Low complexity DCT-based distributed source coding for hyperspectral image

Most distributed source coding (DSC) compression methods for hyperspectral image in transform domain are based on discrete wavelet transform (DWT). However, we point out that DWT domain is not efficient for the on-board acquisition system, which requires simple encoder. We find hyperspectral image is also highly correlated in discrete cosine transform (DCT) domain, and the computational cost of DCT domain is much smaller than that of DWT domain. Therefore, this paper proposes a low complexity DCT-based DSC scheme to compress hyperspectral image. We use a set-partitioning approach on reorganized DCT coefficients to extract bitplanes; then we apply low density parity check-based (LDPC-based) Slepian-Wolf coder to implement our DSC strategy. By this means, the coding paradigm shifts the complexity from the encoder side to the decoder side. Preliminary experimental results for AVIRIS (Airborne Visible/Infrared Imaging Spectrometer) data show that the proposed scheme can reach up to 0.4 dB better than DSC-based coder in DWT domain, despite of the using DCT-based coder inferior to DWT-based one. The performance also improves up to 5dB as compared to that independently using 2-Dimensional (2D) DCT-based coder.