Decision-level fusion of infrared and visible images for face recognition

Abstract

The nature of the imaging environment, illumination plays an important role in the efficiency of face recognition on visible images. Infrared image is independent of the ambient illumination, but it is sensitive to temperatures. Face recognition algorithms applied to the fusion of IR and visible images consistently demonstrated better performance than when applied to either visible or IR imagery alone. An approach based on decision-level fusion of infrared and visible images for robust face recognition is presented, combinatory of linear weighted sum and biggest match score. The combination of PCA and Linear Discriminant Analysis method was used to extract and recognize face feature. In order to achieve the final recognition result, the decision-level fusion was implemented by previous outcome of infrared and visible images recognition and their confidence measure. The experiments have shown it improves the performance and adaptability of face recognition in lots of actual application environments.