Color classification in a multidimensional color space
Résumé
In this paper we proposed a color classification method which consists firstly to select through a multi-stages process the most discriminating color components for a given image among a set of conventional color spaces. To compute the most discriminating components we used the PCA technique on the lightness components, next on the mono-chromatic components, next on the opponent components and lastly on the multi-chromatic components. The second step of our process consists to select from the set of discriminating components a sub-set of color components which maximises the variance inter-classes. This second step is based on a multidimensional classification of color histograms.