This article presents an adaptive min-distance algorithm to classify multi-spectral remote-sensing images. This method approximates the distribution of the classes by dividing the sample sets

and adjusts the parameters of the min-distance classifier adaptively. Experiments with TM remote-sensing images demonstrate that this approach achieves an accuracy of 92.9% in the supervised classification of 16 classes. The experimental results verify the applicability of this approach in classifying of multicategory remote-sensing images.