Multispectral imagery compression by hybrid DWT and partitioning DPCM

Compression of multispectral imagery is based on reducing redundancies in both the spatial domain and the spectral domain. In this paper, a new hybrid compression algorithm using partitioning DPCM and SPIHT is proposed on the base of analyzing the spatial and spectral correlation features of multispectral imagery. A first-order predictor is designed for de-correlating the spectral redundancy and creating error images for later use. Because the image similarities among adjacent spectrum bands are different with the change of spectrum, the whole multispectral image sequence is partitioned into several subsets, and then DPCM predictors are designed separately for each image subset. After de-correlating spectral redundancy, a efficient wavelet coding method, SPIHT, is used to compress error images created by partitioning DPCM algorithm. The experimental results from simulated multispectral images and practical 64-band multispectral images have shown that the algorithm is fast, efficient and practical.