Real-time live face detection using face template matching and DCT energy analysis

This paper presents a method to detect the face of a live test subject in real time using a single camera as the image acquisition source. The objective is to improve the identification capability of biometric authentication systems based on face recognition by serving as a preprocessing tool to reject faces that do not belong to live humans and which can spoof the system into granting access. The face detection procedure involves skin area selection in the YCbCr color space and matching with a reference face template. Morphological operations such as erosion, dilation etc. have been avoided due to the uncertainty in the number of such operations that may be required under a varying real-time environment and also to reduce the computational cost of the algorithm. There are few parameters that can be considered as indicative of a live face, for example- pupil movements, nostril movements, 3D depth, and difference in number of high frequency components in live faces and photos etc. The proposed method involves the analysis of the image´s DCT energy aided by detection of eye blinking and pupil movements. This yields a compact security scheme capable of resisting attacks by photographic imitation.