Pixel Perfect : The digital studio demystified
Digital photography's raw deal
By Lori Grunin
Senior editor, CNET Reviews
June 3, 2005
I've been toddling along, discussing methods for organizing and otherwise herding the various media types--both digital and analog--that bedevil us. And whaddaya know: one of those devilish little buggers has found itself in the center of a storm of controversy. OK, maybe not a storm, but certainly a noisy cocktail party. Who ever thought that people would get worked up over a file format? Raw files have certainly become a critical staple of digital photography, but I suspect the more far-reaching issue is user frustration over the endless hassle of keeping up with the rapid changes in archiving technologies. But I'll get to that later. First, the boring facts.
Our story begins in the gruesome innards of the OS file system. An image file bit stream, the actual binary encoding, is generally composed of several parts. The two key parts that concern us are the raster data, the pixel coordinates and color values that define a picture; and the metadata, which can be as simple as a version number for the format or as complex as a history list of transformations plus the settings of the camera when you shot the picture or essential parameters for decompressing a compressed bit stream. A third component of the file is the header, which provides the info necessary for the file system to recognize the appropriate applications for opening the file. For formats with complex metadata, correctly rendering the image for the screen, a printer, or other output device requires complex software that understands how to read and apply the metadata.
Raw files have very complex metadata, because the raster data in them is so...well, raw. It's the image straight off the sensor, without any noise reduction, sharpening, color-space transformations, gamma correction, and so on. You likely wouldn't recognize the raster as even remotely resembling the photograph you shot. But its utter formlessness is why we like it so much: instead of applying all these transformations in the camera, with its relatively underpowered hardware and constrained algorithms, you can use a powerful computer with sophisticated algorithms to work interactively and nondestructively.
Because the raw data and the types of voodoo the camera manufacturer performs on it reveals important information about the inner workings of the camera (which used to be called trade secrets but now travel with lawyers as intellectual property), raw-file formats have traditionally been proprietary. As a result, you generally had to use whatever utility the vendor supplied, no matter how agonizingly slow or obtuse, to work with your images. More recently, manufacturers have begun to supply software developers with an API (application programming interface) that allows them to use the camera manufacturers' algorithms without ever seeing what's going on in the raw black box but providing the user with a standard interface for working with all the different formats. Of course, proprietary is relative: many third-party developers happily reverse engineer the file formats and build their own tools.
DNG, dong, Adobe calling
As with proprietary hardware, people worry that one day their software and the associated files will become inaccessible. It's not a critical worry when you're talking about a paper you wrote in college 20 years ago, but when it's photographs--either your memories or your livelihood--you get a bit more emotional. As a potential solution, Adobe introduced the Digital Negative (DNG) format, a raw enhancement that separates the metadata into two types: common and proprietary. Thus, software that supports the DNG specification can open and render any manufacturer's raw file, but the manufacturer can pass through any custom info it wants for use by its own raw-processing utility. And because DNG is an extension of the latest version of the TIFF format--one of the longest-lived, most documented bitmap formats around--it theoretically shouldn't take much effort to implement. TIFF was the obvious choice for one other reason: it's the only standard file format with support for color depths beyond 24-bit, which is essential for raw and high dynamic range imaging. Adobe topped off the attraction by guaranteeing that there would always be software available to read DNG files.
Microsoft's plans for supporting raw files within Windows conveniently dovetails with Adobe's DNG strategy. The next generation of Windows' imaging engine, code-named Avalon, is designed to work with a file format that looks eerily similar to DNG; instead of ignoring the proprietary metadata, however, Avalon will check the registry and see if there's a format-specific application on your system (which Microsoft confusingly refers to as a codec). If not, it will try to download one from the appropriate Web site. Unfortunately, though, the next version of Windows, dubbed Longhorn, isn't due until next year--and it's essential, because the current version of Windows doesn't natively support the necessary 32-bit color depth. Microsoft also has a tendency to drop proposed features en masse as it tries to get a new Windows off the ground, though I don't think Avalon will be one of the casualties. As an interim solution, Microsoft will be providing an update, the Microsoft Raw Image Thumbnailer and Viewer for Windows XP, which will allow users to view Nikon and Canon image files as thumbnails in Explorer, as well as preview and print the full images. (I haven't yet researched Apple's OS plans, but I can't help but believe that a closed system deserves proprietary file formats.)
Canon and Nikon rub photographers raw
So everything was going swimmingly until March 2005, when the bits hit the fan. Canon released the new version of its raw software, without legacy support for the D30 dSLR's files. Then, rumors circulated--which Nikon substantiated--that the NEF raw files produced by the D2X dSLR had encrypted white-balance information. Suddenly, the photographic community felt threatened: what if all manufacturers dropped support for older versions of their raw formats? What if we have to give up our favorite third-party raw-processing apps, such as Bibble, Qimage, and Capture One, because they don't license the software developer's kits and can't read that encrypted information?
On the Web, outrage forms communities. This particular outrage resulted in a grassroots movement with an elegantly designed Web site, OpenRAW, devoted to promoting the cause of requiring that manufacturers publicly document their raw formats.
Excuse me a moment while I implode.
Why? First, because the whole brouhaha distracts from a far more important issue that affects a far greater number of people: it doesn't matter if you can't read the file, because by the time the format becomes obsolete, you will probably have lost the ability to read the media on which it's written. Where's an OpenMedia site? As I migrate from platform to platform--Bernoulli, Jaz, Zip, CD-R, DVD-R, DVD+R--I can almost feel myself unwittingly shedding files like a trail of breadcrumbs. I suppose the good news is that I transfer archives so frequently that I never get to find out how long any particular media will really last.
Second, I find many of the arguments specious. The most commonly voiced of these is the property rights argument: "I own the photograph, and I should be able to do whatever I want with it. I can't do that if I can't use the software I want." Now, I'm no lawyer, but from what I know about technology and what I've gleaned about copyright law as a writer and an editor, they've just shot themselves in the ass. According to the U.S. Copyright Office, "Works that have not been fixed in a tangible form of expression..." are not copyrightable. Though you've seen the picture in your head and saved the photo to a flash card, a raw file is not fixed--ironically, that's the reason we love it--until you've opened it and the software has applied all parameters specified by the metadata.
A lot of people also complain about how difficult proprietary raw files make their work flow. I agree that there are tons of inconveniences with having to use several different applications to work with an image file and that distributing the photos requires converting them to another format first. But I think--OK, I hope--that OS support as outlined above will alleviate most of the problems. As for the rest, well, we're in good company: every day, presenters worldwide struggle with the hideous pain of getting content from a PDF file into a PowerPoint presentation. I haven't heard any demands for an OpenPDF or an OpenPowerPoint format. (Surprisingly, I haven't heard of any third-party tools either. Have you? I'd love to know--drop me a line.)
The fact that Nikon charges customers who drop $5,000 on a digital camera another $100 for the raw-conversion software--that's just disgusting. Maybe it's illegal, too, through some byzantine interpretation of the antitrust laws (as an offense known as tying, something I have studied).
I, too, have gazillions of raw files. And all of the above explains why I recommend that, when you archive digital photos, convert them to TIFF. Keep the raw, but count on TIFF. Now if I could just figure out what media to use for archiving...