alan little's weblog

clever photo software

12th May 2004 permanent link

This has that astonishing but at the same time obvious (once somebody else has thought of it) quality that really clever ideas have. French company DO Labs has two new software products out for digital photographers. The first one, DxO Analyzer, is less obviously interesting – basically, using shots of a standard test chart made under controlled conditions, and some fancy maths, it produces a profile of the optical imperfections in a camera/lens combination. Interesting for lens makers, camera reviewers, and people who worry about equipment and technicalities rather than how to compose interesting pictures.

However, DxO Analyzer makes possible the bit that actually is somewhat interesting. The idea behind DxO Optics Pro is conceptually straightforward. DxO Analyzer gives you a profile of the optical imperfections of your lens – in other words, a set of transformations between the image that was actually in front of the lens and the one that got recorded on the digital sensor. And these transformations are presumably reversible, so you reverse them and voilà! Optically perfect images from any old cheapo garbage lens.

Of course it’s not actually 100% effective in real life, and makers of expensive high end lenses probably don’t need to start losing sleep over it just yet. But Michael Reichmann of Luminous Landscape is impressed:

How Good Is It?

In a word – remarkable. I know that some readers who aren’t familiar with the fact that I’m often very critical of products, and am not afraid to say so in print, will think that I’m being hyperbolic. But I’m not. This product is revolutionary. It can’t turn a bad lens into a good lens, but it can make a mediocre lens a lot better, and a good lens can be raised up to being terrific.

… I’ll leave off here by saying that what I see is a dramatic improvement in several aspects of image quality, and no visible negative affects at all. Improvements are most noticeable when using lower quality lenses. But even with very fine (and expensive) lenses improvements are readily visible.

Looking at Michael’s test images, I’m impressed too. He shows an example of a picture shot with one of Canon’s best professional lenses, and even with this lens the version processed with DxO Optics Pro is clearly significantly better than the “before” version.

The pricing is cheaper than buying really good lenses. For example, for the new Nikon D70 and a single lens, such as the perfectly ok 18-70 zoom that comes with the camera, DxO Optics Pro would come in at about $127. A significantly better lens than the 18-70 would cost at least four or five times that much.

This only works for digital cameras. It might be theoretically possible to attempt it for film, but there are so many more variables in film processing that the results would probably be meaningless. Scanned film introduces a whole new set of variables, notably film flatness and thickness, and the fact that the film scanner is a specialised digital camera in its own right. So you would have to profile for every possible lens-camera-scanner permutation, and even then the fact that your film isn’t sitting exactly the same in the camera & scanner every time would probably make the whole exercise futile anyway.

Just to reiterate the key message: I find this quite interesting in that “oh look, somebody’s done something stunningly clever that looks deceptively simple after the fact” way. But even if DxO Optics were perfect it still wouldn’t solve the only photographic problem that’s actually interesting and important: how to compose artistically interesting images. Technical limitations of cameras and lenses are not remotely close to being the main limiting factor in most people’s photography, and you can’t make a boring composition interesting by taking a technically perfect picture of it.

related entries: Photography

all text and images © 2003–2008