The need for a better visualization of color space occupation and partitioning evolved from a study into the exact behaviour of color approximation algorithms. We found the usual way to evaluate the results of approximation methods by presenting test images (originals and approximated versions) rather unsatisfactory: it is difficult to substantiate the differences with the original and the printing process itself may have distroyed most of those differences and introduced others.Especially scanned color images, that contain a substantial amount of noise present a challenge to approximation algorithms; the expected frequencies of color triplets in any of the often used colorspaces (RGB, HSI, CMY,…) are very low, which greatly influences the way color approximation algorithms proceed in their rectangular or spheric decomposition of 3-D colorspaces.Each of the proposed color approximation schemes is particularly well suited for a certain class of 3-D color occupation histograms. This 3-D histogram structure can not be obtained from the test images itself. We agree with (ref. 1) that 3-D visualization tools should be developed for color gamut transformations. In order to characterize the spatial distribution of occupied color triplet values we evaluated several volume visualization methods.The cloudlike triplet density distribution of scanned color images are best visualized using a depth shaded volume imaging method on a resolution reduced 3-D histogram. The presence of clusters and the amount of noise is immediately evident from stereo pairs of such visualizations.The visualization method developed was also used to display the color space subdivision structure of approximation algorithms and logic combinations of occupancy and subdivision.