Browsing Personal Images Using Episodic Memory (Time + Location)

In this paper we consider episodic memory for system design in im­ age retrieval. Time and location are the main factors in episodic memory, and these types of data were combined for image event clustering. We conducted a user studies to compare five image browsing systems using searching time and user satisfaction as criteria for success. Our results showed that the browser which clusters images based on time and location data combined was signifi­ cantly better than four other more standard browsers. This suggests that epi­ sodic memory is potentially useful for improving personal image management. 1 Introduction As digital images are becoming increasingly popular, it becomes necessary to sup­ port users of digital cameras with software tools for personal image management. Most digital cameras time stamp each photograph at the time the image is created. More recently, digital cameras with the Global Positioning System (GPS) function have been developed. These cameras record the geographical coordinates at which each picture file is taken, and then using gazetteers, annotate each picture with loca­ tion data. In this paper we describe a system which combines time and location data for the automatic identification of episodes or events, since these are the main factors in human episodic memory [4]. We perform user studies to show that this identifica­ tion of events helps users to browse their personal collections of digital photographs. Graham et al. [10] and Platt et al. [15] have previously shown that browsing and retrieving images arranged by time is quicker than for unindexed images. Graham [10] also suggested that the use of GPS location data could be used to assist browsing. Naaman et al [14], built a time and location hierarchy clustering system, but did not evaluate it on real users. In section 2, this paper will review some previous studies related to human epi­ sodic memory. Section 3 describes the development of our time and location cluster­ ing system. Section 4 describes our experimental design, where subjects were asked to perform both general and specific scenario image retrieval tasks using five differ­ ent image browsing systems, and retrieval performance was assessed by speed of searching and user satisfaction questionnaires. Both user and system centre evalua­