Unlike most color classification methods, which consist in partitioning the image according to the pixels color attributes exclusively, spatio-colorimetric techniques bring some spatial information directly among the data to classify. However, they usually involve some heavy data structures and a large amount of trichromatic data.To answer this issue, this article proposes a color spatio-classification method performing two successive stages. First of all, the number of colors is lowered through an analysis of the connectedness degrees on the three marginal components independently. Since the number of colors is significantly reduced, it becomes reasonable, in a complexity point of view, to analyze the vectorial connectedness degrees of the trichromatic intervals. Several experimental results will be shown on different images and the method parameters will be discussed.