A Dynamic Extraction of Feature Points in Lip Regions Based on Color Statistic

This paper deals with the problem of insensitivity and low-effect in the detecting feature points in lip regions. We propose a new feature points extraction model based on Harris algorithm, combining ideas of color statistic and graph connectivity. To detect feature points, firstly, we build a proper possibility statistical module of color component, with a specifying color model in lip regions, then we realize feature points filter module under the thought of graph connectivity. Experiments show that our algorithm has a better accuracy and stability, especially a competitive advantage in lip regions feature extraction.