Robust facial element extraction under lighting variation

It is important, in biometrics and human-computer interaction (HCI), to extract features stably from facial elements. In this study, we propose a robust method of extracting facial elements under lighting variation, which is the most important environmental factor. In our experiment, we extract the mouth contours, which are the most deformable elements of the face by using deformable template matching. We conclude that stable extraction from the sequence of facial images was successful on the basis of the results of our experiment.