Photoshop Tutorial, Save for Web 2, Sizing, Sharpening, Saving, f8 Pro Photo Photography

The next step in the 'Save for Web' process should always be the last before actually compressing the image for the web and this is 'Sharpening'
One ought not skip through this essential operation. It is vital to the apparent image quality that you will ultimately transmit and your last chance to optimise the image before compression tries to ruin it.
CS2/3 now provide lots of different methods to Sharpen an image but the good news is you only need one. Try these various methods until you find the one that gives the result you want whilst not being over-complicated or requiring too much process time, then practice using shots of different subject matter.

1. Saved for Web
No Sharpening

This original is a sharp image but because we have resized it for the web interpolation has taken place so Sharpening is needed.
Stood alone it looks fine until you compare it with 2. Take note of the smooth Skin tone.

3. Same as 1 with
'Sharpen' applied

This one click fix Sharpens the whole image by a pre-set amount. Face and Hair are in fact Sharper but notice the Skin Tone around the Eyes and Chin. It's as if a texture were being introduced.
The Girl has aged!

4. Same as 1 with
'Sharpen More' applied

And here she looks decidedly rough. Notice that kind of halo around the back of the Head and the poxy complexion.
This is an example of over-sharpening and why we should apply more critical control with alternate methods.

Increasing the contrast between edge Pixels is all that Sharpening does. It doesn't have much effect on broad areas of uniform tone such as blue sky or an area of skin tone. Where Pixels of one tone meet Pixels of a different tone, contrast can be increased between these boundary Pixels for example, where the pixels comprising this Lady's teeth meet her lips, or where skin tone meets hair.
From the top menu bar > Filter > Sharpen and you should have a drop down box, like this screen grab, showing 5 sharpening options. (In PS Elements there may be only 4).
We'll look at the simplest first.

2. Same as 1 with
'Sharpen Edges' applied

This is a quick, one click fix, that's a vast improvement already. It concentrates on just the higher contrast edge Pixels. Compare the facial features to 1 then compare the Hair where the Lady's Ear would be.
Not much difference.

The fourth option on the Sharpen list is 'Smart Sharpen'.
Here we see some finer control parameters and a 'Preview' image to judge the effect which can be enlarged by clicking the + button.
In 'Basic' mode one can apply any 'Amount' of Sharpening by adjusting the slider and thereby achieve the same results as the above one click fixes however, in this dialogue box we also have control over the number of Pixels surrounding the boundary area, by selecting a 'Radius' quantity. A lower Pixel quantity defining a smaller group of boundary Pixels producing a finer image or a higher count that will seem to give a thicker edge to the boundary. The 'More Accurate' option seems to just exaggerate the applied effect.
Some useful options are available in the 'Remove' box where, by clicking the down arrow, the effects of Gaussian Blur (a kind of Moire effect you get when scanning litho printed dot screen images) or Lens Blur (slightly out of focus images) or Motion Blur (subject movement relative to the camera, and this gives a further option of selecting a perceived angle of movement) can be, if not removed, then somewhat reduced.

And finally we arrive at the last, and my favourite, method of Sharpening - the Unsharp Mask, which sounds like a stupid, contradictory term but has it's origin in the litho printing industry where an unsharp screen mask was applied to 4 colour separations thereby enhancing sharpness. Very similar to 'Smart Sharpen' with 'Amount' and 'Radius' but now we have an extra control feature - 'Threshold' and what this does is to allow us to determine the difference between adjacent Pixels within the boundary area of sharpening. It works in a different way than other methods in that it does not detect edges in an image. Instead, it locates Pixels that differ in value from surrounding Pixels by the 'Threshold' you specify. It then increases the contrast of neighbouring Pixels by the amount you specify. So, for neighbouring Pixels the lighter Pixels get lighter and the darker Pixels get darker. One can also specify the radius of the boundary to which each Pixel is compared. High being lower quality and Vice-Versa except at a 'Threshold' of 0

all Pixels in the image will be Sharpened and this is not desirable, resulting in image 'noise', the roughness visible in the center picture below. Bear this in mind. The same amount of Sharpening cannot be applied to all images. Factors such as image contrast, colour contrast, amount or lack of detail, creative effect, all need different Sharpening amounts. Lastly, pick one method, and stick to it.