See a picture you like but cannot remember the original? Or want to use a motif within a design and need to know its source? Or curious who painted that masterpiece? Thanks to new search and retrieval techniques developed by ARTISTE, you can now find out.
European museums and galleries are rich in cultural treasures, but if you are unable to visit the premises then gaining information on their collections can be a challenge. Many of these institutions have online information facilities, however a lack of standardised systems and techniques has often made access more difficult than need be.
The IST project ARTISTE aimed to tackle this problem by developing content-based search and retrieval systems that could be used by major art institutions. Project participants included several leading galleries across Europe, such as the Uffizi in Florence, the National Gallery and the Victoria and Albert (the V&A) museum in London and Centre de Recherché et de Restauration des Musées de France (C2RMF), the restoration centre for French museums including the Louvre. By the close of the project in 2002, the project partners had developed several new and useful ways of accessing information online on major art collections.
Flexible content-based search
"We developed the 'Fax Finder' algorithm for example, which is designed to match a fax to an image," says Kirk Martinez of the University of Southampton. "We might receive a fax copy of an image from an American professor for example. We scan the fax into the system, which is capable of rapidly sifting through a database of 40,000 images, then receive a list of the top 10 or 20 most similar images."
Another ARTISTE development is the ability to carry out a distributed query across more than one collection, for example those of both the Louvre and the National Gallery. Says Martinez, "Say someone wants to find all 17th Century paintings where one particular colour is used. The same query is sent to both institutions. Each then returns their results, in the form of a set of 10 urls with the highest score."
A search algorithm of a very different type was developed specifically to meet the needs of the V&A museum. Unlike the Fax-Finder search that tries to find the single most likely image, this browser-based facility is designed to carry out a search for colour matches and to present the widest possible choice of original images that match the query. Such a facility can be extremely useful to designers seeking a match with an original design.
ARTISTE has also enabled art institutions to detect potential degradation in their collections. "The Uffizi for example is able to set up a query such as 'What pictures in our collection have developed spider-web shaped cracks?" They were able to find four or five pictures with those issues, and start remedial action before those pieces deteriorated."
Powerful search facilities
James Stevenson of the V&A in London gives a user perspective. "With ARTISTE, we can search our database using combinations of text and image content in the query, which makes the system far more powerful than before. We can submit an image to a database and say, 'Find me an image like this one'. The system can look at aspects of colour distribution for example and find images with similar characteristics."
The system also allows users to go to a colour palette menu and search for images that contain, say, 60 per cent of the same hue. In this way, users can find an original image that is a close match to a design or colour scheme. "We can even take a detail of an image," says Stevenson, "and find its parent. For example in our collection we have thousands of motifs - we can take a motif, or even just a detail of one, and search for its parent."
"We've been working out how to integrate ARTISTE techniques into our current systems," he says. "Although the project made use of a redundant database technology, we have now managed to convert all the features into the open-systems based MySQL.'
Many of the techniques developed successfully under ARTISTE have rolled on into its successor project SCULPTEUR, which has extended the ARTISTE methods of researching two-dimensional images into other media such as 3D models and Quicktime movies. Such abilities are particularly interesting for galleries' sculpture collections.
Providing increased access
ARTISTE project development was driven very much by the needs of the users, emphasises Martinez. "We developed content-based retrieval algorithms for a variety of purposes, say for text, or colour, or colour layouts. These facilities were then tuned to ensure they provided what the users really wanted."
"The project was a success mainly because we were able to see a concrete result which could save people time and effort," he says. "It is not just for professors and professional librarians, but also for the general public. Museums and galleries see it as a good way of increasing access to their collections."
University of Southampton
Southampton SO17 1BJ
Source: Based on information from ARTISTE