An efficient eye location using context-aware binarization method

Abstract

Face recognition system needs the method of locating facial components like eyes, mouth, etc, for extracting features used in recognition process. In this paper we propose a method of locating eyes using context-aware binarization. The proposed method consists of binarization, connected region segmentation, eye candidate area extraction by heuristic rules that use geometric information, eye candidate pair detection, and eye area pair determining by ranking method. Binarization plays an important role in this system that converts a source image to a binary image suitable for locating eyes. We consider edge detection based and segmentation based binarization methods. However, each method cannot be a solution in general environment because these are influenced by the factors such as light direction, contrast, brightness, and spectral composition. We propose a hybrid binarization using the concept of illumination context-awareness that mixes two binarization methods in general environment. We apply this methodology to eye location, and we achieved encouraging experiment results in general environment. © Springer-Verlag Berlin Heidelberg 2005.