The color characterization of digital cameras often requires the use of standard charts containing a fixed number of color samples. The exact choice of such characterization charts–how many (and which) known samples to include–is known to affect characterization performance. This study describes methods to select optimum color samples from a set of 1269 Munsell surface colors. The effect of sample selection on characterization performance is evaluated and compared with performance using the standard GretagMacbeth ColorChecker and GretagMacbeth ColorChecker DC colors. The work confirms that the standard charts appear to have been well selected. However, we show that it is possible to select 24 samples from the Munsell set that outperform the GretagMacbeth ColorChecker and that this selection can be efficiently derived using an algorithm called MAXMINC. It is proposed that this algorithm may have general applicability; for example, to the optimal selection of samples constrained to be a subspace Munsell color solid.