Pose estimation in real time video using Tri-Line method
A. Vanitha, V. Vaidehi · 2016
Human head pose estimation has become an important issue in the field o f computer vision. Over the last decade, many approaches have been introduced to estimate head pose. However, estimating head pose in real-time has proven to be a difficult task. to overcome the drawbacks in existing system, this paper proposes Tri-Line based face pose estimation method for real time videos. The proposed method is faster and accurate compared to other existing head pose estimation methods and also updates the face poses dynamically. In Tri-Line method, pose estimation is calculated based on the distance between the locations of facial landmarks. The proposed Tri-Line method is tested in different databases like YouTube Celebrity database, NRC-IIT Face Video database and manually created Video database for both controlled and uncontrolled environment. Experimental results show that the accuracy of the proposed method is 90 (%) in yaw and 85(%) in pitch on Pointing'04 database.