Gray Scale Projection Based Infrared Facial Image Analysis For Identity Verification

Many existing methods for IR face recognition and verification are based on facial features. However, there are some weaknesses of infrared (IR) image, such as low resolution, no texture and colour information, blurring features. These make the approaches based on facial features difficult for IR image analysis. This paper presents results of IR facial image analysis based on gray scale projection. Through these projections, a 2-dimensional IR image is converted to several one-dimensional signatures. We attempt to extract concise features by analyzing projected patterns of the image along one or more directions. In one paradigm, projections along 4 typical directions are performed to get 4 one-dimensional vectors. Different analysis methods for the projected pattern, such as direct pattern correlation analysis, chain code modelling are carried out and compared. To minimize both extrinsic and intrinsic influence, only the center part of the face is used, which is from the eye brow down to the mouth. The features can provide potential information for classification for IR image based identity verification and recognition.