Multiple Faces Recognition Based on Skin-color Regional Segmentation and Principal Component Analysis

• Main Authors: Chi-Tang Chen, 陳繼棠
• Other Authors: 劉倫偉
• Format: Others
• Language: zh-TW
• Published: 2006
• Online Access: http://ndltd.ncl.edu.tw/handle/22766363094908786942

碩士 === 國立臺灣海洋大學 === 機械與機電工程學系 === 94 === This paper combines Skin-color Regional Segmentation with Principal Component Analysis to realize an automated system for multiple faces detection and recognition. This system is divided into two sections: the multiple faces detection and the face recognition. On the part of multiple faces detection, we utilize the skin and lip colors to segment the original image and transform the images of skin and lip areas into binary images. Next, image morphology and logical operation are used to detect the facial areas. Multiple facial areas are then marked. On the part of face recognition, the facial image training sets are adjusted to the size of pixels by affine transformation and image fix and then served as data bank. Then the height /width ratio of facial image is used to filter out unlike candidates. Finally, we use PCA to compare the facial images to data bank, and minimum Euclidean distance is used as criterion for facial recognition.