A comparative study of face recognition using discrete orthogonal moments

Abstract

This paper presents the comparative study of face recognition using discrete orthogonal moment namely Krawtchouk moments (KMs) and Tchebichef moments (TMs). Both these moments do not require numerical approximation and coordinate space normalization. The complex computation of radial polynomial as order becomes larger is not an issue and this makes KMs and TMs superior compared to continuous orthogonal moments in terms of preserving the analytical properties needed to ensure minimal information redundancy. With these properties, KMs and TMs are well suited as pattern features in the analysis of two-dimensional images. The selection of orders and parameters of KMs determines the ROI and to obtain the reconstructed face image similar to the original face image, both parameters of KMs are set equal at 0.5. As for TMs, only the selection of order is required. The experiments were carried out on the database face images from the AT&T Laboratories Cambridge University consisting of 40 distinct subjects of 10 non-similar images each. For some subjects, the images were taken at different times, varying the lighting, facial expressions (open/closed eyes, smiling/not smiling), facial details (with and without spectacles) and different face scale. Euclidean square distance or Nearest Neighbour (NN) is used as the classifier in the recognition stage. From the experiments, KMs showed better performance as compared to TMs in terms of classification accuracy.