When analyzing the RGB distribution of colors in natural images we notice that they are organized into spatial structures. The use of a linear model for the distribution of colors agrees with the common assumption that the scenes are composed of lambertian objects, for which the emitted surface color depends on the intensity of the illuminant, the reflectance coefficient and the relative orientation of the surface with respect to the light source. However, the color of an object may vary on different parts on the object surface, and the reflection properties of the object may also change, due to illumination changes and to pigment variations in the material body. This fact suggests that a higher dimensional model might be better than the linear model in some cases. Following this line of thought we analyze the distribution of colors of several natural images and discuss the validity of both 1D and 2D models. We propose a dimension reduction algorithm that reveals the underlying 1D or 2D structure of the color clusters and we conclude that, in general, the 2D model fits better the observed distributions.