JPEG XR Could Be Neat, but JPEG2000 is Still Neater

5 minute read

× This article was imported from this blog's previous content management system (WordPress), and may have errors in formatting and functionality. If you find these errors are a significant barrier to understanding the article, please let me know.

On Tuesday, the Joint Photographic Expert's Group (a.k.a. "JPEG") announced a new work item for the standardization of Microsoft's HD Photo as JPEG XR (XR is short for "extended range" -- a reference to its improvement over the original JPEG standard). You can read the publicity details in the Microsoft press release and the JPEG press release, but beyond the public relations pieces I wonder if you are thinking about HD-Photo/JPEG-XR for digital archiving. And if you're thinking that I'll bet your wondering about how HD Photo compares with JPEG 2000. As with many things, the devil is in the details, so here is a first, gut-reaction pass at the details.

Standards Free of Intellectual Property Concerns

In terms of IP concerns, both formats seem to be on par. Microsoft's press release promises "if approved, Microsoft will offer a royalty-free grant for its patents that are required to implement the standard." JPEG 2000 Part 1, developed as a committee work item, is believed to be free of IP restrictions. (More specifically, the official JPEG2000 website says, "In the case of technology submitted for consideration by JPEG for incorporation in Part 1 of the standard, the JPEG committee believes that the individual organisations concerned will make available licences to use this intellectual property, on a royalty- and fee-free basis, under specified conditions which may apply only to conforming implementations of the standard.")

HD Photo's Technical Advantage?

There are claims that HD Photo's compression algorithm is faster and has a smaller memory footprint, yet yields compression results on par with JPEG 2000. The most official statement from Microsoft that I could find was this: "HD Photo delivers a lightweight, high performance algorithm with a small memory footprint that enables practical, in-device encoding and decoding. HD Photo delivers image quality that is comparable to JPEG-2000 and more than twice the quality of JPEG." ((HD Photo Specification Download. Version 1.0 dated November 16, 2006. Date Visited: Fri Aug 03 2007 16:45:27 GMT-0400 (EDT) )) The first statement is about performance that doesn't directly compare HD Photo to other standards; the second statement is a one of quality equivalence with JPEG2000. Taken together, that is not exactly a clear-cut endorsement of the more-efficient-than-while-having-the-same-quality-as-JPEG2000 claim, and clearly we need to see some published metrics support or refute this. (Know of any published studies or statements that address this directly? Tell me about them in the comments.)

This seems to be the only claim of superiority to the JPEG2000 standard. Perhaps it is telling that HD Photo lives in the Windows Hardware Developers Central area of the Microsoft website -- it is a standard intended for embedded devices (cameras, etc.). Even if it were true that JPEG2000 is more computationally intensive for compression and decompression, at the rate of general processor speed improvements is this enough to justify a new standard? Perhaps it makes sense for today's marketplace of digital cameras; I don't think this justification holds much weight in our community.

But What About the Metadata?

Those of us in the cultural heritage community harp on this point: it is not all about the image itself; its context is just as important, if not more so. That context is provided by rich metadata support. The big advantage to the JPEG2000 file format is the ability to include arbitrary things in "boxes" within the JPEG2000 file. Those boxes can contain XML-based metadata of any schema, audio annotations, PDF renditions, whatever. The HD Photo file format specification appears to only offer a dramatically reduced (albeit important subset) of metadata choices: ICC color profile, Adobe's (proprietary) XMP Metadata, (de facto standard) EXIF Metadata, and TIFF-compatible Descriptive Metadata Tags.

And Other JPEG2000 Advantages?

A JPEG2000 practice means fewer derivative files: an archival and delivery system based on JPEG2000 doesn't need thumbnail and web presentation versions since those can be derived from the master file on-the-fly. HD Photo offers this same sort of progressive resolution decoding that enables on-the-fly creation of derivatives to happen. (It has not been as rigorously tested as JPEG2000 has, however; something we should expect to do before we adopt it in our own practice.) What HD Photo doesn't offer, though, is the use of the same image compression scheme and basic file format for both still images and moving images.

As the saying goes, "Never Put All of Your Eggs in One Basket." Digital archival formats is one exception to that rule: simplicity -- fewer baskets -- is important. Fewer baskets -- fewer file formats -- means less to keep track of over time. Part 3 of JPEG2000 family of standards is called "Motion JPEG2000," and as the JPEG committee press release describes, the "JPEG 2000 [Part 3] based DCI specification 1.1 [will be used] for distribution of digital movies to theatres/cinemas worldwide." A file format used by the motion picture industry, with its deep pockets, for the production and distribution of movies is a standard that has a greater chance of standing the test of time. So a JPEG2000 practice -- our one "basket" -- can preserve both our still images and our moving images file formats. HD Photo does not have a corresponding moving image format.

What is HD Photo's Compelling Advantage?

This is where the rubber meets the road. Our community has used a TIFF-based practice for preserving still images for decades, and I've been saying that it would take compelling reasons to change our practice. As posted here before, I believe there are compelling advantages in JPEG2000 that are big enough to cause us to change our practices. At this point, I don't see those same compelling advantages in HD Photo, whether or not it earns the "JPEG XR" moniker.

Truth be told, this analysis of HD Photo versus JPEG2000 isn't based on any empirical evidence -- it is just what my gut is telling me based what I've read so far. What we do for the long term preservation of cultural heritage materials shouldn't be based on anecdotal analysis, though. Gut reactions and/or evidence from others are welcome.