Smart management system for digital photographs using temporal and spatial features with EXIF metadata

Due to the popular use of digital cameras and the growing capacity of storage, managing a large collection of digital photos is a burdensome job for the average customers. One distinct feature of current digital photos is that image contents are embedded with metadata EXIF, which varies among digital camera manufactures. Since these metadata have abundant information about the photographing environment, it can provide useful hints for managing photos. Most previous digital photo clustering methods were mainly dependent on the timestamp of the photo taken, so users can not always find the intended photos especially if adjacent pictures have very small time gap. The timing gap of digital photos is not a sufficient and reliable clustering condition to satisfy the average customer. In this paper, we propose a novel parameterized clustering system for digital photos by exploiting temporal (time gap between adjacent photos) and spatial features (content similarity on color pixel domain), so each user can adjust his or her own clustering parameters according to the preference between event (temporal condition) and people (spatial content condition). In order to compute the spatial similarity, we applied the adapted color weight function depending on the distribution of a quantified color set. This enabled us not to use the dominant background color in image similarity matching. We also propose a new content matching algorithm called the block matching-expansion procedure. In this experiment, we compared the result with Cooper's most recent work. Using a set of testing 54 photos, we obtained 4 different clusterings from 4 average photographers' manual work. For each manual clustering, we could find a near optimal parameter (balancing temporal and spatial clustering), which were all superior to Cooper's clustering using temporal condition only.