Hyper-spectral Remote Sensing Image Compression Research Based On EBCOT Algorithm

Hyper-spectral remote sensing image is a three-dimensional image, that is, add one-dimensional spectral information to the two-dimensional images, which make the Hyper-spectral images have a larger number of spectral channels, as well as high spatial resolution and high spectral resolution. But a large number of data images bring difficulty to its transmission and storage .Thus, compression algorithm of high performance and availability is of great significance to the application of hyper-spectral images .In this paper, we mainly research the compression algorithm hyper-spectral remote sensing images.The paper introduced the following basic theory: Discussed the basic theory of wavelet transform, the image after wavelet transform most of the energy concentrated in a few wavelet decomposition coefficients and easy to get a high compression ratio. We analysis and elaborate the basic principles of EBCOT algorithm. Because of two coding process, it can not only achieve an effective image compression, but also generated streams with resolution and SNR scalability, good random accessing and handling characteristics and so on.This paper adopt two programs on hyper-spectral image compression: IN program one, we used one-dimensional DPCM to remove spectroscopy correlation and two-dimensional integer (5,3) wavelet transform to remove spatial correlation and then combinated EBCOT algorithm to make compression coding. In program two, after using three-dimensional (5,3) integer wavelet transform to remove two kind of correlation, combinated the EBCOT algorithm to compress.Simulation results show that in the aspects of removing the Spectral correlation, DPCM is slightly better than the integer (5,3) wavelet transform, But the two are all good, the Removal proportion of relevance can achieve 90.62% and 93.77%. Compared with the algorithm of DPCM+(5,3)+SPIHT, the nondestructive compression ratio of two programs is all upgraded . And when transmission bit rate is 1.0bpp, lossless compression average peak SNR of programe one can be achieved 45.4328dB,althouth compored with the programe one ,the programe two's had a slight decrease,it can also achive 44.0914dB, and they can reconstructed image better. This shows that these two programs are applied to hyper-spectral image compression.