Fireflies and bump

[Cinemike]

Win 10, S22, V6 daily from May 29

I got this issue in one of my recent files and I am no sure how to remedy it. Actually I do, if I switch off bump the issue goes away - but first things first.
I have a few leaves in this scene (the rest had to go) and just one of it is infested with a nasty swarm of fireflies.
If I lower (very much) the bump value the little beasts go away. They come without and (of course) with DOF.
The final noise in the images is quite a lot below 1%.
What would be the correct step to decontaminate the leaf without sacrificing the bump (it is not needed here, I know, I am just interested in the general solution to the problem)?

TIA
Michael

[Cinemike]

PS
Is using anisotrophy meant to change the color of the problem leaf?

PS
I admit it was a dumb question ^^

[TomG]

The bump here is so strong, it is causing small areas to catch the light from a single point light source (a sun, at a guess). This happens in real life too, to some extent, e.g. sharp points of light from grass fields or leaves, especially if shiny and wet.

Things you could do
- Reduce the bump strength, it does look too strong on the diffuse areas too
- Make the leaves more glossy so reflections and highlights on them are not as small, sharp, and intense
- Use some bloom and glare to soften those points of light
- Use a larger sun for larger points of light

And so on.

[Cinemike]

Hello Tom,

thx for your reply.
Bloom wasn't of much help here, to my surprise I must say. Before it showed an effect on the bright spots the rest of the image vanished in a veil of light.
The sun was already at 5, going up to 20 finally actually helped, but made the scene shadows even softer, of course.
The glossiness I had to increase to 0.99 to get some effect.
To my surprise, too, the denoiser did not have any impact on that area.
So, in the end, just getting rid of too much bump was the only solution (the one I had already used in the actual image).

I had hoped for some magic setting in the render settings I had missed, but obviously the bump value was the critical one here (i had not even seen 25% as "very much", but I was wrong) and the only one that mattered.


Thx again for looking into my testcase,
Michael