CHArt Eighteenth Annual Conference
|
DIGITAL
ART HISTORY? Exploring Practice in a Network Society |
Annette A. Ward, Margaret E. Graham, K. Jonathan Riley, Institute for Image Data Research, University of Northumbria at Newcastle, UK, Nic Sheen, Ibase, UK
Enhancing an Historical Digital Art Collection: Evaluation of Content-Based Image Retrieval on Collage
Searching digital art images is increasingly challenging as collections and users grow. Text retrieval alone may be inadequate. Content-Based Image Retrieval (CBIR) retrieves visually similar matches for a selected painting, sketch, or other image based on colour, texture, and shape. CBIR was added to the Corporation of London Guildhall Library and Art Gallery's 22,000 digital image collection, Collage. Collage is comprised of images of London from the fifteenth century to the present day and includes paintings, drawings, prints, sculptures, and other historical items.
Text-retrieval of Collage images is left unaltered. Instead, the optional CBIR search is conducted after an image is selected. Evaluation of CBIR on Collage is conducted through an on-line questionnaire. Results from 181 respondents are positive regarding the CBIR addition. Approximately 80% indicated that CBIR searching was interesting and would use it again. Nearly 75% reported that CBIR was a good method of image retrieval. Additional results provide information regarding the usefulness of CBIR, satisfaction with the results, enjoyment of the experience, and demographic characteristics of the respondents.
Our presentation will provide detailed description of the technical application of CBIR to Collage, report comprehensive results of the eighteen-month user evaluation, and delineate the potential of CBIR for application to other art collections.
This research was funded by Resource: The Council for Museums, Archives and Libraries (CMAL/RE/103).