Sharpening, what, when and how?
First of all lets us get the term right. When we are talking about "sharpening" we are not talking about repairing an out of focus photo. An out of focus photo is unsharp, and will stay unsharp for ever. Whatever program you use, whatever you do.
What we mean by sharpening is correcting the output from the camera or scanner.
By digitising photo's details get lost, no matter the amount of megapixels you shoot/scan, especially at the places where colors merge. Het digitisingprocess in the camera (scanner as well, but I am going to stop mentioning that from now on) translates the image that reaches the CCD into red, green and blue information (pixels). For each colout there are 12 bits available, meaning 2^12=4096 colorgradations of the colors blue, green and red. In real life there are many more colors, so some loss has already occurred, and we only just took the picture.
Next is the resolution, a CCD that can register i.e. 2048x1536 pixels can not register details smaller than 1 pixels. This means that at edges will become unclear. See the picture below, 1a is reality, 1b shows the edges as your camera sees it.
In other words when digitising information is lost. The same happens when you resize the photo to a smaller size, which actually is nothing but a reduction of pixels. Details are lost, and edges lose definition.
When?
A lot of camera's compensate with a built-in sharpening function. Often you can set the amount of sharpening the camera does, but I think it is better to keep complete control, and do it yourself. That means shooting in RAW, and using Photoshop as a darkroom. RAW is a dump of the data (colors) that reached the CCD and is mostly saved without any postprocessing. So we have to do some sharpening when we get the photo into our "darkroom".
(You may be shooting film and scanning it, or digital but without the option of shooting RAW, but the principles stay the same so I'll act as if you are all shooting RAW.)
Sharpening step 1.
In this step we are going to sharpen the egdes only. In photo's taken at higher ISO values there is a certain amount of grain, which we do not want to sharpen.
Open the photo in Photoshop. Duplicate the layer by pulling the layer over the "New layer" button (See arrow).
A new layer called "Background copy" appears in the box, select it, and make sure Blend is set at Normal, and Opacity and Fill are set at 100%.
Now we are going to select the parts we do not want sharpened. Press CTRL-A (Everything) and the CTRL-C to copy it to memory. Now open the Channels palet (Next to layers) and create a new one by pressing the "New channel" symbol. That is the second one in the bottom, the square with the circle in it. A channel called Alpha 1 (CRTL+4) appears.
Make sure this is the only one selected. Press CTRL-V to paste the selection as a channel. We now have a new channel with all the data from the photo in black & white. The other colors are still there, don't worry!
Now we are going to make the edges visible with a filter, go to the menubar, click on Filter, Stylize and then Find Edges.
Looks a bit weird like below, but this is what you need.
Now we are going to change the levels a bit. Open the levelsscreen by pressing CTRL-L (Input levels) and change the values to 130 1,10 and 235.
Makes quite a difference doesn't it?
Now we are going to filter it a bit more, go to the menubar, filter, Blur, Gaussian blur. Put the radius at 1,5.
Now we are going to do levels again, so press CTRL-L again, and enter the values 90 1 and 250.
Now we have to load this channel (that is all it is, it is not a photo). That is possible in two ways, by clicking on the first little icon in the Channelsmenu, it is the first one in the bottom, the circel, or the faster way. The faster way is by holding the CTRL key while pressing the left mousebutton on our new channel (The one called Alpha 1). You now see the "walking ants" of selection. Invert the selection by pressing CTRL-SHIFT-I. Depending on the photo you now have less or more ants crawling around.
Now that we have selected the edges only of our photo, let's get back to the photo in color, press CTRL-~ or click on the RGB-Channel.
It looks strange, but now you can really see that we selected the edges only.
Now select the copy we made of the background, that is what we made it for. From the menubar select Filter, Sharpen, Unsharp Mask. (Known as USM.)
The values are different per resolution. For a 6 megapixelphoto (app. 2600x1800) I have found an amount of 180% and radius of 0,4 appropriate. For publishing on the Web 200%, radius 0,3 and threshold 5% are better. In the preview (zoomable) you can see that only the edges have been corrected, and not the grain.
You can switch the layers on and off in the layerpalette, just click on the eye (First icon you see per layer) to see the difference.
This was the more correct way of sharpening, but for the quick and dirty there is a faster way, with acceptable results:
The quick and dirty way
Open your photo, and change the mode from RGB to LAB-color.
In the Channels palette you now have 4 channels, switch to the Lightness channel (CTRL+1), and then apply the USM as above. You may want to play with the values a bit.
In this case you are sharpening the light bits, which are often the edges. Do not forget to switch back to RGB-mode when you are done, if you don't you will not be able to save it as JPG, and some other options will be grayed out.
When? Part II
You just did the sharpening upon opening the photo.
You need to do some sharpening for your output as well as a last step.
Different output, different amounts.
The Web has lightly been touched on already.
For printing you need to overdo the edges as your printer will be losing definition in exactly the same way your camera does when taking the photo.
I will write something about the output sharpening later.