I think RAW formats are interesting and their name misleading. For the most part they aren’t nearly as “uncooked” as people expect. It’s quite possible that some use a form of lossy compression; I wouldn’t be surprised. After all, some Nikon RAW digital imaging pipelines involve some filtering that many consider “destructive.”

The whole goal of lossy image compression is to be visually lossless while throwing away or changing data. I see the goal of RAW processing in a very similar way: to store data in a forward compatible way. Who knows what capabilities we’ll have in a decade which might look bad because of destructive choices in the current RAW pipeline?

I think if you can see blocky artifacts in the Canon RAW file — and if that bothers you — then that’s a reason to eschew JPEG lossy compression. If it doesn’t bother you and if we can count on camera manufacturers to use the highest quality settings, then perhaps lossy JPEG can slide by. But I suspect we’ll ultimately wish that we could go back and redo some of those RAW settings.

Who knows, I might be wrong.

I’ll admit that I haven’t actually looked at any JPEG-XR images. I’ve just looked at the file format. As a file format, it’s much better than classic JPEG. As a compression algorithm, DCT may in fact out-perform HD Photo’s PCT. There are several metrics people use for image quality, and no one agrees on which is best; just as there are many opinions on which compression mode “looks best” whatever the numbers say.

I do think that the JPEG file format needs to be “stabilized” and that we should move on to a new format that’s more forward-looking. Classic JPEG is like Ptolemy; we can keep adding features just like early cosmologists kept adding epicycles, but it’s going to make the system more and more difficult. JPEG-XR (or something formatted like it) is Copernican. Let’s move data back to the center of the system and keep attributes on the periphery.