EMoTracker: Eyes and Mouth Tracker Based on Energy Minimization Criterion

We introduce a novel approach for online facial components tracking based on energy minimization criterion. The tracker, known as EMoTracker, employs template matching as the principal technique. As feature appearance changes during tracking, template matching suffers in providing good detection results. Therefore, instead of utilizing only the similarity (correlation values) independently, we add global constraints of facial components placement on face as additional parameters when searching corresponding components. In order to define the correct areas, we first list out n areas (which are the candidates) for eyes and mouth employing template matching technique. These candidates are arranged in high to low correlation order. Selections of correct candidates among these candidates are made based on energy minimization criterion. Additionally, an automatic feature selector and an adaptive face model have been incorporated with EMoTracker to handle tracking from multiple type of faces and non-frontal faces, respectively. Our proposed method also requires no manual initialization and parameter tuning.