I am a sport photographer so for most of my shots I do not really worry too much about post processing. That is, until I get to the Cheerleaders. They have to look perfect. After raw conversion, I use Gimp 2.10 mainly (Wavelet Decomposition) to take care of most of my post processing. Sometimes I get that ahaa (lightbulb moment) watching you tube videos. One gentleman was going over wavelet decomposition and he was explaining it by layer. He said something that just made bells and whistles go off. When he got to talking about layer (scale 1) of the Wavelet, he said that most of the fine detail is in that layer. I thought to myself, Let me try running a few rounds of unsharp mask on that layer to see what it does.
Since that time I have been using that method to sharpen all of my photos at the end of my portrait and cheerleader workflows. I absolutely love it. The file uploaded is an example of a test I took today (on my daughter in studio).
Camera was a Nikon D7500 with a Sigma 50-100 1.8 zoom Lens. 3 stop ND Filter was on to allow me to shoot at f2.8. Alien Bee 400 with Neewee 42 inch Diffused Umbrella was the main light source.
I try to keep things as simple as possible. I may use Linux, but I am not a major computer nerd. I do not know that much about post processing. Just trying to share something simple.
This seems to be similar to highpass sharpening, or am I wrong?
Wavelet decomposing divides an image into separate layers according to their frequencies.
The highpass filter works similar, he only uses one layer and throws away the lower frequencies.
Is the only difference in using frequencies instead of wavelets?
@Ofnuts’ would be if you were to stack the finest detail layer and then adjust its opacity of the copy.
However, it would be different if you were to multiply factors to one or all scales, or if you were to manipulate the scales in other ways such as the one suggested by @254AllStar.com and @paperdigits. This is because the wavelet scales by themselves without modification add up back into the original image. If you edit any of them, they won’t fit like a glove but better than if you were just to make high-, low- and / or band-passes separately.
The duplication of the highpass layers a couple of times (as suggested by @Ofnuts) would be most similar to high-pass sharpening. Just the increase of contribution from the lowest wavelet scale (if you did a wavelet decomposition of just two levels you’d basically be at a highpass/lowpass). This is actually similar to my personal workflow.
A benefit of wavelet scales is being able to target features on mid or other levels as well. I’ve found that skin tones can be considerably improved simply by smoothing the coarsest details scale (just above the residual). It really does a wonderful job of subtly smoothing out tones overall.
