Estimation of Face Pose Orientation Using Model-Based Approach

Abstract

In the domain of computer vision and pattern recognition, though there are numerous methods for face recognition, it is still remaining as a very challenging problem in real life applications. Face detection and recognition suffer from many problems which are caused by the variations in orientation, size, illumination, expression, and poses. This paper mainly revolves around face detection and oriented pose identification. The state-of-the-art Constrained Local Model (CLM) is applied to detect the face from any wild facial image. The extracted feature points are used to segregate the dominant parts of faces. From the dominant feature points, nose tip and eye points have been identified. Applying the geometrical parameters between the nose tip and eye points, the pose orientation of the wild face has been identified. This method is very simple and accurate. The performance evaluation has been done on unconstrained Essex database and internal wild database collected from internet.