Fast classification of V-Q compressed hyperspectral data

An automatic hybrid supervised and unsupervised classification procedure is presented in this paper for fast thematic generation based on V-Q compressed data. The cluster-space formed by the codebook indexes is used for training data representation and image classification (CSC). The class separability can then be visualised and assessed quantitatively. The computational load is significantly reduced at the receiver end. The experiments using an AVIRIS data set show that classification accuracy obtained based on the compressed data is comparable to the original data with possible better performance when only a small number of training samples are available