A short course in correct sharpening ;-)

This document is a guide to help sharpen digital photos according to how they are to be used. The example shows a bolt of lightning which struck a forest on Monday, 16.6.2003 during a thunderstorm which was part of a major weather system over Germany. The image was exposed for nearly two minutes, which explains why some hot pixels appear.

1st image



Post-processing:
- size reduced to half the side length
- from this, selection of a 640x480 pixel section


2nd image



Post-processing:
- brightening using color curves

section unsharpened

section sharpened (RGB 100)

For viewing on screen; the best RGB sharpness range for this example is about 80-120.

section sharpened (RGB 200)

For printing: Here you can see the newly created edges (almost black boundary between the flash and the background). Oversharp for screen viewing!

Another tip:

When should I use which sharpening method?

• Reshaping: Requires much higher values than brightening or RGB sharpening. It does not produce artificial edges, i.e. it does not over-exaggerate (see lightning flash example above). However, it can obscure fine structures. Some parts are sharpened, but small features can disappear through the transformation of the image.
• Brightening: Preserves the colors very well and helps to avoid white patches on fine structures like leaves and grass. Does produce new edges, but these can be coped with.
• RGB is good for really strong sharpening, e.g. for printing. In this case contours may be exaggerated, so make the settings carefully. In some cases the sharpening of individual colors can succeed brilliantly.
• Sharpening edges: Avoids over-exaggeration of the contours, but only sharpens well where there are edges to sharpen. That means that light/dark edges sharpen well, but color discontinuities between relatively similar colors practically don’t sharpen at all. So the effect on subtle structures is very slight. Requires a lot of processing time.
• 3D Sharpness: Only available in PP. This method can always be used because it is relatively good at avoiding steps in oblique edges, so results generally don’t appear oversharpened. The name ‘3D Sharpness’ is used because it makes objects appear more solid.
• Unsharp Masking (USM): Generally the USM technique is considered to be the best, most effective and most universally applicable technique. PhotoPerfect implements standard USM as well as a partially automated method. There are various processing parameters which can be set such as the effective radius of the tool. In this case the radius is ‘soft’, i.e. the tool influences pixels in the neighborhood as well as those directly within the set scope. For instance, with the radius set at 0.9 up to 3 pixels on each side may be affected. Another setting concerns the contrast threshold which determines when a pixel is to be sharpened. With this it is possible (similar to the filter ‘Sharpen edges’) to avoid sharpening distortions in areas lacking in strong contrast. Lastly, of course, the sharpness itself can be set. PhotoPerfect also provides sliders with USM for the three color channels which can be set between 0% and 100%. USM also represents the basis for LCE - local contrast enhancement. This involves generating more contrast for the mid tones. The LCE effect can, however, also be achieved with other sharpening combinations – and a little trial and error. The technical explanation for USM can be found here: http://en.wikipedia.org/wiki/Unsharp_masking
  With partially automated USM PhotoPerfect uses an intelligent algorithm which sets the best radius on its own. This makes the sharpening process much easier to implement. Here, too, there are two variants available, namely sharpening according to the brightness channel (Y) and to the color channels (RGB):
  a) USM-Y is more suited to images with very strong brightness contrasts, because it then simply emphasizes existing light/dark contrast borders more strongly. Example: a sunlit meadow.
  b) USM-RGB is suited more to images with less strong brightness contrasts which contain more uniform color regions displaying a less dynamic range of light/dark alternation. Example: petals become very rich in detail.

Generally: The best results can be achieved with a mixture of methods!

PS: And finally, my special trick for ultra-sharp images that aren’t over-sharp: To put the final touches on an image which has been treated with Sharpen edgesor 3D Sharpness, use a careful dose of Reshaping while observing the image carefully. With this it is usually possible to pick out the finest detail excellently. However, with portraits it is best to mask out areas of skin beforehand.
Cheers
Ralf