A Non-local Means Based Classification Method of Hyperspectral Image

Fourier transform imaging spectrometer can not only provide spatial information but also a wealth of spectral information of the target. The acquired data is normally called hyperspectral images and it normally consists of hundreds of band-images. The Fourier transform imaging spectrometer can be applied in environmental mapping, global change research, geological research, wetlands mapping, assessment of trafficability, plant and mineral identification and abundance estimation, crop analysis, and bathymetry et al. All the application mentioned is related to accuracy of the classification method. Classification of a hyperspectral image sequence amounts to identifying which pixels contain various spectrally distinct materials that have been specified by the user. Several techniques for classification of multi-hyperspectral pixels have been used from minimum distance and maximum likelihood classifiers to correlation matched filter-based approaches such as spectral signature matching and the spectral angle mapper. In this paper, a hyperspectral images classification algorithm is proposed. The Non-local means method is applied to improve the accuracy of classification. NLM method can smooth the image region which has the similar pixel values, and reduce the distance between different targets within one class. In other hands, The NLM method can protect the image edges so that the distance between two classes become larger. In this paper, the Bhattacharyya distance is used to evaluate the separability of the hyperspectral images obtained from different filter methods. The proposed NLM-based method is tested using hyperspectral imagery collected by the National Aeronautics and Space Administration Jet Propulsion Laboratory. Experimental results the efficiency of this new method on hyperspectral images associated with space object material identification.