Three-Dimensional Wavelet-Based Compression of Hyperspectral Images

This chapter proposed a three dimensional set partitioned embedded block coder for hyperspectral image compression. The three dimensional wavelet transform automatically exploits inter-band dependence. Two versions of the algorithm were implemented. The integer filter implementation enables lossy-to-lossless compression, and the floating point filter implementation provides better performance for lossy representation. Wavelet packet structure and bit shifting were applied on the integer filter implementation to make the transform approximately unitary.

Rate distortion results of both lossless and lossy compression of hyperspectral imagery have been presented, and all results were compared with other state-of-the-art three dimensional compression algorithms such as 3D-SPIHT and JPEG2000 multi-component. 3D-SPECK is competitive to 3D-SPIHT and better than JPEG2000 in compression efficiency. The plots of original, reconstructed and error spectral profiles shown that the proposed algorithm preserved spectral profiles well.

The proposed 3D-SPECK is completely embedded and can be used for progressive transmission. These features make the proposed coder a good candidate to compress (encode) hyperspectral images before transmission and to decompress (decode) them at another end for image storage.