The placement of hyperspectral sensors on future spacecraft and unmanned air vehicles is constrained by a data bottleneck caused by limited downlink bandwidth. This proposal addresses this problem by developing a computationally efficient, lossless compression algorithm that is optimized for hyperspectral imagery. In the Phase I, we used a hybrid approach involving the Eigendecomposition to extract spectrtal redundancies, and wavelets Contextual arithmetic encoding was used to efficiently encode the transformed hyperspectral data. This approach, which we term as the EigenWavelet Technique, is further explored in Phase II, with a view to improve system performance characterized by the compression ration and computational speed. The proposed work incorporates automatic parameter selection, and a graceful response to bit errors. We propose a hybrid multiprocessing approach to attain the levels of performance required for developing a working prototype, involving both CPU-based fine grained parallelism and multi-CPU coarse grained parallelism. We approach lossy compression as a truncated lossless compression process. In addition, the algorithm permits progressive transmission. The encoding and decoding complexities are nearly symmetric.
Keywords: Contextual Arithmetric Encoding Error Control Hyperspectral Image Compression Lossless Image Compres