Optimal color sets to represent the colors of natural scenes by k-medoids clustering.

Abstract

The Munsell and Natural Color Systems, as well as the World Color Survey, are standard sets of colors used in many practical and scientific applications. However, the colors of natural scenes exhibit a bias in color and do not have a uniform distribution, making it difficult for these sets to represent natural colors accurately. We derived sets of colors with a small number of samples that are better at representing natural colors than any of these standard sets. Hyperspectral images of natural scenes and a k-medoids clustering algorithm were used to derive representative colors. For the same number of samples, the set of colors obtained by k-medoids is better at representing natural colors than the standard sets. These optimized sets are important for applications that require precise representation of natural colors.