lo optimization model. The collaborative sparse hyperspectral unmixing method can obtain the collaborative sparse abundance vector. Secondly, local reweighting constraints are added to the abundance matrix, which improves the sparsity of endmembers and abundance coefficients. Finally, the total variation regular term is added, and Alternating direction method of multipliers is used to solve the problem. In order to verify the performance of the algorithm, experiments are carried out on synthetic data and Urban data, and compared with algorithms such as CSUnLO, CSUnSAL, CSUnSAL-TV, the proposed algorithm has better unmixing accuracy than other algorithms. Experimental results show that the proposed hyperspectral unmixing model effectively improves the unmixing accuracy of hyperspectral images.