During the first day of my first college economics course, the professor presented us with the fundamental underlying assumption of economic theory: More is always better. "Is it?" he challenged us. Well, one moldy master's degree in economics and an entirely unrelated career later, I find myself returning to that question, albeit in a comfier chair and with a slightly different tack: "In digital imaging, is more always better?" As the resolution of cameras, scanners, and printers rise and increasing numbers of colors fill our digital palettes, where do we draw the line between what is necessary and what's sufficient?
Unfortunately, the debate over digital-imaging specsmanship rarely gets broken down into its three constituent parts: How much data you need for capture, how much for editing/retouching, and how much for output. The answer is different for each stage of your workflow and needs to be calculated backward, from output to input.
A bit about bit depth
Edict: It's always best to work with 48-bit color, but 24-bit color is frequently enough.
The decision about color depth predates your camera purchase. The only way to capture 48-bit data is to shoot and save as a RAW file, which means you need to determine whether you need a camera capable of saving RAW files before you buy. JPEG images are always saved in a 24-bit space, though they frequently don't provide a full 8 bits per pixel of data.
Why does it matter? In my last column, I discussed how to adjust a photo's exposure using Levels and Curves. What I didn't show was the effect the operation had on the image data.
Histogram before (left) and after (right). The image looks a lot better and retains the same number of pixels, but the operation has spread the existing lightness values too thinly, leaving gaps between levels.
In the previous example, I started with a 24-bit image, which defines colors using 256 levels of red, green, and blue values. If you do the math--see Bruce Lindbloom's excellent site for info on the math behind color science--it turns out that when you start with 24 bits and perform operations equivalent to converting the image to a gamma of 2.2 (adjusting the exposure has basically the same effect), you end up dropping from 256 levels per channel to 184. That's equivalent to about 22.6-bit color. By the same reasoning, it takes a starting file with 16 bits per channel (48-bit color) to end up with 24-bit color after the operation. Images that have been compressed into JPEG format are generally left with 4 or fewer bits per pixel.
Though the math may be indisputable, the question of whether you really need to start out with a 16-bit file falls into a gray area. For one, the artifact caused by a drop in bit depth--the appearance of visible transitions between colors referred to as contours--occur in the deepest shadows, where you probably can't see them. And even if contouring was showing up in the midtones, you still might not be able to spot it; after all, dropping from 8 to 7.5 bits per color (184 levels) doesn't make that much of a dent.
10 ways to keep problems at bay
Once you've decided whether to shoot RAW (for 48-bit color) or 24-bit, the rest comes down to careful shooting. It's best to avoid the I-can-always-fix-it-in-Photoshop trap. Even if you do start with a 48-bit file, aggressive edits may still cause problems. Sometimes artifacts caused by editing are immediately visible, but most of the time, they become visible only after a few operations. The less retouching you have to do, the less image degradation you'll incur, and the smaller the initial image file you'll need--that translates to a lower-resolution, less expensive digital camera. Here are some precautions you can take to minimize your postprocessing woes.
- If you plan to edit your photos, slightly underexpose your shots.
- In uncontrolled lighting situations--pretty much anywhere outside of a studio--include a white card or object in the scene (it can be at the very edge of the frame for easy cropping) to make color correction easier.
- If it has one, use your camera's histogram display to ensure proper exposure.
- For tough exposures, take two shots--one exposed for the highlights, one for the shadows--that you might be able to combine later.
- If you intend to crop deeply, make sure your subject will reproduce large enough. Here's a handy calculator to figure out the minimum detectable object size for a given lens and CCD.
- You'll be much happier if you remember to turn on your camera's red-eye reduction than if you have to get rid of it in software.
- Unless you're planning to print directly from the card or camera, don't bump up any of the image adjustment settings, such as sharpness, contrast, color saturation, and so on. You'll have more control in software, and those processes can degrade the image data.
- Frame it right the first time. Yes, you can always crop, but cropping decreases image resolution.
- When you have the shot you want, bring it up on the LCD, and zoom in to make sure it's sharp; if not, shoot it again. You can't make a blurred image sharp, no matter how much processing power you throw at it.
- Make sure your white-balance setting is appropriate for the shot. I frequently get tripped up by this: I shot a 3-year-old's party in a disco (don't ask) and kept alternating between flash and nonflash shots, using a variety of white-balance settings. I'd forget which one was active until I looked at the LCD and realized everything was deep red. Let's just say some of those kids came out mighty scary-looking, and no amount of color correction could get the red out.
Of course, we can't shoot it right the first time, every time. So your next best defense is to understand how different adjustments affect the quality of your image. But that's a subject for another day.
Next up: much ado about resolution.