High ISO Noise Help....

This picture is an example, I don’t mind the noise grain, it can look better than the smooth over processed look. I don’t see how to remove the blue and red colour blotches visible on the dogs nose. I’ve used Darktable for this one, but my attempt with Rawtherapee seemed to have the same issue.
I tested the same image in programs like Lightroom/Capture One and DXO Photolab and they do not show this issue? I’m still trying to find my way around Darktable so I’m probably just missing something obvious.

_DSC1373.ARW (47.0 MB)
_DSC1373.ARW.xmp (20.2 KB)

This file is licensed Creative Commons, By-Attribution, Share-Alike.

1 Like

_DSC1373.ARW.xmp (15.1 KB)

Wonderful dog portrait.
I think noise reduction works well with new denoise in darktable. I first switched to non local auto, there I set auoto adjust and central pixel weight to the highest values, then I switched to manual and tweaked preserve shadows. Needed to remove hot pixels as well.

I only applied denoising and hot pixels, and cropped to the dog’s face. Denoise (profiled) did the main work, first in non local auto, then tweaking a bit. Some color noise remained, I managed to remove it with denoise (bilateral filter) in color blend mode.


_DSC1373.ARW.xmp (8.3 KB)

I agree, DT 3.0 profile denoise works well. Thanks for sharing

_DSC1373.ARW.xmp (6.4 KB)

1 Like

Thanks, it’s good to see differing aproaches to noise reduction. Most commercial programs just give a few sliders, which is fine most of the time and easy, but sometimes having a little more is control can be a good thing when trying to balance noise and a particular image style.
This particular image was taken when it was getting quite dark in a forest at sunset. I’m surprised how good the image as a whole looks even at iso 25600.

@Jean-Marc_Digne
I have the impression that there is a kind of pinkish cast on the white fur of the dog .
Examining the image, as it is downsized and converted in jpg, I cannot decide if the strong artifacts come from the downsize/conversion or from denoising.
Downsizing permits also to average the noise.

detail of full scale jpg Processing with ART

detail of downsized jpeg 1900x1268

_DSC1373.ARW.xmp (14.5 KB) _DSC1373_01.ARW.xmp (14.9 KB)

5 Likes

Great image!

_DSC1373

I think as for @Jean-Marc_Digne, your denoising is too strong. The background is denoised but to the detriment of the black fur , mainly around the right eye, giving a plastic aspect.
Also just above the nose where the aspect of small bristle can be confused with noise.
xnviewmp_Ib0Xl0vtXq xnviewmp_976DCrEokQ

edit: on the right, detail of full size image.

Very good performance of the camera.

1 Like

Hi,
here’s my attempt, using ART. Thanks for sharing!


_DSC1373.ARW.arp (11.1 KB)

4 Likes

My take: minimalist PhotoFlow + G’MIC. Zoom and enjoy!

2 Likes

@ gaaned92
I tried to improve my denoising module test (without any other module just standard DT3.0)

100% scaling (Before / After denoise)

Thanks for your advise, I appreciate your help

1 Like

I liked @afre’s noise reduction amplification.
Seriously.

RawTherapee 5.7

1 Like

I wouldn’t expect wonders with that amount of noise. It is always a compromise between denoising and loosing details.
For the color fringing in nose and fur I used the defringe module with lowest threshold


_DSC1373.ARW.xmp (15.5 KB, dt 3.1)

I don’t deal with noise in my minimalist processing because I don’t know how to do it properly. Low frequency noise is especially challenging. Observe the colour noise of the nose below (exaggerated for the web).

image

Wow, some great examples, I’ve got a lot to learn.
I know it’s not open source, but here’s what Topaz Denoise did with this picture. Similar to the first picture I developed the raw in Darktable 3.0 but with only chroma noise reduction, no sharpening, no luma reduction. Then in Topaz with only noise reduction, no sharpening, no ‘enhance detail’ whatever that does.

3 Likes

_DSC1373.ARW.xmp (18.1 KB)


a-_DSC1373-5.jpg.out.arp (11.0 KB)

2 Likes

@sls141 (!) your bw = best of all in this thread. respect.


quickdirty RT5.7 inkl. Film Simulation ‘Ilford HP5 1-’

_DSC1373.jpg.out.pp3 (11,8 KB)