Face Detection Using the Theory of Evidence
Franck Luthon* Computer Science Lab., University of Pau, 2 allée parc Montaury, 64600, Anglet, France
Abstract
Face detection and tracking by computer vision is widely used for multimedia applications, video surveillance or human computer interaction. Unlike current techniques that are based on huge training datasets and complex algorithms to get generic face models (e.g. active appearance models), the proposed approach using evidence theory handles simple contextual knowledge representative of the application background, due to a quick semi-supervised initialization. The transferable belief model is used to counteract the incompleteness of the prior model due to lack of exhaustiveness in the learning stage.
The method consists of two main successive steps in a loop: detection, then tracking. In the detection phase, an evidential face model is built by merging basic beliefs carried by a Viola-Jones face detector and a skin color detector. The mass functions are assigned to information sources computed from a specific nonlinear color space. In order to deal with color information dependence in the fusion process, a cautious combination rule is used. The pignistic probabilities of the face model guarantee the compatibility between the belief framework and the probabilistic framework. They are the inputs of a bootstrap particle filter which yields face tracking at video rate. The proper tuning of the few evidential model parameters leads to tracking performance in real-time. Quantitative evaluation of the proposed method gives a detection rate reaching 80%, comparable to what can be found in the literature. Nevertheless, the proposed method requires a scanty initialization only (brief training) and allows a fast processing.
Keywords: Belief function, Cautious rule, Classification, Computer vision, Conjunctive rule, Dempster-Shafer, Face tracking, Fusion of information, LUX color space, Mass set, Particle filter, Pattern recognition, Pignistic probability, Region of interest, Skin hue, Source of information, Transferable belief model, Uncertainty management, Viola-Jones detector, Visual servoing.
* Address Corresponding Author F. Luthon:IUT de Bayonne Pays Basque, UniversitĂ© de Pau Pays dâAdour, 2 allĂ©e du parc Montaury, 64600, Anglet, France; Tel: +33(0)5.59.57.43.44; Fax: +33(0)5.59.57.43.49; E-mail: [email protected] INTRODUCTION
Real-time face detection and tracking in video sequences has been studied for more than twenty years by the computer vision and pattern recognition com-munity, owing to the multiplicity of applications: teleconferencing, closed-circuit television (CCTV), human machine interface and robotics. Despite the ongoing progress in image processing and the increase in computation speed of digital processors, the design of generic and robust algorithms is still the object of active research. Indeed, face image analysis (either detection, recognition or tracking) is made difficult by the variability of appearance of this deformable moving object due to many factors: individual morphological differences (nose shape, eye color, skin color, beard), presence of visual artifacts (glasses, occlusions, make-up), illumination variations (shadow, highlight) and facial expression changes depending on context (social, cultural, emotional). Those are difficult to model and do not easily cope with real-time implementations. Moreover, the scene background might disturb detection, in...