Normality assumptions are used in many statistical pattern recognition classifiers. In order to visualize consequences of these assumptions on color class distributions a 3D display is necessary.
Normal distributions are plotted in 3D space for a theoretical percentage of the distribution points. The following figure presents the display of normal distributions for 70% of the points in the previous data base.
Note that some ellipsoids intersect, and that some distribution are truncated at the limit of the 3D RGB space. The truncation is caused by camera sensitivity limits.