Pose estimation system of 3-D human face using nearest feature line in its eigenspace representation

In this paper, a pose estimation system is developed using a minimum distance calculation of projected unknown viewpoint of a spatial image into its eigenspace representation to the nearest line of the two known viewpoints in the same eigenspace representation. In order to have a higher recognition rate on determining the pose position of the unknown image, we developed FullyK-LT and SubsetK-LT methods. The developed system is performed to determine the pose position of 2-D images taken from the human model by gradually changing visual points, which is done by successively varying the camera position from -90 to +90 with an interval of 15 degree. The experimental results shown that the highest recognition rate of the system is about 57.5% when using FullyK-LT method, and could be increased up to 93.8% when using SubsetK-LT method.