World Scientific Publishing Company A SYSTEM FOR HUMAN FACE AND FACIAL FEATURE LOCATION.pdfVIP

International Journal of Image and Graphics Vol. 3, No. 3 (2003) 461-479 ? World Scientific Publishing Company A SYSTEM FOR HUMAN FACE AND FACIAL FEATURE LOCATION JUN MIAO, HONG LIU, WEN GAO, HONGMING ZHANG, GANG DENG, XILIN CHEN ICT-YCNC FRJDL, Institute of Computing Technology, Chinese Academy of Sciences, Beijing 100080, China Email:{ jmiao,hliu,wgao,hmzhang,gdeng,xlchen}@ This paper presents an implementation of a system designed for location of human faces and facial features such as pupils, eyes, nose and mouth. The kernel of the system is an integration of several algorithms, such as human face center -of-gravity template, illumination compensation, and so on . A false-face removal algorithm is proposed in this paper especially for the distinguish ing of cartoon faces from true faces. The testing experiments of the system show quite good results, the average detection accuracy rates for face detection and facial feature location are 97.8% and 87.5% respectively. Keywords: face center-of-gravity template, face detection, facial feature location 1. Introduction Human face detection and facial feature location, as specific issues of pattern recognition, are significant in various applications such as human identification, face retrieval, virtual human face synthesis, and model based MPEG-4 face coding. A lot of work has been done on this issue. Proposed techniques include multi-resolution analysis1,2 , fixed template matching3, deformable template matching4, center-of-gravity template matching5, fuzzy theory6, principle component analysis “eigenface”7 or “eigenpicture”8, neural network9,10,29,30, support vector machine11, and usage of multiple information including color, sound or motion12,13. Quite a few techniques are related to facial feature extraction, such as B-splines14, adaptive Hough transform15, cost minimization16, deformable template17,18, regio