Bryson Leidich

Photography and Photoshop

Smart Sharpen


An excellent sharpening tool in Photoshop is the Smart Sharpen filter. It must be applied to a pixel layer, so the best approach is to make a merged visible layer at the top of the layer stack and apply Smart Sharpen at the end of the process. Since sharpening applied to a file should be relative to the image size this will allow easy removal of the sharpening from a psd which can then be reszied for output. However, this is also one of the reasons why final sharpening of most files is best done at the printing stage, preferably using Lightroom.

Roll over the image to the right to see the unsharpened version. Smart Sharpen was applied rather agressively to this image to illustrate the effect possible. Even so, the lack of undesirable halos shows how well the filter works.

Smart Sharpen is actually a composite dialog that uses unsharp masking (with added edge detection), noise reduction, and blendif capabilities in one dialog. This makes the application easier to use, and works exceptionally well. Absent the need for a different means of sharpening it is the best choice.

The first control is Amount, and in conjuction with Radius it works about the same a Unsharp Mask. It has edge detection as part of the process which makes it less likely to create halos, but still must be used carefully. You can start by lowering the Reduce Noise slider so you can see the full effect of the sharpening.

Increse the Amount slider until you see the sharpening effect you want, either in the image preview or the preview in the dialog box. You can have the primary image at a lower percentage (50% recommended) in order to more accurately judge the final effect. Use the dialog box preview at 100% as a comparison. If you have used unsharp mask sharpening before you will first notice that Smart Sharpen can be used at a higher amount without negative effects, but even more refinement is possible.

Increasing the Radius slider will eventually begin to show a crunchy effect, expecially around fine details in otherwise soft areas, like eyelashes or fine tree details. Here you simply need to work the Amount and Radius sliders to come to a compromise that gives you good results. My suggestion is that lower Radius values are better in most cases and I often use a Radius of 1 pixel or less. Values less the 0.3 have no effect. However, higher values may be needed depending on the subject matter.

Next would be Noise Reduction. In an image with little noise the effect will be minimal, of course, but check for noise in the darkest areas of the image for the most effective use. Noise reduction always softens an image, so a compromise is usually needed between some noise and complete removal. Less obvious noise often simply disappears in the printing process, so the application becomes one of experience.

In the "Remove" box you have three choices as to how Photoshop applies sharpening. Gaussian Blur (inverted) is the default and applies the sharpening to the entire image in a normal manner. Lens blur attempts to sharpen the image in a manner similar to how a lens creates an out of focus image as the result of overlaying circles of confusion. Of course, the result is an inversion of the effect, and the result usually looks more realistic than Gaussian Blur. Often the differences are very subtle. The third option is motion BLur which can be helpful if the blur is slight and you can effectively match the direction of the movement.

Since all sharpening relies to some extent on an increase in contrast the bottom half of the dialog is a simpler to use application of the blendif controls. Amount is simply how much to reduce the effect of the sharpening. The tonal width is the percentage of the image tonal range to use to reduce the sharpening effect. I recommend starting with about 20% which would restrict the sharpening to the middle 60% of the tonal range, blending the effect into the values below 50 and above 190.

Less intuitive, and less obvious is the Radius setting which blends the effect into the image based on the edges detected. In small images a smaller Radius can be used and in larger images the blending can be over a greater range. It can be difficult to see depending on the fineness of the image detail. If you can't see a difference on screen at 100% you certainly will not see the difference in a print.