Face Classification using Optimum Features of LPWT Face Images

N. Kato and S. Wada (Japan)

Keywords

Face classification, Logpolar transform, Wavelet transform, Higher order local autocorrelation, Dimensional reduction.

Abstract

In this paper, we propose a robust face classification method using optimum feature vector of LPWT (Log polar wavelet transform) face images. Face feature vector is extracted from LPWT face image using HLAC (Higher order local autocorrelation). The extracted high dimensional face feature vector is transformed to optimum low dimensional feature space by MNLM (Membership-based nonlinear least-square minimization) method to improve classification rate. In order to show the effectiveness of our method, computer simulations were executed. Actual scale and illumination degraded face images were used to classify face image. The classification accuracy was evaluated compared with several combined methods such as PCA, LDA and spectroface. It is shown that the proposed method using the optimum face feature vectors achieve high classification rate for degraded face images classification compared with the conventional standard approaches.

Important Links:



Go Back