Teton Viewer wrote:
I'm fairly new to post-processing, using Lightroom 4.4 occasionally for a year or two and only on landscapes. I've been wondering what the sharpening tool actually does, at the basic level, to make an image appear sharper. I haven't been impressed that the effect is very great on landscapes but perhaps that's because of my inexperience. Thanks.
There are generally two basic types of "sharpen" tools available with image editors. (And two or three others that are not as commonly included.) One method, commonly called "Sharpen" is a high pass filter algorithm. The other is called Unsharp Mask. Knowing the difference in what each does is important for "advanced" image editing.
You need to understand the concept of spatial frequency domain. Consider a picket fence with alternating white slats and a dark background. If the slats are thin and close together they produce a "high frequency" in the spatial domain. If they are wider and farther apart the frequency is lower. If an area of an image has no variation in tones, it has an extremely low frequency.
If there is a sequence of consecutive tonal variations that exactly match the pixel spacing on the camera's sensor, that would produce the maximum resolution possible. To see it however there has to be at least some contrast between the tones. Increasing that contrast, which is what a sharpen tool does, makes the image look sharper. It does not actually increase the resolution, just the contrast.
The way that "Sharpen" accomplishes that is to detect sequences of tonal variations that are at high frequencies by using a high pass filter. The high frequency part of the signal is amplified to make the bright area more distinct from the adjacent dark area. That is, the contrast is boosted, but only at the edges in the high frequency sequences.
The exact opposite can be done, by reducing the amplitude of the high frequency sequences, to give lower contrast. That will blur an image! In fact blur and sharpen tools use exactly the same algorithm, with different parameters!
The UnSharp Mask (USM) tool does essentially the same thing, but in a very different way that has a very distinct effect. USM works only on single edge transitions. A single edge causes a large spike of high frequency in the spatial domain. USM averages high frequencies (by using a blurred mask). It uses the difference between the image and the blurred mask to apply higher contrast. A sequence that Sharpen would detect is seen as a solid tone due to the average, but a single edge is sharpened.
So Sharpen increases the contrast for fine detail sequences while USM does the same for single edge transitions. Lets say we have that picket fence, with an open gate. Sharpen adds contrast to the pickets, but the transition from the fence to the open area of the gate is not affected. USM will affect the edge of the gate area, but not between the pickets on the fence.
Unlike Sharpen, USM cannot be reversed with a blur tool.
Generally, images produced by cameras using a Bayer Color Filter Array to encode colors will benefit from being sharpened. If the original image is resampled to a smaller size (for the web perhaps) the effects of USM are usually more significant than Sharpen. If the image is resampled to a larger size for printing, Sharpen will usually be more effective than USM. (It very much depends on the type of detail in the image, and usually both methods have an effect.)
When an image is resampled to a smaller size the process is a very effective low pass filter in the spatial domain. Hence any sharpening done before resampling is virtually eliminated by resizing! Therefore it is always best to sharpen as the last step, after a resample operation to set the size.