Researches Of Hyperspectral Image Compression Algorithm Based On Predictive Techniques

With the expanding scope of the application fields of remote sensing in recent years, multispectral images cannot meet the increasing demand for scientific researches. At the same time, hyperspectral images with much higher spectral resolution have been made available with the invention of imaging spectrometer. Hyperspectral images provide far more topographic detailed information compared with multispectral data. However, such a high spectral resolution is on the expense of even huge data volume, which brings new challenges to the current techniques for storage and transmission. Therefore, it is necessary to compress these large data sets. Consequently, it is urgently needed and valuable to explore specific compression algorithm for hyperspectral images. In this paper, a compression algorithm for hyperspectral images is studied systematically based on prediction.First, prediction is deduced to serve as the core step for hyperspectral images compression. Prediction algorithm is a most simple one of all methods on images compression.And it explore directly the spectral correlations of hyperspectral images.Its method is very simple and easy to be implemented.Prediction takes full advantage of the spectral correlations to use transmitted spectral to predict current spectral,then predictive error that original spectral is substracted by predictive spectral is coded.The predictive error is wiped off the spectral correlations,so compression is easier than before.Second, a novel algorithm named model predicition(MP) is proposed to wipe off spectral correlations of hyperspectral images.MP algorithm finds the linear model of hyperspectral images,in which predictive coefficients are set up that is based on SNR.Because predictive coefficients include current spectral band, average entropy of the error data is decreased and SNR is increased after MP.Finally, the error images of MP are coded by SPIHT(Set Partitioning in Hierarchical Trees) to get the final codestream.In order to testify the effectiveness of the proposed algorithm, AVIRIS images are used for computer simulation and the results indicate the proposed...