Reducing Computational Complexity of New Modified Hausdorff Distance Method for Face Recognition Using Local Start Search

Abstract

Average Hausdorff distance that is an efficient measurement is widely used in face recognition method for measuring the dissimilarity between two sets of features. The New modified Hausdorff distance (MMHD) is a face recognition method, which uses average Hausdorff distance for measuring the dissimilarity between two sets of dominant points, which are features of face image. However, the disadvantage of the average Hausdorff distance is high computational complexity. Various methods have been proposed in recent decade with the purpose of reducing the complexity of Hausdorff distance computing. Local start search (LSS) is a state-of-art method for reducing the complexity of the Hausdorff distance computing. In this paper, we present how to use the LSS method for reducing the complexity of the computing the average Hausdorff distance. Firstly, a modification of the MMHD method, namely Least Trimmed New Modified Hausdorff distance (LT-MMHD) is proposed. The LT-MMHD method uses average Hausdorff distance of largest values for measuring the distance between two sets of dominant points. The proposed method gives higher recognition rate than the MMHD method for all conditions of face image. Finally, the LSS method is used for reducing the computational complexity of the proposed method. Experimental results show that by using the LSS method, the proposed method could reduce the computational complexity of 17%.