Robust Lip Tracking with Improved Active Shape Models
Suhuai Luo · 2002
This paper introduces a robust lip tracking algorithm based on improved active shape models. After introducing the principle of active shape models, our two main achievements in dealing with lip tracking are presented. One is that a robust model fitting procedure substantially differing from the original active shape model approach has been developed. The other is that tracking on each new frame is based on a priori knowledge about lip appearance and the resultant position and shape of the pr evious frame as well as the current image structure. This combination has secured a consistent and accurate global lip tracking. The algorithm has been tested on various real video sequences and has consistently demonstrated robust performance on tracking deformable shapes.