LOSSY COMPRESSION OF HYPERSPECTRAL DATA USING VECTOR QUANTIZATION

Efficient compression techniques are required for the coding of hypers pectral data. Lossless compression is required in the transmission and storage of data within the distribution system. Lossy techniques have a role in the initial analysis of hyperspectral data where large quan tities of data are evaluated to select smaller al-eas for more detaile d evaluation. Central to lossy compression is the development of a sui table distortion measure, and this work discusses the applicability of extant measures in video coding to the compression of hyperspectral i magery. Criteria for a remote sensing distortion measure are developed and suitable distortion. measures are discussed One measure [the perc entage maximum absolute distortion (PMAD) measure] is considered to be a suitable candidate for application to remotely sensed images. The e ffect of lossy compression is then investigated on the maximum likelih ood classification of hyperspectral images, both directly on the origi nal reconstructed data and on. features extracted by the decision boun dary feature extraction (DBFE) technique. The effect of the PMAD measu re is determined on the classification of an image reconstructed with varying degrees of distortion. Despite some anomalies caused by challe nging discrimination tasks, the classification accuracy of both the to tal image and its constituent classes remains predictable as the level of distortion increases. Although total classification accuracy is re duced from 96.8% for the original image to 82.8% for the image compres sed with 4% PMAD, the loss in accuracy is not significant (less that 8 %) for most classes other than those that present a challenging classi fication problem. Yet the compressed image is 1/17 the size of the ori ginal. (C) Elsevier Science Inc., 1997.