One of the things digital photographers love to obsess over is our post-processing technique. RAW conversion, noise reduction, color adjustment, upsizing, sharpening, (especially sharpening!); everyone has their own preferred methods. One factor not frequently discussed is bit depth. Bit depth is the number of bits used to describe one pixel. Most computer images are 24-bit--they use 8 bits each for the red, green, and blue color values. This will give you slightly more than 16 million possible colors, which sounds like a lot, and it is. But many high-end cameras and scanners produce more than 8 bits per color channel, and Adobe Photoshop allows editing of images with 16 bits per color channel, a total of 48 bits. But before busting out with a "Way narly, dude!" and proclaiming this the coolest thing since the invention of sliced bread, consider this: twice as many bits per pixel means twice as much data to process when editing. This means twice as much data per image to process, load, and save, and twice the storage is required on disk to save the image. A 200 megabyte image magically becomes 400 megabytes. So what is the point of making everything twice as big and twice as slow? Simply this: All digital image editing is mathematical, To make a photo brighter, you either add to or multiply the pixel color values to make them larger, or do other mathematical operations. 8 bit images have a total of 256 possible integer values per color channel. When doing a color adjustment, what if the result of the equation is 129.725? The computer will round off the answer to 130, since that is the closest whole number. When doing one adjustment to an image, this rounding error is insignificant, but when doing many adjustments and changes to an image, these rounding errors can add up. Or at least that is the way the theory goes. 16 bit mode is still in the realm of integers, but now the allowable values run from 0-65535 instead of 0-255. This is a significant difference, but is this really worth bothering about? Do real-world results exhibit any differences?

Noel Carboni, a regular contributor in the Rob Galbraith forums, recently started a discussion thread suggesting that the benefits of 16-bit editing were not outweighed by the extra overhead imposed by 16-bit editing; that anything doable in 16-bit mode could also be done well in 8-bit mode, if one does it properly. He took a lot of heat for his position (some of it bordering on ad hominem), but Noel has demonstrated on numerous occasions that he can take a photograph and process it using his preferred 8-bit editing methods, and produce excellent results. So his position is obviously not the ravings of an ignorant crackpot; he has some real-world results to back it up. I must admit that I'm a believer in 16-bit editing, and have gone to great lengths to construct my workflow so that the images I work on stay in 16-bit mode throughout the editing process. So anyway the end result is that Noel and I decided to do a head-to-head comparison or contest of sorts; each of us starting with the same RAW image, and editing it via our preferred methods, to see who could produce a superior result. Noel proposed an image of his lovely wife, and I accepted. The image has some characteristics that make it challenging to interpret well:

1. It has some blown highlights in the background.
2. The shadow areas have some noise that has to be dealt with.
3. The dynamic range is pretty wide, keeping detail throughout the image and maintaining a natural look without accentuating the noise is difficult.
4. The lighting is a combination of incandescent and sunlight, making a good white balance difficult.

So anyway, here is a side-by-side comparison of our efforts. Look at the images first, and see if you can decide who did what. The answer is at the bottom of the page...

And while I'm thinking about it, this image is the intellectual property of Noel Carboni, used with permission, yada yada yada. If you want to put it on the cover of your glamour magazine, talk to him first.