Unsharp Mask settings

[JoshD]

Is there a standard setting that you use for most pictures you apply this too? If so, what generally are those settings? I am looking to sharpen up my pictures just a little bit without going overboard or ruining the image. What works for you?

Or do you generally change the settings for every picture?

Thanks

[Medley]

The amount varies dramatically with image content.

The radius varies occasionally with image content, and output medium (ie web or print)

Thereshold almost never varies.

I have a system whereby I can get a good indication of what the image will look like BEFORE accepting Unsharp Mask settings- even if the image is destined for print. It's a royal pain the the %^$ to learn, and a minor inconvenience to execute even now, but for pin-sharp prints, it's worth it.

- Joe U.

[GB1]

So what's the system, Joe?

As said, it varies with content and whether you plan on viewing or printing. Most of the time, I sharpen up my images in Corel PhotoPaint using their "Adaptive Unsharp" function, which seems to work very well, as it has a slider that allows you to select from 1 - 100 %. If I resize first, say to 800 pixels (for this site) then I typically need about 30-40% sharpening with that tool. I like their sharpening a LOT better than PS CS2's. High contrast tends to reduce the need for sharpening a little. Think night shots.

In all though, there is no exact equation as far as I can tell. Maybe Joe has a new equation?

[readingr]

Generally % can range from 20 - 200% depending on what I want, Radius I never seem to go above 1.5 but normally at 0.7, and Threshold is normall 0 sometimes I up it to about 3.

It is so difficult to explain when to set these. Landscapes take more sharpening than portraits and still life inbetween.

Guess that diatribe doesn't help much.

Roger R.

[Medley]

So what's the system, Joe?

I think you and I have discussed this before GB:

The system requires three things:

1) knowing exactly what your screen resolution is,

2) knowing how the percentage views in Photoshop affect that resolution, and

3) understanding how to calculate the effective resolution of a displayed image.

there are a few side rules as well, I'll get to those in a bit.

So, how to calculate your exact screen resolution? That's as easy as measuring the width of your display ( NOT the monitor as a whole, JUST the part that is the display) and knowing what your screen resolution is set at. My display is fourteen and a half inches wide, and the display is usually set at 1680x1050 resolution. so 1680 divided by 14.5 inches gives me a normal screen resolution of about 116 ppi (115.862, if you want to be precise).

So now, lets's say that I open an image in Photoshop, and it comes up in the 50% view. What's the resolution of the image? Well, at 50% view, PS is displaying every other pixel in the image. So. if my screen resolution is 116 ppi, and PS is displaying only half the pixels in the image, then the effective resolution of the image on my screen is 116 X 2, or 232 ppi. I hope that makes sense, because it's the basic building block of the whole system.

I need to mention the first side rule here: you MUST resize the image to 300 pixels X the desired print dimensions for this to work. If you don't, the printer will change the pixel dimension of the image, and the sharpening settings will be off.

So, if we could find a way to make the effective resolution of the image 300 ppi, then we would effective be looking at what the print will look like, since 300 dpi is the resolution of most printers. But I can't get there using multiples of 116. So I need to change my screen resolution.

OK, so let's look: my available screen resolutions are: 1680x1050, 1440x852, 1280x1024, 1152x720, 1024x768, and 1024x640. Some of these are 'stretched' to fit my widescreen.

Now, with my display being 14.5 inches wide ( and once you understand the math) the apparent solution would be to switch to 1440x852 (99.3 ppi) and use the 33% view in PS to get an effective resolution of 297.9 ppi. It would be perfect, except it violates the second side rule: NEVER USE THE 16%, 33%, OR 66% VIEWS IN PS FOR THIS SYSTEM! I cannot stress that enough. The reason is that PS adds a LOT of anti-aliasing to these views to make them work, so the on-screen image is much softer than the actual image. Sharpening at these views almost always result in an over-sharpened image.

So back to the drawing board. I can't use the 33% view, so let's look at the 25% view. To get an effective resolution of 300 ppi with the 25% view, my screen resolution would have to be 75 ppi. And to get 75 ppi with a 14.5 inch display, I'd need a pixel dimension of 1087.5. That about halfway between my 1152 and 1024 resolution settings. I tend to use the 1152x720. That gives me a screen resolution of 79.4 ppi, and (at 25% view) an effective image resolution of 317.8 ppi. I compensate for this by sharpening till it looks good, then backing off just a tad.

Once you understand how to get the effective 300 ppi image onscreen, the rest is easy. You can use any sharpening technique, with any settings you prefer. As long as the 'preview' box is marked in the dialog box, you will see how the settings affect the image, and how they will affect the print- before you click the OK button.

And THAT makes all the difference.

- Joe U.

[OldClicker]

If I am comprehending this right, you could have just reset your View to 39% in the beginning and you would be at 300 DPI, though you would not be able to see the entire image. - TF

[Medley]

The theory is sound, and it may be worth exploring. It presents an interesting question though. How, exactly would I achieve a 39% view? In CS2, my view percentage zooms from 33.3% to 50%. The only time I can get an odd percentage is using the 'fit to screen' view.

Even if I could, I'm not sure it would yield good results. I know for a fact that view percentages that are even multiples of 2 ( 12.5%, 25%, and 50%) work very well for this technique, and view percentages that are even multiples of 3 ( 16.7%, 33.3%, and 66.7%) do not, due to anti-aliasing. So I suspect that there is a type of harmonic effect at work, and if I had to guess, I'd say that at 39%, the anti-aliasing would be a bit better than at 33.3%, but not as good as 50%.

That is just a guess though OldClicker. In the end, I do it the way I do because it's what I know, and because I know it works.

- Joe U.

[OldClicker]

Joe,

I, being new, have CS4 and you can just type in whatever % you want. How would you test for the anti-aliasing (soon as I look up what that is) other than just looking at sharpness results.

Also, I have seen it suggested that printing requires more sharpness than viewing because of the round shape of the 'dot'. Have you found that you need to compensate for images intended for the printer?

TF

[Medley]

TF,

Examination and comparison of actual prints is, in my opinion, the absolute best way to rate techniques. After all, that's what it's all about- a sharper print. In my research, I used a company that allowed back printing- printing the file name and date on the back of the print. Then I kept specific notes about each file name.

Anti-aliasing is, in some ways, the exact opposite of sharpening. It is a lowering of contrast at the edges of an image to provide a smoother transition between the tones. One of the reasons this is done is to prevent the image from looking pixelated. The thing to remember about the above mentioned percentages is that the anti-aliasing is being applied to the display of the image, not the image itself. So the display makes the image look softer than it actually is.

In relation to your second question, certainly there is a softening of the image in converting from square pixels to round dots of ink. For myself personally, I compensate for that in how much I back off the sharpening from what looks good on my screen (remember, I sharpen at an effective resolution of 317.8 ppi.)

But there are literally dozens of variables which can affect the sharpness of an image. The type of output device is just one. For example, you compensate for a half-tone printer differently than you do for an inkjet (which, btw, is relatively easy). Some variables are relatively important, others not so much.

In doing my research (which spanned about 9 months), I found I got to a point of diminishing returns. In the beginning, the sharpness of my prints improved by leaps and bounds. In the end, I found my nose two inches away from 3 separate 5X7's, trying to determine which print was sharpest. At that point, it dawned on me that it really doesn't matter. The three prints are identical to the average viewer.

But my advice to you TF, would be to do your own research using CS4. It may well be that no anti-aliasing is applied to view percentages now (which, to me, would be wonderful).

For those that are interested, my techniques are an adaptation of techniques found in the book "Real World Image Sharpening with Adobe Photoshop CS2" by Photoshop hall-of-famer Bruce Fraser. It's a productive (if somewhat dry) read, and my main reference manual for all things sharpening.

- Joe U.

[OldClicker]

Thanks. - TF

[drg]

I'm not sure we answered much, if any of the basic question about using the Unsharp Mask (USM) and settings 'to start off with'.

Sharpening is at best subjective based on what the image is and of course its final output. With that in mind it might help to just quickly review what the USM settings do!

Amount: The Percentage of strength of the Sharpening. With a high resolution image (such as for a medium to large print) a good starting point for most work is in the 100-150 range. (see below)

Radius: The Radius is the distance used for calculating the amount of the sharpening effect around each pixel. 1 being a whole pixel of 'effect'. Radius can be thought of as how coarse the sharpening appears. The wider, or greater the value, the more edges along the sharpened area will display halos (i.e. the reason this has been called the Halo Width adjustment as well as Radius). Too much of this and the sharpening appears unnatural and degrades the image. For most well focused and sufficiently contrasty high resolution images between 1 and 2 is good starting point. (again, see below)

Threshold: The most easily misunderstood and improperly applied setting, Threshold controls level of difference between pixels (see accutance below) and should be adjusted for smooth areas or fine textures. A setting that is too low can induce noise in areas such as skin or backgrounds that are regular or one color. A clear blue sky that suddenly becomes mottled is a result of a low Threshold setting. Numbers to start between 4 and 10 are good for most content.

'The Below'

What we adjust usually when sharpening is accutance and contrast.

Accutance is the change between tones across an image and how much it changes.

Contrast is a gray scale change and varies from the pure black and white transition to the very subtle and often imperceptible change between gray tones.

Threshold provides a clamp on how quickly or abruptly these changes occur.

Resolution and noise will affect how a image looks both negatively and positively and other than degrading or enhancing an 'effect' the USM tool doesn't really affect the quantity of either of these factors in the final appearance so I won't go into them.

Sharpen as the last step before making a test print. Do all your other adjustments first. Sharpening with the USM has a way of bring out the bad with the good, so leave it for last step before generating the final output be it print, separation, half-tone, or web page illustration.

Some Gotchas ! !

If an image is being sharpened for some types of display such as a large image not to be viewed from any closer than 15 or twenty feet, such a billboard, it has to be adjusted differently. Else it may appear muddy at a distance or the colors may 'run'. Up close it may be gorgeous but further away it doesn't always work.

If an image is being sharpened for web output the first consideration may be that it is now a lower resolution image and the numbers for the USM setting will need to be altered accordingly. The Radius being decreased is the first such as otherwise the 'halos' can overwhelm the end result.

One other consideration for all the 'magic' is that what the resolution is in ppi, dpi, etc. for print isn't best thought of in terms of that 300dpi related number. If the image is for a half tone or offset process, just setting it a 300dpi will not be correct automatically.

The late Bruce Fraser who was mentioned earlier was the first to say that this is all very subjective and often winds up being a matter of individual, the photographer's, taste. His last book or the latest he contributed too posthumously about sharpening sort of modifies a lot of his early work as the technology has changed. Both front end software (Photoshop, Lightroom, and other editors) and the monitors have changed in some degree and certainly how they interact.

One of Fraser's biggest contributions pretty much universally heralded is his idea to pre-sharpen everything. Since digital work whether scanned, generated, or from a direct device(camera) will require sharpening in the end, so just go ahead and sharpen everything by some amount for the best 'initial' comparison. It can always be undone and re-evaluated. Never Sharpen Any ORIGINAL And Save It!!

Thus where do we start with USM sharpening??

For a high resolution (5 Megapixel print or larger):
Amout in the 100-150 range
Radius in the 1-2 range
Threshold in the 4-10 range depending upon the amount of smooth or regularly textured areas.

Then play around and see what really works!!