Spectral Classified Vector Quantization (SCVQ) for Multispectral Images

Multi- and hyper-spectral data pose severe problems in terms of storage capacity and transmission bandwidth. Although recommendable, compression techniques require efficient approaches to guarantee an adequate fidelity level. In particular, depending on the final destination of the data, it could be necessary to maximize several parameters, as for instance the visual quality of the rendered data, the correctness of their interpretation, or the performance of their classification. Based on the idea of Spectral Vector Quantization, the approach proposed in this paper aims at combining a compression and a classification methodology into a single scheme, in which visual distortion and classification accuracy can be balanced a- priori according to the requirements of the target application. Experimental results demonstrate that the proposed approach can be employed successfully in a wide range of application domains.