Using fuzzy adaptive fusion in face detection

Abstract

Face detection, either from still images or video frames, is an essential first step in any automated facial recognition system. A novel approach for face detection is presented in this paper. Multiple algorithms are used to process the same face image, but extract different facial features. Since it does not amplify the errors, the sum rule is applied to the score outputs of multiple detectors. Different from the other approaches that use the pre-set weights, a fuzzy model is developed to dynamically generate the weights based on the image quality. The experimental results demonstrate a distinct advantage of the proposed method - detecting face in a near dark environment.