In my last column, which delved into a couple of useful sharpening techniques, I promised additional techniques this time around, and I’ll make good on that promise. But first we should turn to one of the important questions about sharpening: When in the image-editing process should you sharpen?
The Traditional View
Conventional wisdom has it that you should make sharpening the last operation before converting to your output color space, but there’s a problem with this approach. Sharpening works by increasing local contrast along edges. Typically, after sharpening some pixels that were almost white become pure white, and some pixels that were almost black become pure black. The black ones aren’t too much of a problem, but the white ones can be: Depending on your output device and how you’ve sharpened, sharpening can make light pixels appear whiter than they should, so that edges or diffuse highlights look artificial.
The problem is that almost all our printing devices have a minimum amount of color that they can lay down, and depending on the process we’re printing to, that minimum amount may be quite different from pure paper white. An extreme case would be newsprint, where the minimum printable dot is typically in the 8-to-10-percent range. The “transition zone” — the zone between the smallest printable dot and paper white — is huge.
When we print images, we want to use the smallest printable dot for our diffuse highlights — highlights that still contain actual detail — and reserve paper white for small specular highlights such as the glints on metal or glass, which lack detail. If instead we drive our diffuse highlights into the transition zone, as in Figure 1, we typically produce an ugly result. Note the severe case of “highlight psoriasis” in Figure 1 on the model’s nose in the image at the right. In the image on the left, we’ve maintained a dot in the detailed highlights, reserving paper white for the small specular highlights on the necklace.
Figure 1: In the image on the right, diffuse highlights are too white, leading to an artificial look.
Controlling highlight behavior in the transition zone is critical to producing good images, and is a topic deserving attention in its own right. But it’s also one of the things we need to keep in mind when we sharpen images.
Injudicious sharpening may not produce results quite as ghastly as shown in Figure 1, but it can result in random white speckles along high-contrast edges when you do it as the last operation before converting to output space. Hence my recommendation is that you sharpen before setting your highlight and shadow dots, which should probably be the last thing you do before converting to output space.
Of course, both shadow/highlight dot setting and sharpening for output are dependent on the specific output process — the paper, the inks, and the mechanism used to apply one to the other. If you’re the type who prints the same image to many different kinds of output (even if the difference is only between matte and glossy papers on an inkjet printer), you may find sharpening from scratch each time you make a print a fairly tedious process. The sharpening technique I present below can help reduce the drudgery, partly by splitting the sharpening process into two distinct passes.