A Study On Hyperspectral Image Compression Based On 3-D Wavelet Transform And Fractal Coder

Hyperspectral remote sensing image is three-dimensional, which adds the information of spectrum into two dimensional images. The similarity between and within spectrums is quite strong, so we'd better adopt some methods that can reduce the spatial redundancy and spectral redundancy at the same time to compress Hyperspectral images. At present, the most popular methods are lossless compression methods and almost lossless compression methods, whose compression ratio is not very high. In fact, however, some researches don't need all information of hyperspectral images. Therefore, this paper puts forward a compression method which can get a high compression ratio and acceptable information loss.The author puts forward a new listless wavelet-based coder which is based on quadtree coder. The efficiency of the algorithm is better than others. So the classical eight kinds of affine transformations in 2-D fractal image coder was generalized to 19 kinds in 3-D and implement and octree fractal coder was carried out in 3-D.In this paper, 3-D fractal-wavelet image compression coder is applied into hyperspectral remote-sensing image, which is a combination of 3-D improved listless wavelet coder and 3-D octree fractal coder. Firstly, the data cube of hyperspectral image is transformed by 3-D wavelet transform. Then 3-D octree fractal coder is applied into the low-frequency band and 3-D new wavelet-based coder is used into the high-frequency band. The new compression method had been simulated by MATLAB 7.1. The experiment results show that we can gain high compression ratio and the information loss is acceptable.