Online Active Feature Model for lip tracking

This paper introduces a novel method, Online Active Feature Model (OAFM), which can be thought of as a generalization of Active Shape / Appearance Models (AAMs), in which features (e.g. shape and image features) are modeled sequentially in video sequences. As we are interested in more refined aspects of the object such as contours, we propose a two-stage scheme in which a Light Active Appearance Model (LAAM) is used in first stage to track a target object globally, and an Online Local Feature Active Shape Model (OLF-ASM) is used at the second stage to refine object contours. Experimental results of lip tracking show that our proposed tracker is functioning accurately.