Background Robust Face Tracking Using Active Contour Technique Combined Active Appearance Model ⋆ Jaewon Sung and Daijin Kim Biometrics Engineering Research Center (BERC), Pohang University of Science and Technology {jwsung, dkim}@postech.ac.kr Abstract. This paper proposes a two stage AAM fitting algorithm that is robust to the cluttered background and a large motion. The proposed AAM fitting algorithm consists of two alternative procedures: the active contour fitting to find the contour sample that best fits the face image and then the active appearance model fitting over the best selected con- tour. Experimental results show that the proposed active contour based AAM provides better accuracy and convergence characteristics in terms of RMS error and convergence rate, respectively, than the existing robust AAM. 1 Introduction Active Appearance Models (AAMs) [1] are generative, parametric models of certain visual phenomena that show both shape and appearance variations. These variations are represented by linear models such as Principal Component Analysis (PCA), which finds a subspace reserving maximum variance of given data. The most common application of AAMs has been face modeling [1], [2], [3], [4]. Although the structure of the AAM is simple, fitting an AAM to an tar- get image is a complex task that requires a non-linear optimization technique that requires a huge amount of computation when the standard non-linear op- timization techniques such as the gradient descent method are used. Recently, a gradient based efficient AAM fitting algorithm, which is extended from an in- verse compositional LK image matching algorithm [5], has been introduced by Matthews et. al. [4]. The AAM fitting problem is treated as an image matching problem that includes both shape and appearance variations with a piece-wise affine warping function. Other AAM fitting algorithms can be found in [6]. We propose a novel AAM fitting method that pre-estimates the change of the shape (motion) of an object using the active contour technique and then begins ⋆ This work was supported by the Korea Science and Engineering Foundation (KOSEF) through the Biometrics Engineering Research Center (BERC) at Yonsei University. D. Zhang and A.K. Jain (Eds.): ICB 2006, LNCS 3832, pp. 159–165, 2005. c  Springer-Verlag Berlin Heidelberg 2005