Component-based Thermal Face Recognition

Face recognition based only on the visible spectrum has shown difficulties in performing consistently in uncontrolled operating conditions. Face recognition using different imaging modalities, particularly infrared imaging sensors has become an area of growing interest in recent years. In this paper, we present a new technique for face recognition that exploits the local statistical characteristics of a thermal image. The “whole” face image is divided into components of different sizes. The statistical features of these components, beside the “whole” image are combined together using fusion methods. Decision level fusion finds a combination of multiple statistical patterns to produce an integrated result that is enhanced in terms of information content for pattern recognition and classification. Local representations offer robustness against variability due to the changes in localized regions of the objects. The proposed feature vector consists of different moments’ calculations and thermal components’ histograms. The features found from local analysis are less sensitive to illumination changes, easier for estimating the rotations, have less computational burden and have the potential to achieve higher correct recognition rates. The experimental results reveal that the new system can achieve a success rate of 96.4% when implemented on the AIAOU Database.