國立臺灣大學資訊工程學系Huang, Chen-HsiuChen-HsiuHuangShen, Chih-HaoChih-HaoShenHuang, Chun-HsiangChun-HsiangHuangWu, Ja-LingJa-LingWu2006-09-272018-07-052006-09-272018-07-05http://ntur.lib.ntu.edu.tw//handle/246246/20060927122916835934In this paper, an MPEG-7 based album system for managing consumer-generated photographs is introduced. Face detection and recognition technologies are adopted to implement the query-by-face functionality of the content-aware album. A convenient face database construction and training mechanism is also provided. By utilizing both user attention model for visual objects and face information, the most representative region within a photograph, the so-called photo focus, can be easily extracted. Furthermore, with the aid of photo focus and objects importance calculated based on user attention models, two intelligent thumbnail creation approaches, superior to the commonly used sub-sampling method, are also proposed. At last, the face information is incorporated with MPEG-7 visual descriptors to facilitate the similarity calculation between images, allowing users to perform similarity search operations in our album system. The major functionalities introduced have contributed to form a content-aware album system for consumer-generated photographs.application/pdf887939 bytesapplication/pdfzh-TWMPEG-7a content-aware album systemquery by facesmart thumbnailphoto similarityotherhttp://ntur.lib.ntu.edu.tw/bitstream/246246/20060927122916835934/1/cm_ncs2003.pdf