Lossless Compression Of Hyperspectral Images Based On Prediction

Spatial resolution and spectral resolution of hyperspectral images have been improved as the development of imaging spectrometer technology, adding a sharp increase in the amount of image data. Hyperspectral image are widely used in many fields who attaches great importance to image quality, such as, agriculture, forestry, geological survey, environmental monitoring and military reconnaissance and so on. It is necessary to propose a more efficient method for lossless compression of hyperspectral images. Because compression algorithm based on the prediction can obtain good compression performance and is so simple that can Be implemented by hardware easily. Therefore, research on lossless compression of hyperspectral image based on prediction is very significant.This paper introduces the remote sensing technology, providing a briefframework of remote sensing systems; introduces hyperspectral image and its compression technology, summarizes its research status and compression technologies briefly; then, analyzes the characteristics of hyperspectral images, including its spatial correlation analysis, spectral correlation analysis and information analysis; then introduces the basics of lossless data compression, including lossless compression predictive principles and some existed compression standards; finally, does some in-depth research on the existed algorithms SLSQ and C-DPCM-APL proposing two innovations.One innovation proposes a lossless compression of hyperspectral image based on the diagonal edges and the context ordering which taking hyperspectral image characteristics into account and combining with the median prediction algorithm and SLSQ algorithm. The proposed algorithm contains intra-band prediction and inter-band prediction. In intra-band mode, considering the existence of the diagonal edges in hyperspectral images, a diagonal edge prediction is proposed by introducing the diagonal edge detection into the standard median prediction. And in inter-band mode, a context-ordering prediction is proposed, due to the fact that in a context window the pixels which have a closer distance and more similar pixel values will have a stronger correlation because of the existence of edges. Experimentson AVIRIS 1997 and CCSDS 2006 images showthat the proposed scheme obtains better compression performanceand yields anaverage higher compression ratio of 0.11 than SLSQ.Another innovation proposes a lossless compression of hyperspectral image using C-DPCM-APL with reference bands selection. First, it does in-depth analysis on DPCM and C-DPCM-APL algorithms. C-DPCM-APL subjects to certain restrictions when used in spaceborne compression because it is a lossless compression algorithm which can obtain highest compression ratio but whose processing time is the longest. This innovation takes predictive influence of the correlation between the reference bands and the current band into consideration to select the appropriate reference bands for the current band prediction. Compared with C-DPCM-APL, this algorithm reduces the processing time of the compressionsignificantly, but did not bring significant adverse effect on compression, becoming a lossless compression most practical.