What is digital noise? What causes it? How do we reduce it? We look into the answers.
What is noise in photography?
Digital noise can be caused by low light, high ISO settings, long exposure times, and heat. This can result in variations of either color or brightness. Regrettably, this causes irregular pixels that interfere with the luminance and tonality of the photograph. Night photographers photographing with high ISO settings in dark conditions are particularly familiar with noise.
What kinds of digital noise are there?
- Random noise is statistically random variations in the brightness, as the name implies.
- Banding noise is a type of semi-fixed-pattern noise that manifests as faint vertical or horizontal stripes.
- Fixed pattern noise manifests itself as repetitive patterns of bands of color or brightness superimposed on the image.
How do we reduce noise in the field?
To reduce pattern noise in low light photography, you can try using a lower ISO setting.
Depending on the situation, you could also try a shorter exposure time. Additionally, some cameras have a “long exposure noise reduction,” or LENR, feature. This can help reduce pattern noise by taking a second exposure of equal length with the shutter closed and subtracting the noise from the first exposure. However, doing so effectively doubles the time that it takes to render the exposure, tying up your camera in the field even longer.
And finally, you can do something similar to LENR and take an equal exposure of equal length with the lens cap on, then apply that in post-processing. It’s still time-consuming, but you can likely take only one of these dark subtractive photos and then get on with your business for a while, just doing so occasionally.
Regardless, it’s not always possible to get rid of all noise. For the rest, we can use post-processing.
Denoising on your computer
I have been using Topaz Labs Denoise AI for over a year now. To me, it works almost magically, employing machine learning to determine what is noise and what is not, then effectively getting rid of it.
Is it perfect? No, of course not. No noise reduction software is. But it is really great. And this is coming from someone who is a night photographer. I’ve written about Topaz Labs Denoise AI several times. I’ve tested it in low light with high ISO settings. I have also compared Topaz DeNoise AI to stacked exposures of the night sky. And it keeps getting better as machine learning evolves.
Several other software manufacturers also utilize AI to address noise issues, including DxO PureRAW. AVC AI Image Denoiser works by uploading your image. I have not tried either. Currently, I’m using Topaz although I should mention that the noise reduction in Radiant Photo doesn’t seem to get enough love. It’s surprisingly good, and I believe it too uses AI.
Ken,
Many thanks for this article. I will try your suggestion of using the LENR, if my old camera has such. If not then the lens-cap-in-place approach.
For folks who perhaps still have the free Dfine2 from Google – it is a bit tedious and you have to be reasonably acquainted with it but it works fairly well.
Look forward to your next article.
Ed
My 2005 DSLR cameras had LENR, so I’m guessing that most do.
I was using Nik Dfine prior to Topaz Denoise AI. It’s an enormous leap up. To target the noise reduction when using Dfine, I found that I had to pair it with luminosity masks to target the really dark shadows or the sky. I was able to do this effectively, but it was definitely a bit more work. Maybe there’s a way to do it “in-house”, but if so, I never did it. :D
Thanks for your comments!!
I have DXO’s “Photolab” and Tapaz Denoise and don’t think it’s even a close. DxO is far better at removing noise. The other adjustments in DxO are icing on the cake and are more helpful than Lightroom. It’s become part of my workflow.
A number of people do like DxO Photolab more than Topaz Labs Denoise AI. What sorts of images are you removing noise from? Do you feel it’s more effective for a particular sort of image? Like I say, I’ve never used it but thought I’d mention it since it gets a lot of love.
The first sentence is incorrect. Amplifying the signal does not result in random variation – the variation is already there before amplification, and amplification neither creates nor removes noise. Later in the article, “Random noise is statistically random variations in the brightness, as the name implies.” is correct. Noise is a result of per-pixel samples of photons being smaller at low light levels, which results in greater between pixel variation, which we see as noise (in stats, this between sample variation is called standard error, which increases as sample sizes decrease). This is why, all other things being equal, larger… Read more »
Thanks. I’ll need to fix that first sentence. I’m not sure how that got in there, actually.
Fixed it, but keeping your comment here. I think at the time, I was writing another article about microphones and get crossed up. I appreciate you pointing that out.