Skin region segmentation in color images with dominant GMM component analysis

• Main Authors: Pei, Jung, 戎沛
• Other Authors: 張隆紋
• Format: Others
• Language: zh-TW
• Published: 2003
• Online Access: http://ndltd.ncl.edu.tw/handle/97115305611286939531

碩士 === 國立清華大學 === 資訊工程學系 === 91 === Automatic human face recognition has gained lots of attention in recent years. Before face recognition we need the detection and location of faces. Methods that utilize color information for face detection is called color-based face detection. First the algorithm segments skin-tone region from the source image as face candidates, then it search facial features in these regions to locate faces. Color based techniques have relatively low computational time. However, it suffers from numerous of false alarms under uncontrolled environment, like complex background, extreme lighting condition, or head pose. In this paper, we focus our effort on post processing. We proposed a texture analysis procedure and background eliminating scheme to reduce false positives. We also propose a scheme to correct false negatives caused by extreme lighting conditions. The computational time of the proposed algorithm is very low, and the accuracy of segmentation is satisfactory in most cases.