Request for test

(Ingo Weyrich) #1

In current microcontrast branch I intoduced a new adjuster which allows to limit the effect of Microcontrast and Sharpening (especially useful for RL-Sharpening) to regions which exceed the amount of contrast set by the adjuster.
My intention for this new adjuster was to prevent regions with low contrast from sharpening (for example blurred regions with demosaich artifacts have been affected by RL-Sharpening and showed artifacts).

I would like to get some more tests for this new feature (which I think is really useful). @gaaned92 already included the microcontrast branch in his nightly windows buils (thanks for that :slight_smile: )

For reference, please have a look at the github issue

I already tested with a lot of raw files to find good default settings. Here are first results

(Sebastien Guyader) #2

@heckflosse Ingo I tested the branch a couple days ago, but not very thourughly. The only thing I can say is that generally I avoided using RL sharpening because I found that it introduced too much unnatural artifacts everywhere in my images? At least now I find it more usable. I have to test more, but so far so good.

(Ingo Weyrich) #3

@sguyader Sebastien, I can confirm that using RL sharpening on xtrans-files is not as good as on bayer-files unfortunately :frowning:

(Sebastien Guyader) #4

Well, it’s not a big issue, as I think I’m spoiled with great Fuji lenses and the 24MP sensor, I generally don’t feel the need to sharpen at all if I want to keep a “natural” look. I use sharpening only if I really want to emphasize detail or texture in some more specific shots.

#5

So I’ve tested a little the new contrast treshold and I think it’s very useful!

I’m a big fan of the old unsharp mask and for what I can say the rawtherapee unsharp mask is “optimized” for a radius less than 1 pixel.

In the end I find these settings a little better:
Contrast treshold=1 (yes only 1!!)
Radius=0.5
Top and bottom left treshold=0
Top and bottom right treshold=2000

For the RL deconvolution contrast treshold= 20 is pretty good.

(Ingo Weyrich) #6

Did you try unsharp mask with radius between 0.4 and 0.45 and amount between 800 and 1000? Imho that also gives good results for fine details.

#7

Even better :slightly_smiling_face:

(Gimbal Lock) #8

Sharpening of images are very image dependent so it’s hard to come up with default settings. Pixelshifted images where the aim has been sharpness and details can take a huge amount of sharpening (and now even more thanks to the adjuster). But soft images with lots of out of focus areas just ends up looking weird. So I believe the default values should still be pretty conservative.

I too like the USM better than RL deconvolution since RL seems to produce white dots/single pixels pretty quick. At least in some pictures, maybe I’m not using it right.