Lossless Compression of Hyperspectral Images Based on Searching Optimal Multibands for Prediction

Abstract

This letter presents a lossless compression algorithm for hyperspectral images, which is based on the strength of correlations between bands. First, a searching model is constructed using the tree structure. Second, multibands which have strong correlations to each chosen band are found out and are then used to predict the chosen band in a couple-group manner. Lastly, residual images are encoded using entropy coders. Experimental results show that our compression algorithm provides a competitive compression performance compared with most existing compression algorithms.