An image-based endmember bundle extraction algorithm using reconstruction error for hyperspectral imagery

Although many endmember extraction algorithms have been proposed for hyperspectral images in recent years, there are still some problems in endmember extraction which would lead to inaccurate endmember extraction. One important problem is the variation in endmember spectral signatures due to spatial and temporal variability in the condition of scene components and differential illumination conditions. One category to handle endmember variability is considering endmembers as the bundles. In other words, each endmember of a material is represented by a set or “bundle” of spectra. In this article, to account for the variation in endmember spectral signatures, an image-based endmember bundle extraction algorithm using reconstruction error for hyperspectral remote sensing imagery is proposed. In order to demonstrate the performance of the proposed method, the current state-of-the-art endmember bundle extraction methods are used for comparison. Experiments with both synthetic and real hyperspectral data sets indicate that the proposed method shows a significant improvement over the current state-of-the-art endmember bundle extraction methods and perform best in subsequent unmixing.