Unsharp mask is usually the last thing I do before resizing. And more often than not it is necessary to do another unsharp mask after resizing as well. I find after resizing I don't need to unsharp at as high a setting, just a bitty bit.
Thanks Gary. Wallabies won by the way Gary ;p You should have been here ;)
I'm trying to find a link someone put up here about a week ago Coley. It had a suggested range of settings for USM. I've been using them and getting good results.
Can't find it anywhere Coley, sorry, but I'll post it if I do.
The actual values, I think, depend on the size of the pic. The radius for instance is in pixels, so a 4 pixel change, for instance, will have a greater effect on a 600x400 size pic than say a 3000x2000 one.
If I am wrong there someone please correct me, I have made this assumption on observed results and what seems to make sense, but I could be well off the mark.
Gary... curious as to how you decide what percent to use unsharp mask.
For me it kinda varies shot to shot. Is it just personal preference or what
Umm well in the past I used to use the same settings all the time until my wife asked me what the heck I'm doing :-) Now I first set the amount between 80 and 180 depending on the resolution and size of the pic. Then I adjust the threshold somewhere between 2 and 5 also depending on the resolution and size of the pic. For portraits for example I use a higher threshold to keep areas of skin softer. Then the radius I just adjust by rule of thumb. I don't use any standard settings, it's all just adjusting until I am happy. And whenever I USM a pic, I have the pic open at 100%.
Print:
1800x1350 / 1800x1200 for a 10x15cm 300dpi print
75-250% 1px radius 1 treshold for low noise images
150-350% 1px radius 2-7 treshold for noisy images
When it comes to using the full size 2832x2128 files or anything near this:
Same as the 1800x1350 pics, but with a 2px radius.
These settings depend on the camera used, a much cleaner, natively sharper dSLR file might work better with a smaller radius and low amounts. The rule is: Keep the radius as low as possible and alter the look by using the amount slider. Noisy? Up the treshold and up the amount. The amount varies per photo. I strongly discourage to do a batch USM, unless you use an action that prompts you to alter the numbers.
I sharpen as the last step, right after the resampling, but always before I alter the canvas size (when you want the 4:3 pic to be printed on 3:2 paper). I do this in lab mode.
Just make a few actions and put them under a shortcut. I've got a couple of actions under F2-F11 that prompt me with image size (default settings for certain sizes) and USM dialogs. Saves a lot of time and hand movement. Hit F2 = Resample to 1800x1200, then hit F3 = Convert to lab mode, select lightness channel, prompt with USM dialog (default is 100% 1px 1 treshold), convert to RGB mode.
I sharpen as the last step, right after the resampling, but always before I alter the canvas size (when you want the 4:3 pic to be printed on 3:2 paper). I do this in lab mode.
I've heard this often before. What are the advantages of sharpening in lab mode?
It avoids color halos around strong edges. You can apply stronger sharpening in lab mode before these halos appear. You only sharpen the details and it does not affect the colors. Since it does not affect the colors I think it is useful for any photo you like to sharpen.
Here is an *extreme* example on a small crop from an 1800x1200 file:
The first one is original, no sharpening applied.
The second was in RGB mode, USM 400% 1px 1 treshold.
The third was in Lab mode, lightness channel USM 400% 1px 1 treshold.
The second one has the halos around all the major edges, especially visible around the flagpole.
No one will sharpen a pic this extreme, but when you have for example the petal of a flower on a white or black background, it will make a difference with lower sharpening settings.
(Photo = ISO160, what a *-load of noise....)
I'll remove the examples in few days.
Do you use it BEFORE or AFTER you resize the photo for submission, or does it not matter?
I alway do it after the resize; resizing softens the image.
However, you may want to apply it several times, in stages, with different settings. I always experiment. On another site, someone shared a technique with us, which amounts to applying a USM with a large radius, which increases the contrast, before resizing, and then a smaller, normal level USM (say amount 39%, Radius 1.0 or less, Threshold 4) after you resize. In fact, his recommendation was to do it in four steps, based on the idea that the sharpening "prepares" the image for the reduction. But you have to start with a pretty large image to use four steps.
My personal experience is that I always apply USM to a duplicate layer, and then blend it in luminosity only mode. This has the same effect as using lab mode as mentioned below, but I only have PS Elements, which doesn't have that mode. Doing it in a layer gives you the advantage that you can then adjust the effect using the transparency of that layer. Normally, as well, and this will sound weird, I often add another layer, with a filter which serves to soften the non-edge areas of the image. Again, I adjust the effect by changing the transparency.
Unfortunately, I don't think my usual method is DPC legal, because it says you have to apply all layers in Normal mode. That's kind of weird in itself, because if I had written a program to do what my layering work does and called it a USM filter, it would be legal.
In general, be careful of USM. While many people will say, "needs to be sharper" in comments, IMHO sharpening often ruins the aesthetics of a picture, in the non-edge detail. I've seen so many pictures of beautiful autumn scenes, where they are sharp as a tack, and just about as pretty. ;)
Print:
1800x1350 / 1800x1200 for a 10x15cm 300dpi print
75-250% 1px radius 1 treshold for low noise images
150-350% 1px radius 2-7 treshold for noisy images
The upper end of your percentage range looks huge to me. I set the built-in camera sharpening to high on my D100 and have only once or twice gone as far as about 150%. In 90% of the cases I don't need any more than 90%, and in many cases its closer to 50% (that's with radius 1 pixel and threshold set to 4).
The reason I'm pointing this all out is that sharpening isn't only image dependent, but appears to be dependent on the camera as well.
You are right Ronners, I had already pointed that out in my original post.
The reason why my amount is high compared to yours, is that I don't do any in-camera sharpening. The S602Z already oversharpens in normal sharpening mode. It is agressive, it sharpens all the noise, it creates halos and to make things worse, it compresses files a bit too much at fine jpeg = compression artefacts at some of the sharpening edges. So I shoot in soft and sharpen later.
I usually use settings a little lower than others have cited. I find oversharpened, haloed images more bothersome than a little softness. I also typically prefer to apply 2 "lighter" passes with USM than one with higher settings in order to provde more control over the process.
It is correct that the amount of sharpening needs to be less on a smaller image. That's because each element is made up of fewer pixels. A rope which is 8 pixels thick on a print image might only be 2 pixels wide on a web image; apply a 1.5 pixel sharpening effect and you are affecting the whole element, not just the edges.
I have some examples of sharpening (and over sharpening) in this pBase gallery.
My average settings for DPC-sized images are
66-88% * 0.6-0.8 dia * 5-7 threshold
You should always at least check for needed sharpening after re-sizing, as the re-sampling process by definition is probably going to reduce detail. Resampling is an averaging process which will tend to bring pixel values closer together, whereas the sharpening is somewhat the opposite, in that it finds pixels which are different and makes them more different.
This page describes why there is a need for two sharpening stages in the process: Understanding Sharpening.
The first pass is to compensate for detail blurred by the low-pass filter in the camera, the second to compensate for detail softened by the output process. It also gives some basic guidlines for when in the process to sharpen, and what settings to use.
This process works pretty well for me, but not always. (As with most "guidelines", there are instances where they need tweeking.) I'm going to step through some of the processes outlined above and see how they work in comparison.
Added second question: anyone using PSP know the equivalent of lab mode in Paint Shop Pro? Or, is there an equivalent?
I don't know a lot about it, but here's the help text from PS:
In Photoshop, Lab mode (the asterisks are dropped from the name) has a lightness component (L) that can range from 0 to 100. In the color picker, the a component (green-red axis) and the b component (blue-yellow axis) can range from +128 to -128. In the Color palette, the a component and the b component can range from +120 to -120.
You can use Lab mode to work with Photo CD images, edit the luminance and the color values in an image independently, move images between systems, and print to PostScript Level 2 and Level 3 printers. To print Lab images to other color PostScript devices, convert to CMYK first.
Lab color is the intermediate color model Photoshop uses when converting from one color mode to another.
I do know that it's supposed to be a better mode to work in for sharpening as well as working with curves (at least, based on the tutorial on curves I've used most) - I think something to do with working with luminosity as a whole instead of on seperate color channels. But I may be totally off track here.
Anyway, hope that at least helps a little. Oh, and sorry, I don't know about the equivilent in PSP. :(
In the link ScottK posted it says to be careful you only sharpen twice, not a thrid time, so be careful the camera doesn't sharpen as well.
However .....
When I look at the RAW file in Photoshop CS I notice there is a sharpen option, before you export the RAW to Photoshop. The default value for this is 25.
Do people think this should be left at 25, or set to 0 and the sharpening be done with USM later (which I am tempted to do).
