1 August 2017 Hyperspectral image classification based on filtering: a comparative study
Author Affiliations +
The classification of hyperspectral images benefits greatly from integration of spectral information and spatial context. There have been many means to incorporate spatial information into the classification, such as the Markov random field, extended morphological profiles, and segmentation-based methods. Recently, spatial filtering was introduced to improve the classification accuracy of hyperspectral images. Compared with other spectral-spatial algorithms, spatial filtering is simple and easy to implement. This advantage makes it suitable for practical applications. However, spatial filtering has not been given enough attention. A comprehensive comparative study of spatial filtering is conducted. Specifically, 10 kinds of filters are used to smooth the hyperspectral images and the classified maps, respectively. The experimental results show that most filtering-based classification methods perform well with high efficiency.
© 2017 Society of Photo-Optical Instrumentation Engineers (SPIE) 1931-3195/2017/$25.00 © 2017 SPIE
Xianghai Cao, Beibei Ji, Yamei Ji, Lin Wang, and Licheng Jiao "Hyperspectral image classification based on filtering: a comparative study," Journal of Applied Remote Sensing 11(3), 035007 (1 August 2017). https://doi.org/10.1117/1.JRS.11.035007
Received: 31 March 2017; Accepted: 13 July 2017; Published: 1 August 2017

Back to Top