Skip to Main Content
Widespread distribution of digital cameras and the low price of memory are acclimating people to taking many more photos than before. As the number of photos to be managed grows, photo management tasks become more and more burdensome. One of the most time-consuming aspects of photo management is photo classification. A person must make many folders, group photos, and label grouped photos. In order to assist in this task, we proposed a priority queue-based hierarchical photo clustering method that partitions ordered photos by their temporal context. The temporal context we considered is the variance of time gaps between photos taken in a window (its size is user-designated). A recursive partitioning called by a priority queue enables classified photo groups with maximum variance to be selected primarily. In an experiment, we compared our system with Cooper's temporal clustering method in terms of precision and recall. Our method is superior when the photo timestamp difference has a uniform pattern. For further work, we propose a visualization method for the clustered photos.