Object tracking based on Multiscale Morphological Templates

This paper presents a novel template representation that can be used in template-based object tracking methods. More specifically, the Multiscale Morphological Template is introduced and incorporated in a template-based object tracking algorithm. The proposed template can be updated over time to cope with changing environment/object conditions. The algorithm is applied to face tracking in scenes with complex background. Results of the object tracking algorithm using the proposed and existing template representations are compared using measures based on ground truth data. The proposed template is proved to be superior to existing templates.