On Remote Sensing Hyper-spectral Image Compression Based On SPIHT Algorithm

Hyper-spectral imaging possesses the quality of high spatial resolution, high spectral resolution and more spectral channels. Many problems can be solved by using the high spectral resolution of hyper-spectral image while multi-spectral image can not. However, the high spectral resolution of hyper-spectral image is in the expense of greater data amount and bigger dimension, bringing difficulties in the transmission and storage of the image. Thus investigating compression algorithm that have higher performance and easy to implement is important to the application of hyper-spectral image.According to the characteristics of remote sensing hyper-spectral image compression, this thesis proposes lossless and near-lossless hyper-spectral image compression projects. Firstly, due to the strong correlation between the spectrums of hyper-spectral image, DPCM is used to process the sequence spectrum of the image. Then, integer (5.3) and (9.7) wavelet transform is used respectively to the processed image. For the easy operation of shifting and addition, and the advantage of eliminating redundancies of hyper-spectral remote sensing data more efficiently than common wavelet transform, integer wavelet transform is especially fit for data processing methods required to be real time, high speed coding and loss less compression. Finally, the wavelet transformed coefficients are processed by SPIHT algorithm. The coding efficiency is enhanced the fact that Spatial Orientation Tree and two sets are used in SPIHT to represent the structure of wavelet coefficients.Experiment results show that lossless compression rate can reach 2.34 when using SPIHT algorithm decomposition, improving by 14.1% compared with arithmetic coding algorithm. When using 4-level decomposition in integer (9.7) wavelet transform, and lossy compression transmit bit rate is 0.5 bpp, the PSNR of hyper-spectral sequence image can reach around 44.437dB and the image is well reconstructed. The results indicate that the compression project proposed in this thesis has a better effect on the compression of hyper-spectral image.