Fusion affine moment invariants and wavelet packet features selection for face verification

Abstract

Face recognition under changing lighting conditions is a challenging problem in computer vision. The effect of pose and illumination on face images result in a low accuracy rate in the face verification system. In this paper, we proposed a new framework for solving the above problem and increase the recognition rate of the system. This framework performed on a two dimensional (2D) grayscale face dataset based on the combination scheme that consists of the Voronoi diagram (VD), wavelet packet transform (WPT), affine moment invariants (AMIs) and radial basis function (RBF). The preprocessing technique consists of square morphological, Gaussian low-pass filter, median filter, and histogram equalization, which is applied to solving the problem of pose and illumination on facial images. The face crop is obtained with an arbitrarily window template that is designed by using the distance of eyes, nose, and mouth. All face images reduced the information by transformation with the WPT. The AMIs were applied to select the unique features that are used to classify and verify on the RBF neural network. This work was used on the bed face dataset for experimental use, which yielded good recognition rates and was easily implemented when compared with another technique. The result showed the efficiency of the wavelet coefficient and affine moment invariant features that can be used to verify an image lacking illumination.