Remote sensing hyperspectral images provide more information than multispectral images in the sense that the spectral resolution of the former is much higher than that of the latter. They can solve the problems that the multispectral images can not. Their invention is a leap in the techniques of remote sensing for applications. There are two key subjects-classification and compression—for the researches of hyperspectral images now

which are both independent and dependent. Compression can be viewed as a kind of claasification realized by allocating different code words to different sub-blocks;on the contrary classification also can be considered as a type of compression extracting interesting object information. The main difference between them lies in the different standpoints eveluating the last results. Compression emphasizes the mean error of the reconstructed image

and classification emphasizes the misclassified probability of the images. Because of their inner relation

there are many similar realization algorithms. This paper first summarizes the methods of hyperspectral image classification and compression. Then the similar characteristics and differences for both are compared. Following two schemes for hyperspectral image classification and compression are introduced

and the computer simulations are carried out. Finally

the conclusions are given and the further research techniques are suggested.