Face detection based on a model of the skin color with constraints and template matching

Face detection is now a classical problem in detection, and it was addressed through several methods, including global comprehensive image scanning and structural approaches that focus on the facial structures. Face detection is considered in this paper based on combining both a statistical model of skin color and geometrical face characteristics. The system presented is organized in two parts. The first one consists in skin color detection by a statistical method, based on a Gaussian mixture model in the chromatic CbCr color space. The second part is devoted to processing detected skin regions to select those corresponding to faces. A skin region is considered as a face candidate if it verifies a set of geometrical constraints. And then, a template matching is applied to reach the final decision depending on the degree of similarity between the template used and the region under analysis. Both skin detection part and whole face detection system were tested on face databases.