Digital Photography Sharpening Guide

The act of capturing a photo, be it on film or digitally, is just the beginning of a series of steps that lead up to a final goal, either display or print. In the digital realm, one of those steps is sharpening.

Digital images are inherently somewhat soft. This is due to the fact that the sensors have a low-pass filter (anti-aliasing filter) to help cut down on moire patterns and to keep angles and curves from becoming jagged. These AA filters, however, tend to soften up the important details that we expect from a sharp lens/sensor combo. And that’s before any resizing in post production.

There are millions (OK, maybe not) of different approaches to sharpening, as well as nearly as many pieces of software that claim to be the ultimate. Fact is, if you have a basic understanding of the Unsharp Mask filter in any mid to high-end photo editor you can get very excellent results.

For my examples, I will use Photoshop. The same mechanics apply to unsharp mask in almost all other pieces of software. (Please, don’t get caught up in the hype, unless you’re doing high-end production work, you simply don’t need Photoshop, something simpler and cheaper will work just as well.)

First, you need to understand how digital sharpening works. It tries to find differences between neighboring pixels and exaggerate them. So if you have groups of grey and black pixels next to each other, it will try to increase the apparent difference between them by adding pixels along the border that add contrast between the two, as can be seen in the screenshot below.

Because of this, there are some rules of thumb:

1) Sharpening should be the last filter applied.

2) Always sharpen for the intended FINAL OUTPUT of the image.

Why? Because of different output sizes, viewing distances and output media different settings will look, well, different. What works for the web will likely do nothing for a 12 inch print. For each intended output size you will need different settings. For example, settings for a web post will be very subtle, and the contrast added to amplify the edges might be in the vicinity of 1 pixel wide. If you try to use these same settings on an 8×10 print at 300 dpi, you will likely find that the sharpening gets lost in the tiny details of the image, rendering it effectively useless. Stronger settings would be used for that latter situation.

The reason the filter should be the last thing applied is because unsharp mask is an inherently destructive filter. It changes the pixels of the image to trick our eyes into perceiving more detail, but those changes don’t go away and are affected by any further changes made. If later on you apply more changes, the new sharpening pass will not only sharpen image detail, but it will also sharpen the new pixels created by the previous sharpening pass. In effect, you’re sharpening the sharpening.

The trick with sharpening is to keep it subtle enough to not be noticeable, while noticeably bringing out the details in the image. The pixels added during the process, when applied to strongly, create “halos” around details. Bringing out detail while minimizing halos means understanding the different controls of the unsharp mask filter.

Amount – This is the strength of the newly created pixels. The higher this percentage or number, the more contrast the filter tries to generate between the target pixels.

Radius – This is the actual thickness of the newly created pixels. For a web post this might be tiny, in the 0.3-1.0 pixel range. For a print, this might be several pixels.

Threshold – Think of this as a cutoff. As the filter compares pixels to sharpen, it checks how different they are against this value. If it’s at zero, all pixels that are different from each other are sharpened. This can be a problem if for example a sky has dark blue and medium blue pixels next to each other. If the threshold is low, those pixels will be sharpened and the sky will appear noisy. If this is intended, then fine, but often it might be desirable to have a smooth sky, so by turning up the threshold to maybe 5, we’re telling the filter to ignore any pixels that are 5 RGB values or less apart in value. So the higher the number, the bigger the difference has to be between pixels for them to be sharpened. Careful adjustment of this value helps keep noise out of smooth areas and keep sharpening contained to those areas that need it.

Those are the basics of the unsharp mask filter. By understanding that every output will have different settings and by taking the time to experiment you can quickly master this filter and add that certain dimensional quality that images straight out of a camera sometimes lack.

Tips:

Turn down the sharpening inside your camera if you can. Usually the default settings make the image look good on screen, but can hurt the results as you process the image. See the explanation above about sharpening existing sharpening.

If you find yourself outputting the same format often, record the settings. Either write them down, or create an action. I always post images to the web at the same resolution, so I saved an action that applies the settings I like best.

Experiment! You can use huge radius values on an image to create interesting contrast effects. Just because it’s called unsharp mask doesn’t mean it can only be used for sharpening.

Some images showing different settings: