A Photo Management System Based on Face Recognition

• Main Authors: Bing-Cheng Hsieh, 謝秉承
• Other Authors: 張厥煒
• Format: Others
• Language: en_US
• Published: 2009
• Online Access: http://ndltd.ncl.edu.tw/handle/vt5upc

碩士 === 國立臺北科技大學 === 資訊工程系研究所 === 97 === With the increasing availability of digital cameras, one can easily collect a large number of photos. In this paper, we present a prototype system aimed to achieve efficient management of photos of people. Using current face detection and recognition technologies, we provide both automatic and semi-automatic methods to facilitate the job of associating photos to people. A series of preprocessing methods are taken to improve recognition rate for faces captured in varying lighting conditions and poses. We found that a personal photo collection of a single event may contain only a limited group of people. We can set up a face classifier specific for the group of people. A threshold scheme separates people not in the group or faces that can not be confidently identified apart. The organized result helps pick out incorrectly identified faces. A semi-automatic approach is used to deal with the leftovers by giving three candidates to choose from. Our proposed system is equipped with comprehensive user interface designed for these tasks. We conducted a simulation of this usage showing that it can dramatically reduce the amount of work on a very large set of photos.