Attached Link: http://www.photoshopsupport.com/tutorials/sharpen-an-image/photo-sharpening.html

The link is to a pretty good tut for image sharpening.  This is not the everyday stuff and will take extra effort, extra time and Photoshop©.

Can you recommend settings for sharpening images with unsharp mask or the "smart sharpen" in Photoshop?  Settings for unsharp mask should do fine in other software.

A while back someone here, perhaps Onslow, recommended some settings for unsharp mask to use in web and print images.   Can anyone let us know what really works?

Thanks for any input.     TomD'Art.

Digital sharpening isn't real sharpening. In fact, it's increasing contrast between pixels. If this enhancement follows a certain line or pattern, it gives the impression of sharpening.

The main problem with standard sharpening is that the increase of contrast happens everywhere in your image. Because the algorithm ( =math behind the scene ) doesn't "know" what is on the image, it increases contrast everywhere it was designed to do so. For jpegs for example, this can be dramatic as all the artefacts are increased. Artefacts that are most visible where there are no smooth but abrupt transitions.

Which is why the sharpening should be limited to where you want it, meaning the use of masks and more specific: edge-masks. This is the method I use and prefer:

1/ duplicate your image. ( Image>Duplicate ). This frees us from all concern. You can work on this copy without worrying about your original.

2/ Image > Adjustments > Channel Mixer. Set to Monochrome (bottom-left) and drag the three sliders untill you get an image that has sufficient contrast without losing detail. If the sum of your changes adds up to more than 100% your image will lighten, if it adds up to less, it will darken. Limit both of these. Stay around 100% in total.

3/ Filter > Stylize > Find Edges . We're looking for edges, so why not use the existing command?

4/ Image > Adjusr > Invert . Because for a mask we need the pixels we want to select be white ot at least light, we need to invert the result.

5/ Filter > Noise > Median . The Median filter works in steps, always taking the median value of the pixels it works with.I t's  good to remove noise in some specific cases.  When you apply it here it'll make the edges better for the next steps. Use a value of around 2.

6/  Filter > Other > Maximum . Maximum expands the white parts of images, so here it'll thicken the line that has been prepared by  the Median filter. Use a Radius of around 4. The result is now very stair-stepped and ugly.

7/ Filter Blur > Gaussian Blur . Set to the same value as you did for the radius in Maximum, meaning around 4.

This is your mask. Save it, so you can, if needed, adjust a bit more.

8/ Now press Ctrl+A to select all, Ctrl+C to copy all  (Cmd on the mac)

9/ Go back to your original work. Copy your image to a new layer and add a layer mask.

10/ Alt+click on the icon of your layer mask in the layers panel. Your image will turn white as there is presently no mask there.

11/ Ctrl+V to paste the mask you created in there. Eventually, deselect with Ctrl+D .

12/ Now apply Unsharp Mask with a radius of some 2 and the maximium percentage ( around 500% ). Even jpg's won't suffer much.

It looks more complicated than it is. And the result if really visibly better than most, if not all, other methods. It may look like overkill, but if you don't respect your own work enough to put some time in it to make it really the best possible quality, you may as well have your images printed in the local supermarket. Nothing wrong with that, but you have to accept their standards, not yours.

Ahhh...I can't edit my text anymore. This is a bit exaggerated, no?

I'd like to add these few corrections:

in 2/ See that the image is visually pleasing as a greyscale image without losing too much contrast.

in 12/ you have (of course) got to apply the unsharp mask to the image, not to the mask. Therefore click the icon of the image so as to activate it.
You can go to 500% and a radius of 2 but of course you don't have to. What I wanted to say is that you can use extreme values that normally would produce only very bad and exaggerated results.

I tried your process and it actuallyl goes pretty fast; like you said, looks more complicated than it is.   

Using your method or others, one ingredient stands out:  Be selective in what is "sharpened" and what is not within the image at hand.  

Still, can anyone recommend differences in sharpening for web and for print?  For me, print is using development and photo paper, not an ink or dye printer. ( I use an online service for this. )

For the web I wouldn't sharpen too much, unless you're uploading png. Jpg increases very fast in size when you sharpen because there are more abrupt transitions. One of the advantages of the method I described is that you get large areas that aren't sharpened at all, meaning file size stays under control.
But as said, I wouldn't sharpen too much for the web. UNless you don't have to care about people with slower connections.

As for film, you can do a search for unsharp mask. It does find its roots in film days. But I never used it. What I do remember is that some developers increase contrast, and that films with a lower iso are sharper (and less grainy) than those with a higher sensitivity. But I guess you already knew that.

Is the reason for the gaussian blur to soften the edges a bit, almost feathering the mask?  If not this, then what perhaps?

Tom, thanks for posting this.  Very useful & informative!

In a way, yes. When you apply the maximum filter, all whites get bigger. Because Median samples in  squares, you get lots of white squares. Therefore you apply Gaussian Blur with the same radius as Maximum. This way you get thick and soft area-lines.
This kind of mask may not be sharp. The intention is to get white where the outlines/edges are and black in the large areas. The transition must be soft yet edgy, and certainly not a series of squares.

Thanks, this makes perfect sense.  My trials with your method are successful.  thanks very much.  Tom