The extraction of thematic information from TM images and the selection of optimal composition schemes have been subjects requiring deep research in the field of remote sensing. However, so far, there have been no complete sets of norms or standards to follow, especially, in the quantitative calculation and prediction of image colorimetry, as well as the related determination of optimal composition schemes. Using TM images and applying the separation exposure method, this paper discusses the standardization issue of extracting image thematic information, sets forward the method of quantitative calculation of colour difference and the theory of selecting the optimal composition scheme, and provides a corresponding mathematical model.