Eye color classification for makeup improvement

The development of computer-aided solutions able to suggest the right facial makeup is a recent trend in image analysis applications, from which both amateurs and professionals could benefit significantly. The global harmony of a person is highly valuable when choosing makeup colors to make a person looking lovely. The global harmony is evaluated taking into account the color of the hair, skin and eyes, and among these features, the eyes seem to be one of the most salient features that capture an individual attention. This paper proposes a simple yet effective eye color classification scheme, compliant to the categories associated to the cosmetic software, which are often different than the classification systems used in medicine or biometrics. The color descriptors are histograms of the iris color distribution in the HSV color space, classified by multi-class Support Vector Machines, and the high accuracies achieved recommend it for digital cosmetic assistant solutions.