Background and aims

We have been working with the National Gallery in London and the Louvre for around ten years, developing new imaging technologies. During that time various teams developed the high resolution multispectral scanner in VASARI, which is still evolving today in the National Gallery. In the MARC project we helped develop a colorimetric scanback camera producing 20kx20k images. In Viseum we developed a web-based client-server for viewing these images and helped establish CD-r jukeboxes in the C2RMF and National Gallery. In ACOHIR we turned our attention to 3D objects such as pots and sculptures and helped install a 3kx2k multi-view system with a turntable in the C2RMF. The Viseum viewer was adapted to handle these new image groups and a multiresolution JPEG-in-TIFF image format developed.

The HP-donated servers in Southampton are used to mirror developments in London and Paris, especially the database and image server/searching developments being made under the SCULPTEUR project. Many areas of exchange will be helped by having a smaller version of the HP server in Southampton: code portability, images, tapes, maintenance expertise, image format migration and performance tuning. For example we have used DLT tapes to transfer 20GB of images directly from the Louvre, we also made a SCOPYR to TIFF converter for their old files. At the moment we have around 60,000 images from the Louvre, Victoria and Albert museum and National Gallery.

We are running Linux on the server, which has had to be updated to the latest Redhat Linux plus newest server Kernel and glibc. The 6 CPUs operate together well. Our Apache-based Java image server is installed to serve high resolution images. The server is housed in a local area where there is a possibility of wiring in capture or graphics cards in the future. The system is enhanced with external processing nodes in order to investigate farming out time-consuming processes such as feature vector generation.

The large format printer has been installed (1055CM) and the VIPS to postscript program (vips2dj) adapted for it for direct CIELab image printing, with initial tests providing excellent results. Further work on this driver (options and perhaps binary images) will be useful to the National Gallery and Louvre, where similar printers exist. It is also used to print camera calibration test charts for the V&A and ENST.

A mySQL-driven database has been installed, with a PHP front-end. This is used in order to find images and will possibly be adapted to a full content-based retrieval system. Each year different software architectures and human interfaces are tested, sometimes as part of undergraduate projects. In SCULPTEUR we are moving towards using mySQL plugins and trials with autonomous image processing query/generation services.