What does this have to do with the sensor though? Currently digital sensors are made up of an array of light gathering pixels with each individual pixel measuring the light in a tiny area. If you consider the image below you will see that the pixel only records the average light for its whole area. It does not differentiate between detail within the pixel itself. Thus, in the example given the pixel recording both black and white tonal values will render them as a single grey. To further exacerbate the problem the sensor actually sees in black and white. A layer of filters (red, green and blue) actually lies on top of the light measuring pixels. This layer of filters arranged in the Bayers Pattern only records one colour per pixel. Advanced computing software within the camera then meshes all the information together using a process called Bayer Interpolation to form a colour image. The problem is compounded by another attribute of most sensors, and that is the anti-alias filter.

Due to the nature of the sensor recording light in an array pattern it is possible to come across diagonal lines with stepped rather than smooth edges. In photography these lines are sometimes referred to as jaggies, and I have once heard them called shark’s teeth. Here is how the camera would see a line in relation to how our eye would see a line. To combat this effect the anti alias filter is placed above Bayer filter, essentially defocusing the image ever so slightly so as to smooth out the effect known as aliasing. It is actually possible to have this filter removed by a technician with some cameras. The result is an image with much greater acutance, but with the tendency to alias. Aliasing is a problem that occurswith the reconstructed image from the bayers pattern where colour artefacts, purple fringing and moiré occur. It essentially means that the camera’s computing cannot properly match the spatially different colours in the bayers pattern through interpolation.

Due to the tendency of the camera to lower edge sharpness of an image, it becomes necessary in post-processing to sharpen the image as a FINAL step in the post-processing workflow. It is important that this be done LAST as further processing will tend to degrade the image.

The three sliders interact with one another so it is possible to obtain similar results using different settings (it is quite likely that different authors on photography will recommend different settings or ways in obtaining the settings. It is important for each individual photographer to work through different suggestions and come to a workflow that they are comfortable with). There is no golden setting that can be used for all images (if there were we wouldn’t need the Unsharp Mask!). Images with more or less detail, greater or fewer megapixels, larger or smaller sizes as well as the amount of interpolation all mean that sharpening should really be done on an image by image basis.

There is a small cheat though, one which some stock photographers use when preparing images for stock libraries. A great many libraries require images to be submitted at a size of 48MB (this is just a little smaller than the file size of a 35mm transparency scanned at 4000dpi). The recommendation by some of these libraries, if they require sharpening at all, is to have a setting of roughly 95-100% on a radius of 0.8-1. Remember though, this is sharpening for a very specific output. Sharpening for prints, the web or even final digital files can be very different. The reason the above settings are suggested is because they have a very slight effect on the image, meaning that further post-processing can still take place by a client if necessary (without compromising the quality of the image).

Putting theory to practice (This requires an understanding of layers - see tutorial)

Whenever you sharpen an image make sure that the sharpening is taking place on a new layer copy (new merged layer if there are more than 1 layer below). Hit ‘SHIFT + Ctrl + ALT + E’ to merge the layers into a new separate layer at the top of the layer stack. You can rename this layer sharpening so that it is easy to find.

I have already mentioned the ‘cheat’ used by some photographers. This is not the best way to get the optimum quality out of an image though. One technique I have found that seems to work in getting your settings right is to slide the amount slider right across to the right and then slowly slide the radius across until edges seem to be ‘damaged’ by the USM. The sharpening will still be too severe (for a natural looking image) so the next step is to reduce to the amount until the image looks crisp. Once you have your sharpened image you can toggle back and forth between the sharpened layer and the layers below to see the difference that the sharpening has made.

Note that the sharpening amount has been left to 500%. After obtaining the radius move the amount down until you are happy with the final 'look'. For this image I would use a radius of 0.9 or 1 for sharpening a full sized image from a 12 MP camera. I'll adjust the amount closer to 100% for my final sharpening.