February 26, 2007
Non-destructive JPEG: An oxymoron?
When cameras capable of shooting digital raw files started hitting the mainstream (roughly five years ago, give or take), one of the advantages of shooting raw was that editing had to be non-destructive. That is, because the pixel data hadn’t yet been converted into traditional RGB channel data, applications like Photoshop couldn’t poke at it directly. This in turn meant that conversion parameters had to be stored as sets of instructions, rather than as burned-in as pixel edits.
Photographers have now become familiar and comfortable with the idea of moving & storing the captured bits along with the "special sauce" used by their raw processing app of choice. The XMP files that are (optionally) parked next to images by Adobe Camera Raw & Lightroom make this particularly easy. The fact that the DNG format supports built-in metadata & rendered previews turns it into a kind of envelope (or "job jacket," to borrow Peter Krogh’s phrase)–a container that stores your negative, your processing instructions, and your rendered print. As editing tools get richer–for example, with Lightroom’s ability to store multiple settings per file–the benefits of this approach grow.
But what about non-raw files? Both Lightroom and Camera Raw now offer the ability to edit JPEG and TIFF files, so that no matter what format(s) your camera generates, you can use the same non-destructive tools. So now a photojournalist or sports shooter, say, could shoot JPEGs, apply edits in the field (soft crops, non-destructive dust busting, tonal corrections, etc.), and upload the original files plus their processing instructions.
This poses some tricky questions, however. Fundamentally, is it okay that Adobe is putting "special sauce" into the metadata of JPEGs, causing them to appear differently when viewed in the latest Adobe editing tools than in other apps? Is it okay to extend the JPEG standard? A few things to consider:
- Adding this metadata to JPEGs doesn’t damage the files in any way, or degrade other tools’ ability to read the pixels. The data is simply ignored by tools other than Lightroom/ACR/Photoshop/Bridge. Adobe tools are leveraging the flexibility that’s already in the format.
- Generating a copy of the image with the edits burned in (i.e. with the pixels changed) is a one-click task.
- Putting the metadata into the files makes it more easily portable than requiring a sidecar file.
- One alternative would be to bake Lightroom/ACR edits into JPEGs immediately, thereby negating the advantage of non-destructiveness. Another would be to force the JPEG to be converted to another format, making it clear that something had changed, but rendering those images unreadable by other tools. Forcing either approach, however, seemed like a bad idea.
So, there are pros and cons to any approach, but the one we’re pursuing makes it possible to enjoy the portability and non-destructiveness of raw editing using non-raw files. It’s done in a way that lets JPEGs be extended easily & without damage. If you’re concerned about using this approach, you can convert JPEG & TIFF files to DNG (an option I’ll address separately in a bit)–but that conversion isn’t forced on anyone.
My take is that the flexibility it opens up is more than worth the cost. What do you think?