Lossless Compression Of Hyperspectral Image Based On Prediction And JPEG2000

Hyperspectral images, which are obtained by imaging spectrometer, are datacubes. They contain both2-D spatial information and1-D spectral information.Because of large amount, they must be effectively compressed for the convenienceof transmission and storage. Moreover, hyperspectral images are the key datasource for ground object spectra analysis. Lossy compression is bound to loseuseful information, so they should be compressed lossless.In this paper, lossless compression methods hyperspectral images werestudied, using prediction algorithm and JPEG2000standard. First, the principles ofimage compression and basic algorithms were introduced. In the beginning, thebasic information theory of image compression and commonly used entropycoding methods were given; Then, the principles of three major categories ofcompression algorithms (prediction algorithm, transform algorithm and vectorquantization algorithm) were introduced, and highlights the JPEG2000standard; Inthe last, the quality evaluation methods of compression algorithms were given.Secondly, the lossless compression algorithms of hyperspectral image wereanalyzed and implemented. In the beginning, the spatial correlation and spectralcorrelation of hyperspectral images were analyzed, which provides guidance forthe compression; Then, the feasibility of hyperspectral image lossless compressionwas given from the perspective of information theory. In the last, the threecompression methods for hyperspectral image lossless compression were implemented, and the compression results shows that the3D compression methodbased on prediction and JPEG2000could achieve higher compression ratio (2.06).Third, write a MATLAB GUI-based software for hyperspectral image analysis andcompression. The analysis module can realize the analysis of high spectral andcorrelation (including spatial correlation and spectral correlation); the compressionmodule can realize first-order linear lossless prediction coding, JPEG2000losslesscoding and lossless coding based on the combination of both. Experimental resultsshow that the software can be simple, functional, effective and so on. Fourth, areal-time image compression system for a convex grating imaging spectrometerwas designed. In the beginning, the convex grating imaging spectrometer wasintroduced, including its optical structure and imaging principle; Then, eachfunctional module of the system was designed based on the combination of thefirst-order linear prediction and JPEG2000; In the last, the system’s performancewas analyzed and the result shows that it can meet the requirement of real-timeimage data compression.