Abstract
There exists a large amount of movies that record the brilliant performance of many actors. However, most existing methods that generate 3-D face models of an actor require specific camera setting. A method that estimates a sequence of 3-D face models is proposed. It uses a generic wireframe face model as the initial guess to represent the 3-D structure of a face. In order to estimate a more accurate 3-D face model, we have developed an algorithm that can estimate the z-coordinate of fiducial points of the wireframe by using an additional frame that contains the head at a different pose. A pose estimation algorithm is also developed to estimate the pose between two adjacent frames. Experimental results show that the algorithms can estimate the 3-D face models and the pose with good accuracy.
Original language | English |
---|---|
Title of host publication | the Regional Inter-University Postgraduate Electrical and Electronic Engineering Conference |
Number of pages | 6 |
Publication status | Published - Aug 2003 |
Externally published | Yes |