Write in Guestbook Read Guestbook

Sharpening with USM

Copyright Carsten Arnholm. May be reproduced if 100% unchanged, name of author and authors home page
http://carnholm.home.online.no
is included.

07. nov. 2001 CA
Thanks to Alan Browne for proofreading
the English version.

What is Unsharp mask?

Unsharp mask (USM) is an editing tool for digital files. Simplified, USM is used in order to make digital picture, which is sharp, also to appear visually sharp. (Be aware that the abbreviation "USM" as used in connection with Canon lenses means "Ultra Sonic Motor").

Is there a significant difference between sharpened and non-sharpened images? See for yourself!

Unsharp mask strange name?

A dear child has many names: Unsharp mask, abbr. "USM", the "USM-filter" etc.. The use of this tool is called sharpening. The term "sharpening" does however cover several types of sharpening tools, many of which do not allow for the same amount of complete control as the sharpening tool: "Unsharp mask".

The odd name "Unsharp mask" is a relic from earlier repro techniques, also used in the darkroom. In contact with the original reversal ("slide") one made a relatively thin, unsharp B&W negative. This unsharp negative was called a mask (unsharp mask), as it partially masked off the original. This unsharp negative was positioned in exact register with the original (every point in the original and the negative in precisily the same position). This results in an increase in the contrast between contours. The effect corresponds largely to what today is found in editors such as Photoshop, where it is called the "unsharp mask".

In Photoshop, here Photoshop LE, you find USM under "Filter->Sharpen->Unsharp Mask.




The parameters T, R and A shown with a typical setting for small files, such as 600x500 pixels.

"Filter" in Photoshop refers to a series of different functions for editing digital image files.

For the beginner USM may be difficult to understand and difficult to use effectively. I have therefore decided to publish this article containing practical advice. This advice is wholly based on my own practical experience, and the advice given is accordingly my own personal opinion. BTW, I'm making no attempt to produce a complete description
of USM.

What "Unsharp mask" is NOT!

Using USM is a totally normal and necessary part of editing and finishing a digital image file. Hence, it is certainly not a tool designed for manipulation etc.. But of course, USM may be used for other purposes than those it was designed for, as may all other editing tools. Apply USM several times on a file, with parameters set for maximum effect, and the file will very quickly become very different from the original. Neither is USM a sort of digital "trick" which can cover sloppy photographic techniques. If the original is unsharp, USM cannot be used to recover what was lost at the time of exposure. As already said, USM is used in order to transform a digital picture, which is sharp, to also appear visually sharp. Be aware that USM is a tool that needs care and understanding. With careless use, USM is quite able to destroy almost anything. In experienced hands it is however a priceless tool.


Why is it necessary to use Unsharp mask?

To repeat myself, the term "sharpening" may mistakenly be taken to indicate that USM is a tool which corrects from "unsharp" to "sharp". This is not so, and whoever supposes anything in that direction will be guilty of a grave misunderstanding. When it comes to photographic images, two separate elements will influence whether the human eye perceives the image as sharp. Fine detail must of course be correctly recorded. This is a job for the optics and the photographer. The second element is reasonably good contrast between contours, rather, between adjacent levels. It is this last "element" which usually suffers somewhat during digitalization.

It is quite normal, actually the rule, that digitalization (directly, in a digital camera, or indirectly with a scanner, including a film scanner), leads to a certain amount of "softening". Contrast between adjacent areas may appear indistinct. This effect does not need to be strong to have a negative influence on brilliance and visual perception of skarpnes. However, when USM is used in a sensible way, the contrast adjacent areas will increase. This is exactly what heavily influences the human eye when it perceives what is "sharp".

In other words: USM is a tool to ensure that the real qualities of the image to the best of its advantage.

How should Unsharp mask be used?

The USM-function in Photoshop has 3 parameters,
        "Threshold"
        "Radius"
        "Amount"
It is important to understand the effects of these parameters and their mutual influence.

Threshold

Threshold (T) will in practice determine which differences the USM function acts on. Threshold at 0 (zero) means that all differences are relevant. With T at zero contrast between all adjacent pixels may increase. "Difference" is to be understood as difference in luminosity between pixels.

USM may in many instances be used effectively with T set to zero. However, many images may have apparently smooth areas (skin tones, sky etc.), where T set to zero may create a muddy, grainlike structure ("skin disease" in portraits). The reason is that normally there are small variations between pixels even in apparently completely smooth areas. When such differences are increased, contrast between pixels is made visible. In fact, apart from digital noise, "grain" created by careless (over)use of USM is a very frequent occurrence among beginners. It is not always easy to find a good balance, namely one where muddiness or "grain" is at a minimum, while the desired increase in visual contrast between levels is achieved. Coversely, if threshold is set above zero (say 10, 20 or higher), the influence on smooth areas decreases or disappears. As the value of T increases, USM will increasingly only have affect where contrast between levels already is distinct. In other words, it will sharpen edges and little else.

Radius
Setting radius depends in my opinion primarily on two conditions:
The number of pixels in the file. Not file size (disk size), but the number of pixels in the actual file.

The character of the file or image. The character of the file or image, or more precisely, is the image really sharp and does it contain much fine detail.

The number of pixels
As the value of radius determines how many pixels the function should evaluate when one pixel is changed, it follows that the setting of radius to a large extent must be influenced by the number of pixels in the file. If radius is set to 1, it represents 1/400 in a file of 400x400 pixiels, while that would be 1/2400 in a file of 2400x2400 pixels.

In a file with few pixels (500x600, a typical websize), radius set to 0
.3 or 0.7 is often suitable. With a file of 2610 x 4060 pixels (about 30MB) radius may well be set up to 3-7, although a lower setting should usually be tried first. Visual evaluation of the effect visavis the individual image is absolutely necessary. In Photoshop (or other program for editing) the file must be displayed in the exact size as a print or the size it is to be displayed on a computer screen. This is necessary to be able to evaluate the effect realistically.

The character of the file or image.

When an image (as recorded on film or digitally) has a high degree of sharpness, it is my experience that a relatively low setting for radius is suitable. And if such a file has many sharp and fine details (hair, branches, fibers etc.) it is often necessary to test several values for radius. If radius is set to a high value, the effect on fine detail is low. Conversely, a low value combined with a high value for "Amount" (see below), may have a more pronounced effect on fine detail. This is not simple!

Images with smooth tones (skin tones, sky etc.) may get a lot of "grain" if values for both radius and threshold are kept low. If threshold for some reason cannot or should not be set to a higher value, it may be advantageous to increase value for radius. As already said, with a high value for radius the effect on small detail is diminished. Small detail in this respect may be small variations in smooth areas (skin, sky).

Amount.
"Amount" is largely self descriptive. Amount determines the amount(!) of effect from threshold and radius as these are set in the individual case. In actual practise it seems to be many ways to reach the same goal. Low radius with high "Amount" may result in effects that are not easily distinguishable from a high radius with low "Amount".

All advanced graphic editors (Corel Photo Paint etc.) have similar parameters for their USM functions, whatever the functions and parameters may be called in those applications.


When to use Unsharp mask?

USM should definitely be used as the very last editing step (the only exception is removal of artefacts created by USM itself). USM has a destructive effect, to what extent depends on the setting of its parameters. Sharpening with subsequent editing (such as correction of colour balance or contrast) is definitely to be avoided. This is the reason why I am deeply sceptical to allow a film scanner to do sharpening during scanning, since there is always some editing, often a lot, to be done in Photoshop. This argument should apply equally to in-camera sharpening with a digital camera.

Be aware that for optimal quality USM should be applied to file types such as TIFF or PSD (with non-lossy compression. The usual file type with "lossy" compression is JPEG (*.jpg). When an image is to be saved as JPEG, the sharpening should always be undertaken before the file is saved as JPEG.

Errors when using Unsharp mask

In all editing of digital images caution is paramount. Nothing is simpler than the abuse of good tools. Less is more.

Various forms of excessive use of USM.

"Grain" has already been mentioned, usually caused by USM applied with excessive parameters, the most usual cause is to low threshold value. It is of course possible that film grain (after film scanning) has been made (more) visible by USM. The increased pixel contrast in apparently smooth areas has already been mentioned. The remedy may be to increase the value of threshold.

Another error, not easily avoided, is distinct, visible contours between dark and light areas in the image. This may appear where a dark hill stands against the sky, dark branches against the sky etc. A distracting light, often almost white contour appears. Dark clothes against skin, spectacles, flowers, text etc. may also be subjected to this unnatural effect. The remedy as almost always to reduce Amount, possibly also Radius.

A typical error, after USM, is one or more isolated pixels with values that are markedly different from the surroundings, the may be totally white or black. Again, caused by USM applied with excessive parameters. A revision of parameters is usually the only remedy. When such effects are few, the "Rubber Stamp Tool" in PS may be used for spotting in the usual way.

Illustration, excessive use of USM

What to do when one part of the image demands one set of USM parameters, and another part exhibits errors as described with these parameters? Any advanced graphic editor (typically Photoshop, also the LE version) allows for selecting just one area, where editing steps, including USM, can be taken with effect only in the selected area (possibly put on a separate layer). Use "feather" (in PS, Select->Feather) so that the special treatment of different areas blends well.

It is possible that a particular area in an image needs more sharpening than other areas. An example may be a sharp portrait with a relatively unsharp background. Also, some users may like to do a little extra sharpening around the eyes etc., etc.

Non-traditional use of USM

Those of us who are more concerned with creating pictures other than just photography may of course us "Unsharp mask" constructively to create effects that have a positive pictorial (or "artistic") effect. As an example, USM may be used with a very high setting for radius (50) and a very low amount (50). The effect is rather similar to advanced use of the "Curves" tool, but often with much less work. Depending on the individual image, this may cause various levels to be very well separated.

Illustration, non-traditional use of USM

Summing up

Using USM is certainly not a science. Hard, absolute rules do not exist, nor should anyone try to establish anything in that direction. In this article I have attempted some advice based on personal experience. In the end it is a question related to a basic technical understanding about what one is doing, training, personal experience with respect to effects and pitfalls. Of course, pictures have to do with a pictorial sense and an ability to see. But that is a highly personal thing.

07. nov. 2001 CA