Are there tools or algorithms for denoising directly on RAW Bayer data?
Asked 7/6/2011
11 views
2 answers
0
I’m looking for noise-reduction methods that work on RAW Bayer sensor data before demosaicing, rather than applying denoising later on a processed RGB/YCbCr image. I use a Canon EOS 450D and currently work with Photoshop CS5 and GIMP, but I’m specifically interested in RAW-domain approaches because I may eventually implement my own Bayer-data denoising.
What algorithms, tools, or research papers should I look at for denoising directly in the RAW/Bayer domain?
Originally by Photography Stack Exchange contributor. Source · Licensed CC BY-SA 4.0
Photography Stack Exchange contributor
15y ago
2 Answers
5
This is well beyond my understanding of the technology but I have read an article on this here. http://www.csie.ntu.edu.tw/~fuh/personal/NoiseReductioninRawDataDomain.A2-2-0103.pdf
It was an interesting read on the subject.
Also I notice you are using GIMP and CS5 for noise reduction, I found both to give me sub par results. I ended up settling on Nik Dfine for that process when I really need it and using Lightrooms built in noise filter as part of my process when I'm feeling lazy.
Originally by user6051. Source · Licensed CC BY-SA 4.0
user6051
15y ago
0
Generated from our catalog & community — verify before relying on it.
Yes — RAW-domain denoising is a real area of research, and there are methods that operate on Bayer data before demosaicing.
A good algorithm to study is Non-Local Means (NLM). It works by averaging pixels or patches that look similar elsewhere in the image, exploiting image redundancy. It can be adapted to Bayer data by comparing only pixels from the same CFA color positions, or it can be applied after demosaicing.
For papers and demos, IPOL is a useful source for image-processing algorithms, and one answer also pointed to a paper specifically about noise reduction in the raw data domain.
If you want examples in commercial software, DxO PRIME is an example of a RAW noise-reduction approach aimed at high-ISO images. It is known for strong results, though it can be computationally heavy.
So, in short: look at RAW-domain denoising literature, especially Non-Local Means variants for CFA/Bayer data, plus raw-domain noise-reduction papers and implementations.
Recommended products
UniqueBot
AI15y ago
Your Answer
Related Questions
What are the pros and cons of different Bayer demosaicing algorithms?
Is there any image-quality benefit to converting RAW directly to black and white instead of demosaicing first?
How can I estimate an illuminant’s color temperature from a photo of a color chart?
Can noise reduction be applied to RAW data before demosaicing?
How can I view the actual sensor sample values in a RAW file?