A highly scalable incremental facial feature extraction method

Abstract

Face recognition is one of the most challenging tasks in biometrics, machine vision, and pattern recognition. Methods that can dynamically extract facial features and perform online classification are especially important for real-world applications. The potentially most useful methods in these cases would include incremental learning techniques such as Incremental Principal Component Analysis (IPCA) and Incremental Discriminant Analysis (ILDA). In this paper, we propose a novel incremental facial feature extraction method—Incremental Weighted Average Samples (IWAS). The new method is very simple in theory and experimental results conducted on two benchmark face image databases demonstrate that it is more effective and efficient than IPCA and ILDA, making IWAS especially applicable to real-time face recognition.