Influence of color spaces over texture characterization

Abstract

Images are generally represented in the RGB color space. This is the model commonly used for most cameras and for displaying on computer screens. Nevertheless, the representation of color images using this color space has some important drawbacks for image analysis. For example, it is a non-uniform space, that is, measured color differences are not proportional to the human perception of such differences. On the other hand, HSI color space is closer to the human color perception and CIE Lab color space has been defined to be approximately uniform. In this work, the influence of the color space for color texture characterization is studied by comparing Lab, HSI, and RGB color spaces. Their effectiveness is analyzed regarding their influence over two different texture characterization methods: DFT features and co-occurrence matrices. The results have shown that involving color information into texture analysis improves the characterization significantly. Moreover, Lab and HSI color spaces outperform RGB