Image sharpening is an artificial transformation that can make images appear more in focus by exaggerating transitions that have been dulled by scanning. Most scanned images benefit from sharpening however too much sharpening can produce unwanted artifacts.
Sharpening is performed with an unsharp mask filter, a counterintuitive name that originated in the printing industry when it was performed optically via a process in which a purposely blurred (unsharpened) negative copy of an image was used as a subtractive mask in combination with the image itself to produce a sharpened copy of the image.
The unsharp mask filter places a dark halo at the edge of the darker of two image elements and a lighter halo at the edge of the lighter element. The process modify's the original image and should be the last operation performed before the image is saved or printed.
Parameters
The selection of radius, strength, and threshold parameters to be used for unsharp masking before display or printing is largely a trial-and-error process and image editors provides a slightly different variation of them for the user to adjust. The goal is to determine a set of parameters that enhances the image without introducing excessive noise.
Unsharp mask filter parameters (radius, strength, threshold)
RST = 4, 100%, 0
RST = 1, 100%, 0
RST = 0, 0%, 0
(original image)
Radius (pixels) determines the width of the transition area in pixels. Larger radius values produce wider halos.
Images with fine detail, e.g. fabric or skin, usually require a smaller radius to prevent loss of detail. Larger images with more pronounced detail can benefit from a larger radius. Typical values range between 1 and 3.
RST = 4, 200%, 0
RST = 4, 100%, 0
RST = 4, 50%, 0
Strength (percent) controls halo contrast. Larger values darken the border around the darker element and lighten the border around the lighter element.
RST = 4, 100%, 25
RST = 4, 100%, 0
Threshold controls where the unsharp masking effect will be applied. Sharpening is performed only if the tonal values of neighboring pixels differ by more than the threshold setting. This lack of action is important to prevent smooth areas from becoming speckled.
Human faces usually require values greater than 1 or 2, e.g. 5 or more. For inanimate objects 0 or 1 can be useful. For general work try 3 or 4.
Unsharp mask examples
Original image
RST = 2, 100%, 0
Oversharpened
RST = 2, 500%, 0
Notes
- Sharpening is an irreversable transformation that disorts the original image. It should be the last editing operation and only lightly used on archival images.
- Sharpening during scanning usually offers less control than in an image editor.
- Sharpening can not correct a blurred original.
- On color images, avoid or minimize sharpening the darkest image channel, e.g. green or blue, to reduce the effect of noise and unwanted image detail.