Can cameras of the same model share a usable sensor noise profile?
Asked 10/8/2019
61 views
2 answers
0
Darktable’s “denoise (profiled)” tool uses camera-model and ISO-specific noise profiles contributed by users. I expected sensor noise to vary from one individual camera to another due to manufacturing tolerances, so I’m confused how a shared profile can work.
Do cameras of the same model actually have the same noise characteristics, or are these profiles only approximations? How can software reduce noise using a model-specific profile if the exact noise pattern is random and changes from shot to shot?
Originally by Photography Stack Exchange contributor. Source · Licensed CC BY-SA 4.0
Photography Stack Exchange contributor
7y ago
2 Answers
4
No, it doesn't mean that the noise is predictable for any sensor or any sensor type: something which is predictable isn't noise, almost by definition.
What it does mean is that sensors of a given type used in a given way may have similar statistics to their noise: although the actual noise varies randomly, it obeys a 'noise model' which tells you what sort of noise there is. If you know that noise model, and particularly if you know its various tweakable parameters, then you can make use of it to attempt to remove noise from the image.
Originally by user82065. Source · Licensed CC BY-SA 4.0
user82065
7y ago
0
Generated from our catalog & community — verify before relying on it.
Cameras of the same model do not produce the exact same noise pattern, and even one camera won’t produce identical noise in two shots. The random part is still random.
What shared profiles capture is not a fixed pattern, but the statistical behavior of the noise: how much noise is typically present, how it changes with ISO, brightness, and color channel, and what kinds of noise dominate. Cameras of the same model are often similar enough that one noise model works reasonably well across many samples.
This is why profiled denoising can help: the software uses a noise model to estimate and suppress likely noise without needing the exact pixel-by-pixel noise in that specific frame.
Some noise sources are inherently random, such as shot noise from the random arrival of photons, so they cannot be perfectly predicted or removed. Other components may be more repeatable or measurable, which is why techniques like dark-frame subtraction can also help.
So the answer is: no, the noise is not identical or fully predictable, but yes, cameras of the same model can share similar enough noise statistics for a common profile to be useful.
Recommended products
UniqueBot
AI7y ago
Your Answer
Related Questions
Why does luminance noise get worse after editing, and how can I reduce Darktable denoising artifacts?
How can I reduce chroma noise in RAW files while keeping some luminance grain in open-source software?
Can noise reduction be applied to RAW data before demosaicing?
What causes front- or back-focus in a lens and camera body combination?
How do cameras apply lens distortion correction to JPEGs, and where do the correction parameters come from?