Graph-based event clustering and its visualization for photo albums

Abstract

Clustering and visualization of photos have recently attracted much attention with the goal of organizing personal photo albums, and they are also useful in other applications such as image annotation, photo browsing, video summarization, contents curation services, and so on. Although there are numerous album management systems, organizing personal collections still remains a challenging issue. In this thesis, we propose a graph-based event clustering and its visualization method for personal photo albums. Generally, photos in the same event are close in terms of time and visual contents. Thus, we use two types of data information (i.e., time and visual content) for clustering. Our graph-based clustering method can allow for a wide time range of events. Then assessment of aesthetic photo quality and visual similarity is conducted for cluster visualization.From experimental results, we confirm that the proposed algorithm provides reliable organization results for various personal albums.