Acquaintances Clustering for Social Relationship-Based Indexing of Digital Photos

Abstract : One of the effective ways to manage large collections of digital photos is to tag names of people appearing in those photos. However, the number of people appearing in photo collections may range in the hundreds and the names of tagged people are usually presented in alphabetical order or on a first-tag first-place basis. As a result, it is difficult to quickly search for name tags that a user wishes to find. In order to solve this problem, we developed a digital photo management system that automatically groups the name tags based on their social relationships. This system was tested on users' own photos against three other comparison interfaces. The average searching time for name tags was significantly faster with our system. Also, the user satisfaction was higher than the others.