Automatic facial motion tracking and retargeting

Abstract

With the growth of computer animation, facial motions tracking process is becoming an important task. There are several solutions to this problem. However, most of them require to set up a motion captured database which is either store as parameter set or some fixed 3D models. All this pre-process information need extensive amount of time together with expensive hardware to obtain. In this report, we propose a new approach to this problem by using Work Minimization Morphing technique. It helps to automate the motions tracking process without using any other information than the input video frames. The method, however, does not aim to track the motions in real time. In order to make up to the pre-processed information, it requires a substantial amount of time to process. Therefore, only offline facial animation should be considered. Our approach includes several steps from capturing, normalizing and scaling to retargeting motions. This report will explain and summarize the algorithms that we use, including an adapted version of auto-morphing technique.