Overview

J. Birgitta Martinkauppi, Matti Pietik inen

University of Oulu, Department of Electrical and Information Engineering, PO Box 4500, Oulu FIN-90014, Finland, jbm@cs.joensuu.fi, mkp@ee.oulu.fi

Color information has been commonly used to assist face detection. Its popularity can be attributed to several factors such as computational efficiency and discrimination power. However, because of its sensitivity to illumination variations, changes in illumination (especially in chromaticity) can be detrimental to some models.

Use of color in facial skin detection may not be enough: It cannot necessarily distinguish faces from other objects (e.g., hands, wood) with a similar appearance. Other cues are therefore needed to verify that the selected area is indeed a face. Color is, however, still useful as a preprocessing step because it may significantly reduce the number of candidates by eliminating obviously false targets.

In this chapter, we try to answer the following questions: What factors should be taken into account when selecting a skin color model? How can we find a suitable model?

The chapter is organized in the following way: Section 1 provides an introduction to the problem. In Section 2, the behavior of skin color in different color spaces is considered. Section 3 presents a review of skin color models, and Section 4 describes the skin locus approach and compares it to other methods. Finally, discussion is presented in Section 5.

1 Introduction

Color is a low level feature that can be a powerful, highly discriminative and computationally fast cue. It is easy to understand and robust toward some...