Face Detection

Face detection in still images and videos has been extensively studied over the last two decades. Attributed to the recent proliferation of cameras in consumer applications, research in face detection has gradually transformed into more unconstrained settings, with the goal of achieving performance close to humans. This presents two main challenges: (i) in addition to modeling the facial characteristics, understanding the information portrayed by the surrounding scene is important in resolving visual ambiguities, and (ii) the computational time needed for decision making should be compatible for real-time applications, since detection is primarily a front-end process on which additional knowledge extraction is built upon. This chapter begins with a review of recent work in modeling face-specific information, including appearance-based methods used by sliding window classifiers, concepts from learning and local interest-point descriptors, and then focuses on representing the contextual information shared by faces with the surrounding scene. To provide better understanding of working concepts, we discuss a method for learning the semantic context shared by the face with other human body parts that facilitates reasoning under occlusion, and then present an image representation which efficiently encodes contour information to enable fast detection of faces. We conclude the chapter by discussing some existing challenges.