A User-Adaptive Image Browsing System with Summarization Layout for the Personal Photo Collections

Abstract

Users spend much time organizing photos into small groups as part of photo management. Selecting good quality photos and organizing them is burdensome, as photographers amass large number of photos. This paper presents a new photo layout system with representative photos considering multiple features. Our approach consists of three steps to deal with hundreds of photos. First, we construct photo clusters by user-adapted criteria: temporal context, the number of faces, blur and luminance metrics. Then, we construct a bipartite graph that consists of photo nodes in a partite set and the constructed cluster nodes in other partite set. The representative photos of each cluster are selected by a maximal matching algorithm based on user-controlled multiple criteria. Finally, our system places the selected representative photos on a 2D grid using the placement algorithm of PHOTOLAND. Other photos in each cluster are displayed in an upper layer of a screen when the user clicks the representative photo. We conducted an experiment based on a user study; it used nine photo sets taken on a trip. The experiment showed that our system conveniently managed hundreds of photos, summarizing and visualizing them.