A multi color space approach for texture classification: experiments with Outex, Vistex and Barktex image databases

The color of pixels can be represented in different color spaces which respect different properties. Many authors have compared the classification performances reached by these color spaces in order to determine the one which would be the well suited to color texture analysis. However, the synthesis of these works shows that the choice of the color space depends on the considered texture images. Moreover, the prior determination of a color space which is well suited to the considered class discrimination is not easy. That is why we propose to consider a multi color space approach designed for color texture classification. It consists in selecting, among a set of color texture features extracted from images coded in different color spaces, those which are the most discriminating for the considered color textures. In this paper, we experimentally study the contribution of this multi color space with three well-known benchmark databases, namely Outex, Vistex and Barktex. Comparison and discussion are then carried out.