Abstract: On the basis of the hybrid algorithm constructed by cluster analysis and principal component analysis a newer multi-spectral space for color representation is developed.When cluster analysis applied using self-organized mapping the spectral distributions in each cluster are substituted by the mean distribution of those in the corresponding cluster.Further,principal component analysis is implemented on the compressed spectra set.Using the eigen spectral distributions as base to spanning the wide-band multi-spectral space for color representation.Results of experiments show that this newer multi-spectral space can represent and reconstruct the spectral distribution of the original color in a high accuracy.