Lower Facial Curves Extraction for Unconstrained Face Detection in Video

Abstract

Face recognition plays vital role in many of the biometric as well as other scientific applications. Face needs to be detected prior to recognition, and many researchers introduced methods for face detection along with unique features. These methods have been used in constrained and unconstrained environment. But it is difficult to tackle challenges such as occlusion, pose variation, illumination order to detect faces in unconstrained environment. Here in this paper, we are going to extract facial curves as features which will be further used for face detection. We have extracted lower facial features such as chin curves, lips curves, and ear curves by obtaining face contours using edge detection techniques in images and video frames from YouTube Faces Database. By matching majority of features, it is possible to detect face in unconstrained environment.