An improved M-NVQ algorithm for the compression of hyperspectral data

Mean-normalised vector quantization (M-MTQ) has been demonstrated to be the preferred vector quantization technique for the lossless compression of hyperspectral data. The authors optimise the M-NVQ parameters for application to lossy compression. While slight improvement is shown to be gained by the implementation of spatial and spectral discrete cosine transform (DCT) techniques for coding of the M-NVQ residuals, much greater compression can be obtained by optimising the M-NVQ and DCT techniques simultaneously, rather than sequentially. Results for a spatial M-NVQ/spectral DCT coder are between 1.5 and 2.5 times better than the compression ratios obtained by the M-NVQ technique alone. The data used in this investigation was acquired by the Airborne Visible/Infrared Imaging Spectrometer (AVIRIS), which simultaneously acquires 224 channels of data. Channels are originally recorded with 12 bit resolution but, after radiometric correction, data is stored as 16-bit words.