Region based KLT for multispectral image compression

In this paper we present a new approach of spectral decorrelation for multispectral image compression. It is based on the merging of two main tendencies such as the use of KLT as spectral decorrelator and object based image coding schemes. The use of the principal component in multispectral imagery is described and used to perform a multispectral segmentation. This se­gmentation is taken as the basis for a specific spectral decorrelation for each segmented class. The resulting eigenimages present lower variance than classical KLT approaches, leading to better compression ratios.