Adaptive coding of hyperspectral imagery

Two systems are presented for compression of hyperspectral imagery. These systems utilize adaptive classification, trellis-coded quantization, and optimal rate allocation. In the first system, DPCM is used for spectral decorrelation, while an adaptive wavelet-based coding scheme is used for spatial decorrelation. The second system uses DPCM in conjunction with an adaptive DCT-based coding scheme. In each system, entropy-constrained trellis-coded quantization (ECTCQ) is used to quantize the transform coefficients. Entropy-constrained codebooks are designed for generalized Gaussian distributions by using a modified version of the generalized Lloyd algorithm. The wavelet-based system compresses an AVIRIS hyperspectral test sequence at 0.118 bits/pixel/band, while retaining an average peak signal-to-noise ratio (PSNR) of 41.24 dB. The DCT-based system achieves the same bit rate with an average PSNR of 40.72 dB