The use of digital cameras is prevalent. Although the cost of digital photographs is low, managing numerous digital photos is burdensome to most users. An intelligent management tool for digital photos is needed. We propose a novel clustering algorithm for concurrent digital photos obtained from multiple cameras. Since previous photo clustering methods can be applied to a single camera, a group of photos obtained from different cameras cannot be classified to meet user preference. We newly define temporal/spatial combined clustering for the set of group photos taken from different cameras to solve this situation. We define a new spatial similarity using block alignment for two independent photos. If a user submits photo clustering that shows preference between spatial and temporal clustering, then we can cluster other photo sets according to the reference clustering characteristics. In this method, the EXIF metadata plays an essential role. We tested more than one thousand photos taken by tourist groups. The final result was satisfactory compared to previous methods based on temporal (spatial) criteria only.

According to the popular use of digital camera, a traveler group carries multiple cameras for the same event. Previous studies on digital photo focused on massive unrelated photos or collections of a private user. But group travelers can carry several cameras and take photos simultaneously at the same time. Thus, an effective management for group photos is main issue for a traveler group. Since group photos from different photographers share their content, people need to collate the photos and classify them. We propose several supervised and unsupervised clustering methods for group photos. Previous studies are not applicable to group photos, because group photos do not guarantee clear relevance between photos which shown in private photo album. The proposed supervised clustering method, using spatio-temporal similarity, obtains a true cluster set of a specific camera from a user. It extracts discriminating features from given clusters and applies them to cluster other photos. Unsupervised methods use temporal photo blocks to compensate spatial variation of photos from a user. We use hierarchical clustering and neural network based clustering. In experiments, we show clustering results from real nomadic photo data. People can use a method suited to their needs.

No abstract

No abstract

No abstract