PhotoSim: Tightly Integrating Image Analysis into a Photo Browsing UI

Current photo browsers for personal photo collections mostly use the folder structure or camera-recorded time stamps as the only ordering principle. Some also allow manually provided meta-data (tags) for organizing and retrieving photos, but currently, only professional tools allow a pre-grouping of photos by image similarity. We believe that similarity is indeed a useful criterion both for image retrieval and casual browsing, and that the tight integration of content analysis techniques in media UIs in general can lead to more powerful UIs. In this paper we present a prototype, in which we have tightly integrated image analysis techniques and user feedback into a zoomable user interface for browsing and sorting digital photos. We have discussed our system with domain experts and received encouragement as well as valuable ideas for future research.