Abstract
Preprocessing is a major area of interest within the field of hyperspectral endmember extraction since it could provide relatively better quality data prior to endmember extraction. In this paper, a novel spatial energy and spectral purity based preprocessing algorithm is presented, named SESPPP. For a given pixel, we first define its spatial energy and spectral purity index (SESPI) according to its spatial correlation, the extreme of projection, and local spectral variability. A high-quality pixel will have high SESPI values, whereas the heterogeneous pixels are not. Then, based on the well-determined SESPI values for whole pixels, we develops a new filter strategy to remove heterogeneous pixels with low SESPI values and retain the homogeneous pixels with highest SESPI values in the sliding window. Experimental results on real hyper-spectral Cuprite dataset indicates that the SESPPP is superior to current state-of-the-art preprocessing techniques.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call