Sharpening
ebwest
Registered Users Posts: 416 Major grins
I know there are many ways to shapen pictures in Photoshop and I've tried all that I could find. This is one (which is new to me) I thought I'd pass on. I got this via an email after asking about another method. The photographer, Manuel Librodo Jr., said to pass it on as his way of paying forward. I've tried it on several pictures so far and it has improved them all and I like the results much better than LAB sharpening which I have been using. Here's the method, let me know if you like it.
Background copy
USM 18-40-0
USM 150-.3-0
Edit>Fade USM 100% in Darken Mode
USM 150-.3-0
Edit>Fade USM 50% in Lighten Mode
Flatten image
You don't have to do the first and last step, but it's safer that way.
Background copy
USM 18-40-0
USM 150-.3-0
Edit>Fade USM 100% in Darken Mode
USM 150-.3-0
Edit>Fade USM 50% in Lighten Mode
Flatten image
You don't have to do the first and last step, but it's safer that way.
0
Comments
Wouldn't you still see some improvements from doing it on the L channel in LAB mode?
http://ebinfl.smugmug.com/
No difference that I could see between RGB and L*A*B. Nice tip by the way.
Anthony.
Sharpening in RGB can shift colors in an image, whereas sharpening only the L channel in LAB doesn't--the L channel doesn't contain any color information.
Admittedly it'd usually be pretty subtle, but I don't see any reason not to avoid it anyway.
Actually there are two different techniques offered here; that they both use the USM filter has been a cause for confusion from time to time.
The initial settings are a process that Dan Margulis calls HIRALOAM for High Radius, Low Amount, and, truth be told, it's not a sharpening move at all. Elsewhere you can find it referred to as Localized Contrast or some such, and it does a great job of enhancing contrast between items that are next to each other, without the blown highlight and plugged shadows that curevs will threaten. It's useful to think of this step as something separate from the sharpening process. I will sometimes insert it into my work flow before any color moves, and then shapren at the end. This thread contains a good description of the procedure (and why a fixed set of numbers won't work):
http://www.dgrin.com/showthread.php?t=22524
As for sharpening, here's a technique that I've used to great effect, that eliminates the worry about what the "right" values are for the various parameters in the USM dialogue. It's based on the same theory found in the above approach, but it's much more flexible and can be customized for the needs of specific images. Like any formula approach, it's not going to give the best results for any image, but, for someone uncertain about the values that work, it's a great way to get started, producing useful results every time.
I always dupe the image and convert to LAB, so I can run my sharpening moves on the Lightness channel. There is some disagreement with this from various quarters; suit yourself, it's not a crucial thing. But the way I sharpen, the lightness channel is safer. If you work in lab, select the lightness channel in the Channels palette, then toggle the eye icon back on for the a and b channels. This way you're only working on the Lightness channel but you can see the result in the actual image.
Step 1: Use a low radius and a high amount. I start with .9 pixels for the radius and 200 for the amount. This amount needs to be steadily increased, depending on the resolution of the image. The goal is to increase it until the halos that are characteristic of the Unsharp masking process become painfully evident. This will make the image look like crap. At this point, that's the intention. The halos will look like they are all white; black halos are there as well, but they aren't as noticable. The reason I like LAB is that the drastic color shifts aparent with RGB won't happen here.
If you've reached 300 and the image still looks halfway decent, increase the radius up to around 1.5 max. That should be enough to blast out anything.
To repeat: the image will look grossly oversharpened at this point, but the halos should be small enough that they won't really obliterate detail, they'll just call way too much attention to themselves.
Step 2: If in lab, convert back to RGB
Step 3: Drag the sharpened image back onto a new layer in the original. Dupe this layer.
Step 4. You will now have three layers in your file: the original, and two identical overshaprened layers. Set one of these layers to darken mode. Set the other to lighten mode. Reduce the opacity of your lighten layer to 50%.
Step 5. Group the two sharpened layers into a layer set and reduce the opacity to taste. Usually somewhere between 40% and 70% will give a good result.
Practice and experience will allow you to target ranges more precisely for specific images, but the process is guaranteed to absorb variations without affecting the overall result.
—Korzybski
Thanks for taking the time to post your technique. I tried it on one image and may have noticed a very slight improvement. I will give it a try on a few more.
I have tried a variety of sharpening methods, and will continue to try new ones, but believe it or not, I have a little plug in called Focus Magic that seems to do a great job, (most of the time) with one mouse click.
Sam
Fun to play with at any rate.
Thanks for the comments.
http://ebinfl.smugmug.com/