Unsharp Mask
I’ve been meaning to post a link to this article for some time: Understanding Sharpening. (Although this article is on Microsoft’s web site, it’s actually very good!)
As some of you know, I got my first digital camera a few months ago. I was a little disappointed with it in some ways, and one of the reasons was because I felt the images were a bit on the soft side. Anyway, now I’ve read this article I’m much happier about it all, and following Rob Galbraith’s advice in this article has certainly helped me get more pleasing results.
I’m yet to actually print any of my photos, but I will be doing so shortly, so as always, watch this space for updates!
By the way, Rob has got quite a few really good articles about digital photography on the Microsoft web site, and they are well worth a read!
Update 30 December 2005 The Understanding Sharpening article has gone off the Microsoft website. Doing a search of the site turns up a copy of it under the israel section, so in case the article goes completely, here are some interesting bits from it:
…
To compensate for the blurring effect of the low-pass filter, sharpening must be applied before the picture is resized/resampled for a specific output process, but after other corrections—especially noise reduction—have been completed.
To compensate for the softening effect of the output process, sharpening should be applied after the picture has been resized/resampled
…
Amount. Sharpening doesn’t actually bring back lost detail, it simply adds edge contrast to the detail that’s already there. The Amount controls how much contrast is added to light-dark transitions in a photo. As it happens, light-dark transitions are almost always edges; increasing contrast along these transitions enhances the detail and makes the picture appear sharper. The higher the Amount value, the greater the contrast increase.
Radius. This setting determines how many pixels out from an edge will be affected by the edge contrast adjustment. The higher the Radius value, the greater the width of the affected area. If you’ve ever seen a strange halo along edges in your photos after sharpening, you’re seeing the effects of what was probably a too-high radius value.
Threshold. This setting dictates how different in tone adjacent pixels have to be before they’re considered a light-dark transition to be sharpened. The lower the Threshold value, the greater the number of pixels that will be sharpened throughout the photo.
…
To compensate for the blurring effect of a camera’s low-pass filter, the key is a low radius value: 0.3 to 0.5. Only cameras that shoot particularly soft photos will require a radius higher than 0.3. The Threshold setting should be in the 0-2 range. Move it above 0 only if you see noise appearing in dark backgrounds as you preview your sharpening. The Amount will be between 200-500%. Examine the image at 100% magnification, and pick an Amount that gives the photo the kind of snap you would expect to see from the good lens you used to shoot the photo. Landscape photographer Stephen Johnson once referred to this sharpening step as restoring the native sharpness of the lens. The volume of sharpening you apply should be to accomplish just that, and nothing more.
When it’s time to sharpen the picture for the printed page, the Unsharp Mask filter settings are different, and vary widely with the intended output. In general, printed output of any description will require a higher radius value than described above. The coarser the printing, the higher the radius. For instance, when sharpening an average-sized photo for newsprint reproduction, the radius will be at 1.0 to 1.5 typically. To print the same photo on a premium quality inkjet photo printer, the radius need be no more than about 0.6 to 0.8. As print size increases, so too does the Radius value. The Threshold in most cases should stay low—under 10 - while the Amount will float from picture to picture, and is highly dependent on the output process. With practice, you’ll develop a sense of where the Amount needs to be set for your printing workflow.
OK, quite an excessive amount of copying there, but it’s useful stuff and it would be a shame to see it vanish.









