To all RAW image processing devs: When implementing custom White Balance pickers do allow much larger sampling windows, hundreds or even thousands of pixels. All images have noise that messes it up and you have to average over much larger areas. 32 pixels is nothing @RawTherapee
No. That's a RAW image (from Sony A9 III). On the right is the cranked up saturation to visualize the noise splotches. And it makes it obvious that any averaging window must be much larger than the characteristic spatial size of a noise blob.
Interesting what they are doing in camera. Raw sensor noise don’t have a color clusters
Yes. All of them do. Not on purpose of course. But every single RAW image I've ever seen has this. That shot though is after median denoise (left) so the actual RAW color noise if finer grain (right) but still has features going way beyond 32 pixels in size.
You can kinda see all that in the raw RAW even without debayering. That's a "perfectly" flat gray ball.
Hard to guess but this is how raw sensor data looks after basic demosaic. Maybe RawTherapy uses a strong color denoising, subsampling? Photone noise can't create color clusters.
No. This is nothing RawTherapee specific. That "noise" is present in raw sensor data and pretty much follows the Poisson distribution. You can actually simulate that.
United States Tendances
- 1. StandX 4,995 posts
- 2. #NationalCatDay 1,992 posts
- 3. Aidan Hutchinson 3,251 posts
- 4. Huda 20.6K posts
- 5. South Korea 144K posts
- 6. FOMC 35.7K posts
- 7. #SpaceMarine2 N/A
- 8. #wednesdaymotivation 4,648 posts
- 9. Hump Day 19.2K posts
- 10. Good Wednesday 33.6K posts
- 11. Jay Z 8,609 posts
- 12. $XHLD $0.27 N/A
- 13. NBA Street N/A
- 14. Marcedes Lewis 1,073 posts
- 15. Happy Hump 12K posts
- 16. #Wednesdayvibe 2,340 posts
- 17. Brad Holmes N/A
- 18. Mandy 16K posts
- 19. #MAYATVAWARDS2025 1.72M posts
- 20. $NVDA 106K posts
Something went wrong.
Something went wrong.