5

I was just editing some tit photos I made on our balcony, and I found really strange patterns of noise over the bird's feathers:

enter image description here

The photo is from Nikon D5600, settings 85mm F4.0 1/1000 ISO 160

These are individual pixels exported from RAW and even if I went all the way to the beginning of my edit history, these patterns are still there. Is this an error in demosaicing, or is it a feature of the photo? I can provide the RAW file but I am not sure where to upload it.

I am using darktable, if it's software related, it the cause might be there. I would like to get rid of these.

RAW file: https://drive.google.com/file/d/1gDMyE5UaAvhvcFhv0qoRbf9ada6XtHvK/view?usp=sharing

Tomáš Zato
  • 1,050
  • 1
  • 11
  • 25
  • You could put the original RAW up on something like Dropbox etc & just post the link here. I'd definitely like to see it for myself. – Tetsujin Sep 14 '20 at 17:43
  • 2
    darktable does have several different demosaic algorithms - have you tried them? Some may do better at minimizing/eliminating it than others. But, otherwise, it looks like a Moire interference pattern, which happens with fine textures that are near some small multiple of the sensor element size. Definitely would be easier to see with a full-size raw image, though, rather than a scaled down and compressed jpeg.... Also may be exacerbated by what appears to be higher-ISO noise on the background, but that could be other noise sources as well.... – twalberg Sep 14 '20 at 17:46
  • @twalberg It sounds a lot like you're right. I uploaded the RAW on google drive, so you can check and confirm your suspicions :) – Tomáš Zato Sep 14 '20 at 18:27
  • 1
    It may be my system but when i click on the link i get googles spinning circle of eternal limbo. No file loads. And i so wanted to see the RAW tit. – Alaska Man Sep 14 '20 at 18:30
  • @AlaskaMan I cannot reproduce. At least one person managed to download it. I might try to think about better upload alternative for the future. – Tomáš Zato Sep 14 '20 at 19:43
  • I was able to download the NEF. In order to reproduce your image, I had to over sharpen and over saturate in Darktable. – user10216038 Sep 14 '20 at 22:35
  • @user10216038 The example above is over sharpened and over saturated deliberately, but I could see it on the original image too. I checked and it only happens with PPG demosaic algorithm. – Tomáš Zato Sep 14 '20 at 22:41

4 Answers4

3

Using Darktable 3.2.1 I get this jpg with default settings. enter image description here

Then, I get this jpg with VNG4 demosaicing algorithm, default "sharpen" option and automatic exposure correction.

enter image description here

I don't believe Darktable is to blame here. Maybe you are pushing too much some settings, since histogram shows an underexposed photo:

enter image description here

Edit: Playing with 'L' in Lab color space I got a brighter image, without messing with the colors. 200% crop: enter image description here

vsis
  • 1,251
  • 4
  • 18
2

This is the RAW image cropped but otherwise untouched, saved from Nikon's own ViewNX-i as a 100% quality jpg.

It shows some hint of interference, as we're right down at pixel-peeping, but not the extra 'crunch' darktable seems to have added. The colour fringing is vastly reduced too.
It has, of course, had the snot cropped out of it - it's down to 493x362 out of a 6000x4000 original.
I'd really suggest you need a longer lens for birding ;)

enter image description here

I'm going to go back to my usual stance on this - don't use 3rd party software for initial processing of Nikon camera images.
Compared to all the other Darktable/Lightroom examples so far - this is the only one that hasn't just softened all the detail out of it.

Tetsujin
  • 23,252
  • 3
  • 46
  • 97
2

Based on the comments on the question and some answers, I figured out the issue. The issue can be removed by enabling Match Greens: local average. If you also use AMaZE demosaic algorithm, the artifacts are completely gone:

enter image description here

Tomáš Zato
  • 1,050
  • 1
  • 11
  • 25
  • Which is exactly what the answers to the duplicate from over three years ago said... – Michael C Sep 15 '20 at 03:36
  • It's still at the cost pf a lot of detail compared to just using Nikon's own software, as are all the other Darktable, Lightroom tests. – Tetsujin Sep 15 '20 at 06:19
1

This is the result of Lightroom, default settings +1 Exposure:

enter image description here

Looks much more reasonable to me, with a little but of colour fringing which could be removed easily.

Conclusion: the demosaicing algorithm of Darktable has trouble with the fine details in the feathers.

ths
  • 7,161
  • 1
  • 11
  • 25
  • 2
    One of the multiple demosaicing algorithms that are user selectable in Darktable has trouble with the fine detail in the feathers. – Michael C Sep 15 '20 at 03:35
  • i'm not familiar with darktable, but the other answer demonstrating different algorithms shows pretty terrible results as well. – ths Sep 15 '20 at 11:59