An approach for fast human head pose estimation
Yessenia Yari, Jacob Scharcanski · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 2011
This paper proposes a simplied fast method to estimate the head pose in monocular images. The head pose is estimated comparing the position of the landmark points localized in the face, which correspond to pre-selected points of a 3D face model. To localize these points on a face image (such as the eyes, mouth and nose) we use Active Shape Models. These points are detected in the 2D space, and next the 3D face model is adjusted by geometric transformations to the 2D points on the face image to estimate the head angular position in the 3D space. Our preliminary experimental results are encouraging, since they show that our simplied approach has a competitive accuracy when compared to more sophisticated methods available in the literature.