https://upf.hal.science/hal-03132308Loesdau, MartinMartinLoesdauGePaSUD - Géopôle du Pacifique Sud - UPF - Université de la Polynésie FrançaiseChabrier, SébastienSébastienChabrierGePaSUD - Géopôle du Pacifique Sud - UPF - Université de la Polynésie FrançaiseGabillon, AlbanAlbanGabillonGePaSUD - Géopôle du Pacifique Sud - UPF - Université de la Polynésie FrançaiseChromatic Indices in the Normalized rgb Color SpaceHAL CCSD2017color space theorynormalized rgb color spacechromatic indicescolor object classificationcolor feature vectors[INFO] Computer Science [cs]Delcambre-Maillard, Elodie2021-02-05 01:13:272021-02-19 03:08:202021-02-18 09:59:41enConference papershttps://upf.hal.science/hal-03132308/document10.1109/DICTA.2017.8227415application/pdf1In typical applications, chromatic indices are calculated as linear combinations of the normalized r-, g-and bchannels and used as features for a later classification based on chromatic appearance. But the variety of indices used in the literature is very limited. Furthermore is the choice of which index to use justified either empirically, based on false mathematical assumptions or not justified at all. The reason for the lack of mathematical justification is that so far no formal definition of chromatic indices existed. Such a definition is provided in this paper. An experimental classification with 180 different index combinations shows that the index choice has a significant impact on the classification. The results stand in sharp contrast to the very limited variety of indices used in the literature. The results imply the need for deterministic methods to estimate ideal indices for a given application, for which we hope the provided formalization might be useful.