Digital Imaging with JPEG2000

The digital imaging with JPEG2000 (jp2) session was essentially divided into four sections: an introduction to jp2, a description of the standard and its uses, product demonstrations from vendors, and a panel of practitioners. Unfortunately there was some mix up with the busing from McCormick (one bus didn’t take anyone (maybe it was having mechnical problems)) and I missed Peter Murray’s introduction. When I walked into the session, Robert Buckley from Xerox Labs was describing the standard and its uses.

The standard is complicated and Robert Buckley’s description was very technical. He did a good job, though, explaining a complex open standard. The take away point about JP2 is that it uses an arrangement of packets to allow much greater flexibility in how images are used. For instance, when a thumbnail is desired, an image processor only needs to select a small group of packets in the jp2 file to create the thumbnail (rather than having to access the whole image to generate a thumbnail).

The same is true for generating images of higher or lower quality (a lower quality version of the image is retrieved by accessing the right combination of packets). The JP2 standard also supports ROI (regions of interest). An example use for this would be if someone wants to enlarge just a small part of a picture. Using a drawing tool, a random region in the JP2 could be selected and enlarged. This JP2 feature allows for great flexibility when navigating and exploring a complex image.

Another feature of JP2 is its ability to losslessly compress images. Unlike the original JPEG standard, whose compression will lose data, JP2 images can be compressed without losing any image quality. Buckley explained JP2 compression takes into consideration the types of content in an image. If a JP2 image contains text, a color image, and a black and white image, each content type is recognized and compressed differently. Different layers in each of those content types are also supported.

During the course of Buckley’s presentation, he explained the structure of a JP2 file. The jp2 “box stack” contains six main components: jp2 signature, file type, header, codestream, XML metadata, UUID. The file type identifies the JP2 version and profile. The header contains the compression type and color palette. The codestream part of the box stack contains the image data. The XML metadata secton allows any type of XML metadata to be stored in the image.

Whether an XML metadata section is useful to the library community remains a question in my mind. We have complex metadata management systems and we do a good job of managing that information over time. Putting this information in the image itself would be like putting the metadata associated with a digital copy of a book into the copy. I can see it as useful for conveying user supplied metadata, before it has been cataloged, or as a place to put the most recent version of the metadata before sharing the image, but I don’t see a reason for moving image metadata out of our controlled, and managed, metadata environments.

Next, there were presentations by three JP2 vendors. While, of course, they would like to sell their products, the presentations didn’t have to feel of being sales pitches. The LuraTech vendor showed many different sites using JP2 in their digital projects (showing what could be done with the JP2 standard). The Aware representative demonstrated the ability to select regions of interest within an image and also showed examples from a site using the Aware product.

One key point that was reiterated by the vendors was that they would like input from the library community about the types of products that we want. If there are features that would be useful to us as a community, we should let them know. One person from the audience stated that he would like more open source products. It was mentioned that some JP2 open source products exist, but these are mostly just the encoding/decoding tools. For image servers, only commercial ones exist.

The last vendor was a new company, SandCodex. Their representative explained that their focus was on the presentation of JP2 images (they don’t deal with the encoding/decoding issues like the other vendors). He showed a very impressive demonstration of their product that allows one to zoom in from viewing a large collection of images into a particular region of interest in a single image. This product is not currently on the market though. It also, as I found out later on their website, requires IE (has an ActiveX dependency). A future Mac version is planned, but there is no mention of a version for Linux.

One thing I noted from the vendor presentations was that the most visually impressive demonstrations were using client side applications. The other alternative, also demonstrated by the vendors, is to provide the same functionality, but with less interactivity, in the browser itself (with no plugins). I would be curious to learn how/if more interactivity might be brought to online JP2 interfaces without requiring plugin dependencies.

The next part of the session was the panel of practitioners. Wangyal Shawa from Princeton University and Patrick McGlamery from the University of Connecticut showed projects that have been using JP2 in production. There was also a video presentation of a JP2/annotation protoype that looked very interesting. The protype, Vannotea, uses the annotea protocol from the W3C.

Wangyal Shawa, from Princeton, demonstrated layering GIS information over JP2 images to show, as an example, the Princeton area and its development over time. These images can be mapped onto larger GIS/JP2 images using embedded GIS data. Wangyal also explained the workflow involved with associating metadata with images, and moving both into production; he showed system architecture diagrams of his current JP2 system and diagrams of changes illustrating changes he would like to make.

Patrick McGlamery explained that they were in the process of moving several of their MrSid image projects into JP2. He talked about their Olson Collection project and their upcoming project, Connecticut History Online. Notably, UConn has also linked their image projects to their Voyager catalog so that links in the catalog take the patron directly to the image collection. There are also two other collections at UConn that are in the process of moving over to using JP2.

Overall, the session on JP2 was very technical but also very informative. It was interesting to hear what libraries are doing with JP2 and to learn about the advantages that JP2 offers over other image formats. There is a JP2 interest group for those in the library community interested in learning more. Also, as one of the vendors recommended, it would also be good for libraries to let the JP2 community know what we want from the standard. This, reportedly, can be done at the JP2 Info site.