Distributive segmented PCA: A novel approach to hyperspectral image compression

We propose a segmented PCA based distributed framework exploiting the high correlation and dimensionality reduction properties of PCA in conjugation with the distributed source coding paradigm for onboard hyperspectral image compression. The hybrid scheme deploying Lossy Wyner Ziv coding in conjunction with statistically segmented PCA strategy exploits the two fold intra group and inter group correlation for establishment of multiple independent processing elements evenly distributing the cumulative encoder load over the entire system. The net performance of this flexible infrastructure scheme is further increased by utilizing the superior rate distortion performance of JPEG2000 for high fidelity reconstruction of side information crucial for corresponding Wyner Ziv decoding. The inherent parallelism and low encoder complexity of the proposed scheme not only makes it a viable candidate for onboard implementation but also conforms it to subsequent multi-node transmission. Experimental analysis demonstrates that the proposed scheme offers a novel perspective and competitive performance as compared to other techniques in the same domain.