An image processing and belief network approach to face detection

This paper describes an approach to face detection, which is the first stage of any fully automated human face recognition system. We propose several enhancements to a feature-based approach described by K.C. Yow and R. Cipolla (1997) in an attempt to obtain more accurate results: namely, the attentive feature selection and grouping phases are modified in order to deal with incomplete feature detection while, at the same time, reducing the number of candidates and candidate groups considered