SharpeningWhy, How, and When?

If we got a nickel for each time we've been asked what sharpening settings someone should use, we'd probably be sipping wine on a tropical island rather than writing books on Photoshop. Of course, we couldn't really come up with a short answer any less lame than "it depends."

The longer answer, of course, involves explaining the various factors on which it depends, which we'll now proceed to do. First, a necessary disclaimer: Bruce and his colleagues at PixelGenius publish a sharpening plug-in, called PhotoKit SHARPENER. While we'd be pleased as punch if you considered purchasing PhotoKit SHARPENER, you should know that a lot of what it does is based on the information in this chapter. The plug-in just does it a lot faster than you can do it manually.

Why We Sharpen

Wethat's a collective "we," which we hope includes you, toosharpen images for several reasons, including the inevitable loss of sharpness from the capturing process. Every reason to sharpen imposes its own demands. Fairly often, though, these demands contradict one another.

Sharpening the capture

Whenever you turn photons into pixels, you lose some sharpness, because no matter how high the resolution of your capture devices, they sample a fixed grid of pixels, turning the continuous gradations of tone and color that exist in the real world into discrete pixels. Each capture device imposes its own noise pattern on the image, whether it's from film grain, from digital noise, or, as is common in many film scans, from a combination of the two.

You need to sharpen the image content to restore what was lost in the conversion to pixels, but you don't want to also sharpenand hence emphasizethe noise and grain. So effective sharpening must take into account the source of the image.

Sharpening the image

People often sharpen for creative reasons: to tell a story, to make a point, to emphasize an area of interest, or to sell a product. To do this successfully, you need to match your sharpening to the content of the image. A busy, high-frequency image with lots of tiny details, like a forest full of trees, has much narrower edges than a close subject with soft detail like a head shot or pumpkin (see Figure 9-7).

Figure 9-7. Unsharp Mask settings for high- and low-frequency images

When sharpening image content, you want to emphasize the edges without overemphasizing textureslike skin tonesand without introducing spurious texture into flat areas like skies. So effective sharpening must take into account the image content.

Sharpening the output

When you turn pixels into marks on a substratein other words, when you printyou lose sharpness again. In most cases, individual pixels aren't translated into individual dots of ink or dye, and even in those cases where they are (such as photographic printers like the Durst Lambda or Fuji Pictrography) the printed "pixels" tend to be round rather than square. In either case, the output loses some sharpness. Because you want to make the print as sharp as the output device can render it, effective sharpening must take into account the output process.

Image Detail and Sharpening Radius

You can achieve the same apparent sharpness with many different combinations of Amount, Radius, and Threshold settings, but the difference between good and bad sharpening lies largely in matching the Radius setting to the image content.

Look closely at the image at hand. How big, in pixels, are the details that you want to sharpen? You need to match the size and intensity of the sharpening halo to the size of the details in the image.

High-frequency images contain a lot of detail, with sharp transitions between tonal values, while low-frequency images have smoother transitions and fewer small details. Whether a given image is high frequency or low frequency depends on the content of the image and on its pixel density. High-frequency images, where the edges of objects are reproduced using only one or two pixels, need a smaller Radius setting than low-frequency images, where the edges may be a dozen or so pixels wide.

An image containing fine detail, such as a picture of trees, is likely to have many more high-frequency transitions than a head shot, for example. But if you scan the trees at a high enough resolution, even the edges on the tiniest leaves will be reproduced several pixels wide in the scan. So it isn't just the content that dictates the sharpening, it's the relationship between content and resolution.

Unpleasant settings. Too large a Radius is the prime cause of oversharpened images. Moreover, an overly large Radius can actually wipe out the detail it's supposed to be accentuating. Too small a Radius can result in too little apparent sharpening. This might in turn seduce you into cranking up the Amount setting so far that you create spurious specular highlights, and overemphasize textures such as skin in undesirable ways. With very extreme settings, you can change the overall image contrastwhich in most cases isn't what you want.

Figure 9-7 shows two images that need quite different sharpening settings. The trees image contains a lot of fine detail that needs a low Radius setting and a fairly high Amount setting to bring it out. If we apply the same sharpening to the pumpkin, it fails to bring out the necessary detail (while threatening to create unpleasant mottling).

Conversely, sharpening settings that work well on the pumpkin don't work at all well on the trees. The larger Radius sharpens the larger elements well, but the more delicate elements are lost. It creates a very confused appearance, where the same element in the image appears sharp in some places and soft in others.

If you get the Radius correct first, it's easy to set the Amount to achieve the degree of sharpness you want. Then you can adjust the Threshold to suppress noise, and to avoid oversharpening patterns, film grain, and the like.

When We Sharpen

Needless to say (but we'll say it anyway), the chances are exceedingly slim that you can satisfy all these different sharpening criteria with a single round of Unsharp Mask, applied globally to the image. Of course, if one pass of Unsharp Masking is all you have time for in your workflow, it's better than not sharpening at all. But for the highest quality, we recommend using a two- or three-stage approach to sharpening. Rather than try to satisfy all the criteria simultaneously, this workflow approach to sharpening addresses them separately.

First, apply a gentle round of sharpening very soon after image capture to take care of the source-sensitive aspects. After you make your major tone and color corrections, apply some localized creative sharpening (we skip this step in automated workflows). Then, once the image is at final output resolution, apply sharpening tailored to the chosen output.

At the risk of drawing an analogy from one incomprehensible subject to another, we liken the sharpening workflow to the color management workflow:

• Capture sharpening is like converting from the source profile to the working spaceit compensates for the quirks of the source and puts the image in a good state for editing.

• Creative sharpening is like doing color correction, using creative skills to make the image do what you want it to.

• Output sharpening is like converting to an output profile, creating a device-specific version of the image that is designed to work only for the designated output process.

Essentially, the capture and creative sharpens make a file that is repurposable (our mothers always taught us to keep our options open as long as possible), and responds well to resizing and final output sharpening.

Of course, if approached carelessly, this workflow can create some very ugly images. Just hitting the image with three rounds of Unsharp Mask using different radii is a recipe for certain disaster. Instead, retaining optimum quality requires finesse and some fairly advanced sharpening techniques. We look at individual sharpening techniques in detail later in this chapter, but here's the 30,000-foot overview.

Capture sharpening

The first round of sharpening in the workflow must be done very gently indeed; otherwise the result is likely to be a hideously oversharpened mess. It's often helpful to sharpen through an edge maskso that only the high-contrast edges get sharpenedand to focus the sharpening on the midtones, protecting highlights and shadows so that they don't get driven to solid black and solid white.

With very grainy or noisy originals such as high-ISO digital capture or fast color negative, you may first want to apply some noise reduction (essentially unsharpening) using the Reduce Noise filter or a third-party noise reduction plug-in.

Creative sharpening

For creative sharpening, you can build "sharpening brushes" to paint your sharpening just where you want it. As you'll see, you do this by creating a new merged layer, setting the layer's blending mode to Luminosity (to avoid color-fringing), applying a global Unsharp Mask to the layer, then adding a layer mask set to Hide All. As you paint on the layer mask, Photoshop adds or removes the sharpening.

Creative sharpening effects are really only limited by your imagination. For example, one way to make an object appear sharper is to blur its surroundingsyou can create "smoothing brushes" using the same techniques as sharpening brushes, but substituting a blur for the sharpen.

Output sharpening

Since the image-specific and source-specific concerns were already addressed in the capture and creative sharpening phases, output sharpening can concentrate solely on the output process.

Note that you can only sharpen the image's pixelsPhotoshop has no control over how those pixels are rendered to ink on paper (or any other output process). So the key factor in output sharpening is the relationship between the pixels and the resulting hard copy. Hence output sharpening must be done at the final size and resolution, often as the last step before converting an RGB file to CMYK and saving the file to disk. Note that unlike capture and creative sharpening, output sharpening is something we always apply globally. Here are a few other considerations:

• A rule of thumb that has served us well is to aim for a sharpening halo of approximately to -inch (.5 to .25 mm) in width, the thinking being that at normal viewing distances, a halo this size falls below the threshold of human visual acuity, so you don't see the halo as a separate featureyou just get the illusion of sharpness that it produces.

  A good starting point for the Unsharp Mask filter's Radius setting is image resolution ÷ 200. (Remember: we're talking about final image resolution, after it has been placed on a page and scaled to fit.) Thus, for a 300-ppi image, you'd use a Radius of 1.5 (300 ÷ 200). For a 200-ppi image, you'd use a Radius setting of 1. This is a suggested starting point, not a golden rule. As you gain experience, you'll find situations where the rule has to be bent. When sharpening using methods that don't involve Unsharp Mask, you'll have to look closely and do some math yourself.

• Of course, there's really no way to get an accurate on-screen representation of how a sharpened halftone output will lookthe continuous-tone monitor display is simply too different from the halftone. An image well-sharpened for halftone output will typically look "crunchy" on screen. And the monitor resolution and view percentage can help or hinder your appraisal of an image's sharpness. See the sidebar "Sharpening and the Display" for more on the subject.

  Sharpening and the Display

  Back in the days when all our monitors were CRTs, we thought we had a good idea of how to judge sharpness from the screen. But the vast differences in apparent sharpness between LCD and CRT monitors, and the research we've undertaken in applying output sharpening, have caused us to reevaluate that position.

  LCD monitors are much, much sharper than CRTs at any given display resolution. Moreover, an image will appear quite different in terms of sharpening at a lower display resolution than it will at a higher one.

  So where color management lets us compensate for a huge range of different display behaviors, we have no such solution for sharpening.

  What we do have is a new set of very general rules of thumb. Use these with caution: You need to learn the relationship between what you see on your particular display at your preferred resolution and the resulting output (just as you had to do with color in the days before color management). With that caveat in mind, here are some very general guidelines.

  Zoom percentage. We believe it's a good idea to look at the Actual Pixels view to see what's happening to the actual image pixels, but unless your output is to a monitor, Actual Pixels view may give a fairly misleading impression of the actual sharpness on output.

  For halftone output, bear in mind that each halftone dot may be comprised of four image pixels. Viewing at 25-percent or 50-percent view may give a truer impression of halftone sharpness. Avoid the "odd" zoom percentages33.3, 66.6, and so on, because Photoshop applies fairly heavy anti-aliasing to those views. For inkjet output, the key factor is the resolution you're sending to the printer. Look at the even-divisor zoom percentage that comes closest to reproducing the image at actual print size on the display.

  How sharp is sharp? For the first two passes of sharpeningcapture and localized creativeour general rule of thumb is to apply sharpening that looks good on a CRT display, or very slightly oversharpened on an LCD display.

  For output sharpening, you can really push the sharpening far beyond what looks acceptable on the monitor at Actual Pixels view, particularly when you print at higher resolutions (like 350 ppi for a 175-lpi halftone or 360 ppi for an inkjet print).

  The key here is bear in mind the actual size of the pixels on output. At 360 ppi, each pixel is only of an inch, so to produce a of an inch halo, you'd need a dark contour approximately 3.6 pixels wide ( of an inch) and a light contour the same size.

  
  

• On very large prints, you may have to use a slightly larger sharpening haloif the resolution is below 100 ppi, the halo will be larger than -inch because it takes at least two pixels, one light, one dark, to create the halo. But large prints are generally viewed from further away, so the longer viewing distance tends to compensate for the larger halo.

Someday, RIPs and printers may even be able to apply output sharpening on the fly, particularly if things like color conversions and trapping are also going to occur there. But for such an approach to succeed, the device will somehow need to know the state of the incoming images (what kind of sharpening has already been performed, and so on).

The Sharpening Workflow

We'll be the first to admit that taking a workflow approach to sharpening is a fairly radical idea, but the more we use it, the more we find that it makes sense. We've done a great deal of testingBruce reckons he sharpened about 5,000 images to build and fine-tune PhotoKit SHARPENERbut plenty of work remains to be done.

The results of two- or three-pass sharpening often justify the extra pains, especially with images we plan to reuse for several different types of output. However, if you're in a hurry, and you're preparing an image for one-off reproduction (particularly with a low screen frequency that can only show a limited amount of detail anyway), one-pass sharpening may make just as much sense.

We don't claim to have solved every conceivable sharpening problem. The techniques that follow are ones that we use every day in our sharpening workflow, and as we describe them, we'll tell you how we use them. But feel free to pick and choose, and to adapt them to your own work.