Human Face Detection Using Skin Color Segmentation and Watershed Algorithm

Face detection receives immense interest in computer vision to improve security and authenticity of a particular system. Color provides useful information at the early stage of face detection in a complex view. Such detection involves many complexities such as background, illumination, and poses. This study provides depth analysis on most prominent color models. The use of those color models can handle well-defined problems in face detection such as occlusions, poses, and illumination conditions. The application areas, techniques used, remarks as well as statistical conversion of the color models from Red Green Blue (RGB) color model are demonstrated. Moreover, a new framework for efficient face detection using skin color segmentation is proposed. The process involves transforming the face images from RGB to the selected color models; then segmentation is carried out by selecting a threshold value for each of the color models. Watershed algorithm is applied to isolate the facial feature from the background. Finally, lips area is localized as it may be missing during the detection process. Detection rate of up to 97.22% was obtained using standard database. The proposed framework targets a range of applications such as PC login security, passport authentication, and pornography filtering.