Distributed compression for hyperspectral images

An efficient lossy compression algorithm was presented based on distributed source coding. The proposed algorithm employed multilevel coset codes to perform distributed source coding and a block-based scalar quantizer to perform lossy compression. Multi-bands prediction was used to construct the side information of each block, and the scalar quantization was performed on each block and its side information simultaneously. According to the principles of distributed source coding, the bit-rate of each block after scalar quantization was given. To reduce the distortion introduced by scalar quantization, skip strategy was employed for those blocks that containing high distortion in the sense of mean squared errors introduced by scalar quantization, and the block was directly replaced by its side information. Experimental results show that the performance of the proposed algorithm is competitive with that of transform-based algorithms. Moreover, the proposed algorithm has low complexity which is suitable for onboard compression of hyperspectral images.