RLS classification of hyperspectral image after fusion and recursive filtering

Hyperspectral image classification is a challenging task due to its large dimension. Dimensionality reduction has an important role in hyperspectral image classification as it reduces redundancy by mapping the higher dimensional data into lower dimension without losing any valuable information in the spectral signature of each pixel. In this paper, dimensionality reduction is achieved by band partitioning and averaging, followed by classification using regularized least square(RLS) classifier. Adjacent bands in the hyperspectral image are partitioned into multiple subsets and then are combined together by averaging which follows transform domain recursive filtering for edge preservation. These dimensionality reduced, edge preserved images are classified using a RLS classifier. Our experimental results show that the proposed method has high accuracy, utilizes spectral and spatial information and is robust compared to other conventional classification techniques.