An Enhanced Face Anti-Spoofing Model using Color Texture and Corner Feature based Liveness Detection

Abstract

In recent years, Biometric security systems have extended their uses. The systems are able to identify humans by analyzing their behavioural characteristics. Face recognition is the most popular biometric techniques, which widely used nowadays. They are treated as a suitable replacement for PINs and passwords for regular users. It is very easy to use a photo imposter to fake face recognition algorithm. To ensure the presence of real human face to a photograph or 2D masks, an enhanced face anti-spoofing model is proposed using Color Texture and Corner Feature based Liveness Detection (CTCF_LD). From the input video, the frames are extracted and cropped for the specific facial landmark points. The texture of the 2D masks and real face is analyzed by changing its colorspace. Then, the corner points are detected using various corner detection algorithms. Based on the corner points, the fake face is differentiated from the real face using a threshold value. Empirical study shows that the proposed CTCF_LD face anti-spoofing model with HSV_FCD algorithm gives better accuracy of 88%.