Noise reduction for Highlight jpg

Hi, I am using RT for create JPG from RAW, some times for enhance JPG.

Currently I am working with some JPG images contains highlights and color noise. I found a problem in preview even after processing whole image, RT doesn’t remove noise. Same it is for RAW where are big highlights.

There are some screens.

  1. Disable Noise reduction - same preview

  2. Enable Noise reduction - same preview

  3. Enable Noise reduction - preview move to highlight and doesn’t remove noise from preview

When I export image color noise is there.

Can somebody know how set Noise reduction to remove noise, or some possibility how to fix source code.

Thank you

Actually, if you compare your images closely, you can see that noise is reduced when the noise reduction tool is turned on. There’s no problem with RT here, the only problem is that you try to pull too much lightness from the shadows, which is bringing the noise up, naturally. You can remove some more noise in Luminance, but you’ll never get noiseless but usable image.

http://rawpedia.rawtherapee.com/Noise_Reduction

It would help if you could upload your file :wink:

e.g. on this test image RawTherapee Noise reduction doesn’t work when process full image. Noise is still remain especially on the table edge.

1 Like

I suspect that your problem is understanding how to use the various adjustments but as some one mentioned it would help if you put the complete file raw or jpg some where to be downloaded.

John

In your first examples, noise reduction is clearly working but possibly not how you expect it to work.

If you look at your settings, I noticed that you have the chrominance slider set to “Manual” and cranked up to 100, which might make sense as the image definitely shows improved chrominance noise reduction. But you have the “Luminance” slider set to “0”, effectively removing any luminance NR. This is exactly what I would expect to see in this case.

If you want to reduce the luminance noise as well (the ‘graininess’ you are seeing), you’ll need to modify the Luminance NR value to something other than 0.

Aggressively applying for both chrominance and luminance:

It is hard to say without having the full size file, but I guess, he means that noise reduction in preview ‘works’ when there are not much bright areas in preview and it doesn’t work in full processing (because there are a lot of bright areas in the image). This may indeed be true because the way rt noise reduction works.

I’m also a bit concerned that with chroma NR set to 100, there is still chroma noise. @spex please upload the original raw file using https://filebin.net/

Please try this

Hi all,

There is new test RAW image CR2:
https://filebin.net/n6dusb32cfnn46mm/0_IMG_0011.CR2

@patdavid:
Luminance noise is not important for me.

@ Morgan_Hardwood:
It is exactly what I meen. Chrominance denoise set to full and no effect show on export (preview).

@heckflosse:
This is best result on export. But in preview it still working strange, even for
Chrominance : “Preview multi-zones”.

When I have preview filled with zoomed “Lungo box” (from test Raw image) and only little white color is in preview, denoise work fine for “Preview multi-zones” even “Manual: Chrominance - Master 15.0”.
When move “Lungo box” up in preview and preview filled half to half Lungo box/white color (below Lungo box). None of Chrominance method remove color noise from preview.

Steps to reproduce
1] RawTherapee Processing Profiles set to Neutral
2] Set Exposure +4EV to show color noise
3] Enable Noise reduction
4] Changing only Chrominance method and moving image in preview window

My problem “Export image to jpg without noise” resolved with Processing Profiles Neutral and Noise Reduction choice “Chrominace Method: Auto multi-zones”. Thank @heckflosse

Thank you all

1 Like

Ping @jdc

Here is a prototype to fix the issue. But I didn’t check for side effects. The idea is to exclude clipped values from median calculation in denoise (not the median you see in denoise gui, but an internal median calculation). I know it’s not the final solution because, though it excludes clipped values, it also excludes grey values and also values where only one part of a or b (from Lab) is close to zero. But it works fine with the example picture from above and maybe is a hint for the right direction…

Please follow this