Hyper-spectral Image Compression Research Based On SPIHT Algorithm

Hyper-spectral image with high spectral resolution, narrow spectral features, can provide more abundant information about ground object for us, but the enormous data has brought certain difficulty for storage and transmission. Therefore, studying efficient compression method is very necessary.In this paper, hyper-spectral image compression technology has been studied, we adopt KLT transform to eliminate spectral redundancy and wavelet transform to eliminate spatial redundancy and then adopt two SPHIT coding scheme, one is that we adopt two-dimensional SPHIT algorithm to encode each band, the other one is that we adopt 3D-SPHIT to encode the whole image. For 2D-SPHIT algorithm, we adopt a non-uniform rate allocation method based on KLT transform feature vectors, experimental results show that the peak signal to noise ratio obtained using this method is about 3db higher than the method of uniform distribution rate. For 3D-SPHIT algorithm, we analysis the performance of hyper-spectral image redundancy between KLT transform with wavelet transform and asymmetric 3D wavelet transform, the results show that, in general, the signal to noise using wavelet transform KLT transformation programs available ratio is higher than the direct three-dimensional wavelet transform scheme, at low bit rate, little difference between the two, but with the increase rate, using KLT transform wavelet transform scheme can get a better signal to noise ratio.