Internal-to-internal transition method for consecutive hierarchical template matching

Abstract

This study proposes a method reducing the tree traversal cost by first investigating the most probable node instead of the root node when a hierarchical template matching is consecutively applied to the object. In particular, this study gives a novel viewpoint that a consecutive hierarchical template matching could be regarded as a transition of a hierarchical finite state machine, and then it proposes a novel method considering the transitions between the internal nodes of the hierarchical template tree. The proposed method is verified by applying it to pedestrian silhouette detection.