Face Recognition using Geometric Measurements, Directional Edges and Directional Multiresolution Information

Abstract

This paper propose the extraction of geometric measurements of face regions, different directional information and multiscale information offered by real tree dual discrete wavelet transform to generate facial features for face recognition. The proposed scheme calculates feature sets: geometric measurements such as the area measures of eye, nose and mouth, horizontal, vertical, and diagonal directional edge information and multiscale information using whole face information. Existing feature-based or local features based methods, depends on characterization of individual facial features (i.e., eyes, nose, and mouth etc.) and their geometrical relationships. Our approach automatically calculates the measures of important areas of interest that describe the information present in an individual face image. The area of different segmented regions is calculated that describes the information of facial components. The proposed geometric features, directional information using Fast Wavelet transform and multiscale image formation gives the rotation, pose and expression invariant face recognition. Fusion of various generated features improves the recognition considering the different segmented face information. The success of the scheme is the automatic calculation of various geometric features, directional edge information and whole face information. The main features of our approach are detection of important regions, calculation of geometric features and fusion with information offered by edge and multiscale information. We demonstrate the success of our approach by experiments.