Please stop capitalizing “not”.
<deleted by author 7/9/2024>
1. Not (*not*) is the preferred style for emphasis, but a single capitalized Not is hardly shouting.
2. USM and other filters can be implemented differently among software algorithmically and parametrically. Do not sweat on the details. Use parameters that work for you.
Technical notes:
The smallest radius is usually 1, giving the filter a neighbourhood or window size of 3x3 pixels (r*2+1). The centre pixel is the one being acted on. For USM, the sharpened image is the difference between each pixel and the average of all 9 pixels from the 3x3 window (then normalized or cut in some manner to match the original image value range).
Amount is the number that represents the degree in which the filter works, often called regularization, or simply amount to keep it simple.
Developers set parameters to reflect the purpose of the filter, so GIMP devs figure that 1500 radius seems like a good limit, whereas RT dev consider 200 suitable. It does not mean that more is not meaningful to the user. Ditto regarding the amount.
The parametric degree and range of the amount is also arbitrarily set. RT’s ‘1.0’ could be more than GIMP’s ‘300’. It is just normalized to be [0,1]. To me, what is important is the granularity in setting the amount. As a hypothetical extreme, imagine we were given the binary choice of 0 and 1. Now, that would not be useful at all, giving use off and on. Two or three decimal points would be practical. RT devs are known to expand the slider ranges on request if it is reasonable to do so.
Application:
Conventional USM is context unaware as evident in the simple averaging and subtraction, causing artifacts such as halos of edges and over-texturizing of noise and extreme neighbourhood differences. Consider applying it judiciously, e.g., applying it to lightness channel only (L*a*b*) and only a masked portion to make it more tone and context aware.
I forgot to address standard deviation, which is the parameter for Gaussian blurring. I recall some versions of GIMP USM using standard deviation as the parameter instead of radius. This means that it is using Gaussian blur as part of the USM algorithm instead of the simple averaging I gave as an example in my previous post. Although it can be confusing for some people, exposing the SD parameter is just being truthful about what is driving the filter. I am not sure about the tool tip for radius though. Looks like it may be an artifact of a former iteration of the filter.
As with radius, move the slider up and down incrementally to see its effect. The effect is not linearly related to radius, but I suggest you benchmark radius ↔ SD equivalents for small to medium sized blurring.
In general, when you make changes in detail and contrast, make sure you are in 100% zoom and your monitor and graphics configuration is in their native resolution and compatible setup to prevent artificial blurring or distortion that may lower the accuracy of what you see on the screen. Make sure your monitor and workflow is colour calibrated.
no comment
Alan, thank you. Your explanation does elaborate on the documentation that I’ve been able to find.
I’m using a fairly good high definition (27") monitor that has been color calibrated.
It might be worth pointing out that my photography is mostly landscape. What I might refer to as water-scape. Reflections from interesting cloud formations is one of the things that tend to be what make such pictures interesting. You can find some examples here. Of course those are low resolution .jpg files intended for display devices. My real goal is maximum resolution lossless files intended for large format printing.
When it comes to zoom factor I have been zooming in when the goal is sharpening for which I pretty much always use the Unsharp Mask filter. However, when I apply Local Contrast it is the overall affect on the whole image that causes me to always try it at the end of the pipeline when done with everything else just to see if it helps which it very often does. Now it is true that there are likely certain portions of the image where the affect really matters and it does make sense to look closer at that when making adjustments. However, the result on the whole image is still what matters. I need preview to work on the whole image.
As I mentioned earlier the Local Contrast Enhancement tool provided by the GMIC Plugin works very much like the other Local Contrast enhancement tools I’ve become accustomed to using. But GMIC no longer allows preview of the entire image. With their zoom factor having to be 100% or more I can only view a very small portion (maybe 10%) of my 50+MP photos. For me, this makes it unusable.