Abstract: With the development of spectrometer, the research of hyperspectral remote sensing image has come into a new stage-processing and utilizing the acquired image efficiently. Now many processing methods are centered on arithmetic analysis processing of hyperspectral, such as atmosphere correction, dimensional reduction, information extraction, classification, compression, and so on. Whereas dimensional reduction of hyperspectral image is an important step to continue subsequent procedure. To select a dimensional reduction method correctly will affects the arithmetic and software which deal with plenty of data and abundant information of hyperspectral image. This paper generalizes and analyzed various methods from the point of view of band selection, subspace decomposition, feature extraction and fusion both at home and abroad.