Heads up brick lovers - Noise reduction


(nosle) #1

I frequently photograph brick walls for non testing purposes and have noticed that the RT Detail tab Noise Reduction is quite aggressive using the standard settings (Quality: High) I havent hesitated switching it on for other subjects as it magically removes chroma noise withough any issue.

See below for what happens to brick walls. Second frame shows nr enabled.

This might only be an issue for the tiny group of photographers that are fans of swedish brick brutalism but I reccon any photo with bricks in it might get squashed. The algo seems convinced that bricks should not vary in colour that much, a bit of an architecture critic some might say…


(Mica) #2

Have you tried fiddling the settings or tried the other noise reduction tool?


(nosle) #3

Yes I gone manual but I’m not in control yet. It’s probably very doable. This post is more of a heads up that nr can gobble up the sort of variation a brick wall produces. Been a bit sloppy with the chroma nr as I hadn’t seen any adverse effects on most of my photos. Just something to look out for when turning nr on.


#4

Without the org. image and the processing profile (.pp3) it’s not possible to help you.
Can upload these files, please?


(nosle) #5

TooWaBoo and anyone else interested can download a dng and pp3 from https://filebin.net/srfkjp2rnfiyis9d

That’s not a proper development pp3 just a basic one with the results of nr showing. I’d be interested to know if the brick result is as unusual as I imagine or if i’ve just failed to see it in non brick phototgraphs.

The effect is best seen on the brick pier at the center of the image.


(Alberto) #6

fwiw, I almost never use auto chroma with the “high” quality setting, as I find it way too aggressive. I much prefer STD quality and live with a bit more noise. however, you can mitigate a bit by playing with the chroma curve and/or the gamma – at least that is my experience


#7

Here is my try:
IMGP1317_1.DNG.pp3 (9,9 KB)

Like @agriggio mentioned I also never use “High quality” and I always use “False color suppression steps” in the Demosaicing tool over “Chrominance noise reduction”. Avoid values higher then 5 for the Chromonance - Master.


(nosle) #8

Thanks thats much better! Good tips and tricks as well. I always assumed Quality: High meant more computationally expensive but also more accurate settings. Completely random assumption but there you go.


(Morgan Hardwood) #9

https://filebin.net/g3s2sku7w9ik1jjb


(Simon Frei) #10

Sorry, I am going slightly OT, if it’s consider intrusive, tell me and I will shut up (and maybe move to a new thread).

Your assumption isn’t random at all. Firstly this is simply what quality means and secondly that’s what it says in the Rawpedia:

[…] higher quality at the cost of processing time.

@agriggio’s and @Morgan_Hardwood’s statements contradict this. Maybe a warning in the Rawpedia, that it is another tradeoff between quality and NR, not quality and processing time (which I don’t care about), would be in order?


#11

I’ve also had quite a few occasions where I disliked the “high” quality results because it desaturates muted colors too much. “Normal” quality helps somewhat, but I noticed a bigger improvement when switching to RGB instead of LAB mode for NR for these cases, although IIRC LAB is supposed to give better quality.
Maybe the LAB “high” quality algorithm needs some fine-tuning?


(Alberto) #12

Some more discussion about this in this thread: