A Hierarchical Fuzzy Based Approach for Face Detection

In this paper, we propose an algorithm for face detection in 2D intensity images. The method has two steps: extracting face candidates and filtering out the false positives. In the first stage we provide a face model by combining intensity and edginess information of face image. The extracted features are fuzzified to overcome the problem of variations in imaging system. We provide a fuzzy model for the most representative parts of face (lip, eyes and forehead) . The applying first part on input image results a number of locations candidate as face. In the second part we use a neural network to filter out false positives from the list of candidates for face. We train this part using face and non-face images.