dante stella stories photographs technical guestbook

Looking backward into the future...
The Failure of Firmware

Are people being overrun by post-processing issues because digital SLR firmware is short of what it needs to be to deliver usable proof JPEGs? Is the next revolution in in-camera image processing? My suspicion is that the proliferation of batch RAW processing software is a response in part to the lack of practical JPEG generation in-camera.

Having come from a Kodak/Nikon FF background, I was a little bit surprised at how much more time-consuming it is to generate good, printable JPEGs (the kind you would make 4x6 proofs from) using Nikon's current D2x firmware (and I imagine the situation is the same with other Nikon and Canon pro bodies). There is little that you cannot do, given enough time either in setting up the shot, or processing it. But the efficiency is, in a word, low. You can shoot quicker, but unless you really take your time to set things up, you end up paying for it on the back end. Nikon's in-camera controls are, in a word, basic. Downloading a custom curve is ok, but you have to be able to predict ahead of time what curve you will need.

The missing piece seems to be dynamic curve processing. Kodak, whose first digital SLRs emerged before LCDs and on-camera histograms were invented, approached the lack of feedback in two ways. First, it emphasized extreme DR in its chips (the 14n, for example, does 11.5 stops at 80 ISO). Second, and more importantly for this discussion, it developed a system in which the camera automatically recovers highlights and changes the curves on a shot by shot basis to make sure that what you output (in-camera) to JPEG was at least within bounds - and a taste of what you would get if you took more time in Photodesk to perfect it (these settings would be saved into the DCR metadata, too). This is an excellent shortcut if you know that most of the pictures will remain proofs but you still want the proofs to look good without recorrecting each one. This process, which they call Digital Image Correction, has a very high hit rate, and you can basically plug your CF into a Frontier, perform no further corrections, and get pictures that look good in a book of proofs about 98% of the time. It does make in camera processing slower - but not nearly as slow as getting an underexposed JPEG and having to correct it later.

A lot of people seem fixated on numbers - megapixels, focusing zones, fps, DR (at least as much as they understand it), and the like - but where is the clamor to make the image processing more convenient? Someone should get on digital camera manufacturers to license this Kodak technology, which seems not unlike Photoshop's Auto Levels for a camera.