Sharpening-What constitutes alot?

rjbuzzclick

rjbuzzclick

Well-known
Local time
7:38 AM
Joined
Jul 20, 2009
Messages
415
Hi,

I'm not looking to get an answer for "Is this too much sharpening?" or "What setting should I use?", as those questions can only be answered when talking about a specific image. Rather, I'm asking from a general technical workflow point of view as to what constitutes a lot of sharpening. When I'm processing my film images, I seem to end up in the following ranges.

Amount: 40-75%
Radius: 2.1-3.0 pixels
Threshold: 1 levels

If one is going for a sharp image, does this level of sharpening constitue heavy use of the tool in your opinion, i.e. trying to correct for sharpness issues that might be better corrected "upstream" in the workflow, by using a tripod, smaller aperture, flatter film, etc?

How much sharpening do you find your images seem to need? I only ask to try and establish a personal point of reference in my own workflow.

Thanks,
 
Last edited:
it depends on the scanner, etc.

The main point is to avoid halos around oversharpened edges. You have to evaluate your images at 100% pixel to pixel magnification.
 
The numbers alone don't say a lot - it depends on the size of the picture (# of pixels h x # of pixels v), and when in your workflow you are applying this sharpening step.

I would generally advise to use sharpening as late as possible in the workflow because sharpening is a destructive processing step (meaning that mistakes made in this step are very hard to correct in subsequent processing steps).

The best place to use sharpening (other than as part of a very specific 'creative ' manipulation) is immediately before printing, after you have applied the final crop and scaling operations. Please note that there are different recommendations for sharpening, depending on the printing technique you might use. Inkjet printers require a different amount of sharpenung than other printing techniques (search for appropriate info e.g. on Luminous Landscape).

When you apply sharpening, following a number of simple rules will help you avoid the most obvious mistakes:
  1. Open the image in your image processing program and set image magnification to 100%. This displays a rather small part of the image (choose appropriate detail), but shows you in highest clarity which effect your sharpening operation will produce.
  2. There is a typical sign for oversharpening: Bright-to-dark contrast edges will develop haloing, i.e. black lines will show an additional white line, or white lines will exhibit a black shadow line. The larger the radius, the more easily will you get haloing. Try to start with values between 1 and 2 pixels. As an aside, if you use very large radii , you can influence overall image contrast without oversharpening artifacts (but results when overdone can get ugly).
  3. The numerical value for 'amount' depends on the software you use, so I can't offer a meaningful recommendation for this parameter. But - the larger the value, the more risk of haloing!
  4. Try to experiment with levels. Start with zero - in this case, sharpening will be applied to all image areas irrespective of whether the exhibit significant brightness differences (contrast). If you increase the value, you will see that sharpening will gradually concentrate just on those areas that exhibit larger brightness differences. Play with this parameter to gain a better understanding!
rjbuzzclick said:
If one is going for a sharp image, does this level of sharpening constitue heavy use of the tool in your opinion, i.e. trying to correct for sharpness issues that might be better corrected "upstream" in the workflow, by using a tripod, smaller aperture, flatter film, etc?
Click to expand...

'Unsharp masking' - as this process is called correctly - is no remedy for blurry or out-of-focus pictures, as the USM corrections to mend such deficiencies would have to be so severe that you would invariably obtain sharpening artifacts. Better make sure to use a sufficiently high ISO value / fast film, and have a tripod ready.
 
Last edited:
One thing I've never understood is how there can be a radius of fractions of a pixel. I think of a pixel as being an atomic unit. What am I not understanding here?
 
A pixel is indeed an atomic unit, they cannot be split or divided in any way. CRTs can make them appear less distinct, but they still are. The only reason they appear less distinct on a CRT is basically because CRTs are brighter and the pixels "glow" more, and also CRTs are somewhat blurry compared to LCD.
 
The answer is it all depends. For the size and type of images I usually take, I think your amount values are conservative but your radius values are a lot.

You will find that the amount of sharpening you are able to do will depend a lot on the type of image. For example, I have found that its easy to over sharpen an image with a lot of detail in it (like small tree branches) so that it looks harsh and unpleasant to the eye. I also find that generally when using USM I get better results with the radius set low (below 1) and the amount set higher (it can be 100% to 200%). The threshold sets the amount of variation in tone that the software looks for to define an edge. Thus if set high less edges are sharpened other things being equal. This mean that smoother areas like sky do not get sharpened as much. Also I have found that using a radius of much more than 1 makes it too easy for edge artifacts like haloing to appear in the image. But it also depends if you are planning to view the image on screen or print it - a printed image needs and can take more sharpening.

I have come to prefer high pass sharpening to USM. It is better at only sharpening edges than USM where even with the threshold pushed up a bit some of the areas you do not want sharpening never he less do not get it.

If using USM however I suggest trying our some settings like the following to see if they work with an image (be prepared to experiment and adjust from image to image to get best results.)

Amount 100-150%
Radius 0.6
Threshold 2-5
 
USM can do more than just sharpen an image!

USM can do more than just sharpen an image!

Roger Hicks said:
Arjay: thanks for a wonderfully clear summary. Some of it I knew, but some I didn't. Thanks again.
Click to expand...

Thanks for the compliment. Actually, I just found a series of Photoshop tutorials which gives a brilliant explanation of the idea of sharpening.

The key term for sharpening is global/local contrast. USM is normally used with small radius values for the purpose of sharpening. But used with larger radius values, USM can modify the local contrast of an image - as opposed to levels & curves which is used for global contrast manipulation!

Ron Bigelow has published a series of very interesting articles on this topic. Look for 'Localized Contrast Part I to III'. Highly recommended reading!
 
reid, here's what i sometimes do when i'm unsure on sharpening:

set the amount to maximum (500%), set radius to 0.0, threshold to your taste

set magnification to 100%

increment the radius in steps by 0.1 and watch for the image to "pop"

once at the "pop" settings, adjust amount downward to the minimum level that retains the amount of "pop" you want. tune the radius and threshold slightly, if necessary.
 
Last edited:
Sharpening is at least a two step, once when you import or digitize and once at final size /resolution. Some images need a step inbetween.

Initial is in camera raw even it it is scanned TIFF, 75 amount .8 pixels. or USM same numbers.

A final size JPEG is done again. This needs to be done at final size and rez and 100% image size. This is size/medium dependent, glsosy & matt are different. The best thing to do is increase until you see halos, then back off radius. With smaller radius, you can increase the amount. YOU MUST OBSERVE THE EFFECT ON THE PRINTED MEDIUM. a MONITOR WILL ONLY GET YOU SO CLOSE. Don`t let it get crunchy looking.

Real world Sharpening by Frazer and Schrew, Amazon.com $35.

There is far more in it than I can type here.
 
Beyond the edge sharpening, I find it is useful to use local contrast sharpening, especially when dealing with film, as this gives a pop to the image without causing the grain aliasing. Instead of using amount 50-200 and radius 1-3, use amount 20 and radius 40-60. This kind of sharpening alters the curves somewhat, so it should be done at the very beginning of the editing. You have to be careful though, because sometimes, when you have dark areas against bright background, you can get the "saint" effect, like in case of this shot here, which I edited somewhat mechanically...

4104051226_5a8aa3c0de_b.jpg


Then, as suggested above, you can apply an appropriate amount of conventional output sharpening right before you print.
 
i agree with the local contrast pass, which i favor for B&W conversions, usually 20/50/4-5. also use it for scans, depending on the subject. but i usually do the local contrast pass earlier in my post-processing, after white balance and level adjustments, but before any other contrast work.

something else i employ is to do my sharpening within Noise Ninja, chroma only, especially for portraits and for high ISO, low light images.
 
You must log in or register to reply here.
Back
Top Bottom