Abstract

The advent of digital cameras has provided photographers, with varying levels of expertise, the opportunity to accumulate large repositories of digital images. However, this expansion has also brought the attendant difficulty of image retrieval. This paper reviews the considerable work already carried out on image retrieval and identifies critical constraints in attempting to handle the underlying semantics of photographic images. The authors address the issue of how an amateur photographer, storing several thousand images a year, can effectively and efficiently manage a personal collection. A number of surprisingly simple techniques are proposed utilising a sparse lexical ontology alongside iconic visual clues to identify key representative events in the collection. The photographer’s associated knowledge and experience, often shared by others who are attuned to the context of the collection, are critical to the effective management of the images. The authors recommend the use of existing browsers and a simple keyword table to facilitate concise and effective image storage and retrieval. Photographers can further enrich the representation of their collection by clustering images in accordance with their perceived relevance and aesthetic qualities. The authors argue that these properties can also be exploited to enhance retrieval