3
\$\begingroup\$

I have an Canon 200D (Rebel SL).

I am concerned about the red noise in the darker parts of the picture, at low ISO values. Especially in the blue sky. Could this be a hardware problem with a sensor?

I post screenshots and two original RAW files:

image1

image2

https://megaupload.nz/U0k69aufbd/IMG_1590_CR2

https://megaupload.nz/a9la96u0b9/IMG_1673_CR2

\$\endgroup\$

3 Answers 3

6
\$\begingroup\$

No, this is not a problem.

I see red noise. I see green noise. I see blue noise.

That's not a surprise because there are three main colors in the Bayer filter, and the pixels beneath all of those three main colors are noisy.

The red and green noise are however more prominent, because the image is blue.

This noise is called chroma (chrominance) noise, as opposed to luma (luminance) noise.

You need to take a look at various noise reduction algorithms. Chances are your camera has already one for JPEG files but it won't be used for RAW files.

For RAW files, you can use the noise reduction algorithms of RawTherapee. Darktable too has some noise reduction algorithms, but I have found them to be much poorer than the ones in RawTherapee.

When you turn on noise reduction in RawTherapee, it should remove chroma noise effectively. To remove luma noise, there is an adjustable slider.

If you want less noise, consider using ISO 100 which is slightly less noisy than ISO 200. You could also consider switching to a camera that has Sony Exmor sensor which is somewhat less noisy than Canon sensors, but that would be expensive if you have already invested lot of money into Canon. You could also consider a full frame camera (or a full frame camera with Exmor sensor), but that would be even more expensive.

Typically there is less noise when:

  • The sensor is big (full frame)
  • The sensor is good (Sony Exmor)
  • The ISO level is low and thus exposure is higher¹
  • Noise reduction algorithms are used

¹ ISO actually has very little influence on the amount of noise in an image. It is the lower amounts of exposure we use with higher ISO settings that does that. If we shoot the same scene at [ISO 100, f/2.8, 1/30] and at [ISO 1600, f/2.8, 1/500] the difference in noise isn't because we used ISO 1600, it is because we exposed for 1/16th as long and thus allowed 1/16th as much light into the camera. Please see What are the advantages and disadvantages of using auto ISO on a DSLR? and Why would using higher ISO and faster shutter speed yield more noise than using lower ISO and slower shutter speed? for more.

\$\endgroup\$
8
  • \$\begingroup\$ I wouldn't say that other sensors are necessarily better (outside a laoratory) in noise handling: dxomark.com/Cameras/Compare/Side-by-side/… \$\endgroup\$
    – flolilo
    Commented Feb 24, 2019 at 14:08
  • \$\begingroup\$ Thank you, juhist, for your comprehensive reply. \$\endgroup\$
    – Wedrowycz
    Commented Feb 24, 2019 at 14:10
  • \$\begingroup\$ @flolilolilo Well, I have seen some low-light photos taken with an iPhone camera using Sony Exmor sensor, and I was positively impressed for being so good given the sensor size. It may be the case that for APS-C or FF size the differences diminish to marginal. \$\endgroup\$
    – juhist
    Commented Feb 24, 2019 at 14:12
  • 2
    \$\begingroup\$ I'd think that is down to noise reduction. \$\endgroup\$
    – flolilo
    Commented Feb 24, 2019 at 14:34
  • \$\begingroup\$ Some cameras do more on-die NR than others, which will affect the scores at DxO where the on-die NR affects the score but no off-die NR is done. It will also affect the way things that are not actually noise may be mistaken for noise and processed out of an image, even the raw file. Sony Exmor sensors are not called "star eaters" for no reason. The on-die NR that gives them such good DxO Mark scores also tends to eliminate dim stars as "noise". \$\endgroup\$
    – Michael C
    Commented Feb 24, 2019 at 20:47
3
\$\begingroup\$
  1. Your screenshots show ~300% magnification. For comparison, I have a 24" 1920x1200px (so ~94 ppi) monitor and viewing your picture at 30% zoom (which would roughly be the size at 300dpi), it just fills my monitor from top to bottom! What you are doing is known as pixel-peeping - and 300% is an extreme version of that. While it is not inherently bad to pixel-peep, always keep in mind that it is unlikely that you will ever behold that picture at such a magnification 'in real life'.
  2. I think there was no noise reduction applied at all. There is always noise, so at least a little bit of noise reduction never hurts in my opinion.

There is nothing wrong with your sensor and it does not even perform that badly - here is a screenshot-comparison between your photo and some random one of mine with made with a 5D Mark III:

Note: Unfortunately, there will be some artifacts due to imgur's transcoding. While not too bad, one could mistake them for luma noise. Also, I have to confess that when watching the photos again, it seems likw choosing a picture of a forest to compare it to a solidly-colored house was not a good idea - sorry for that.

In 100% views it slowly becomes difficult to even spot the noise with as little as 25% noise reduction (C1 defaults to 50%). In my opinion, that noise level would be good enough for a ~A3-sized poster even at very close distances - and beyond most people's requirements.

\$\endgroup\$
1
\$\begingroup\$

The sensor used in Your Canon EOS Rebel SL2/200D is basically the same sensor used in the EOS 80D, the EOS Rebel T7i/800D, EOS-M M6, EOS-M M100, and 77D. It's Canon's current off-the-shelf APS-C sensor. Some of Canon's lower end APS-C models use older sensors that were used in previous versions of the higher level APS-C models, such as the EOS Rebel T6/1300D (introduced in 2016) that uses the same basic 18MP sensor that was introduced with the EOS 7D way back in 2009.

All digital sensors result in image noise. Some of them, such as Sony's Exmor sensors, do a significant amount of noise reduction on the imaging chip itself before the data is moved off the sensor to the camera's image processing circuitry. Other sensors, such as most of Canon's line, do less on-die NR and allow varying decisions about how noise should be handled up to the processing routines used to convert the raw image data into pictures that we look at.

There are advantages and disadvantages to either method. The Sony sensors are sometimes referred to as "star eaters" by those that do astronomical photography because the on-die NR done by them often eliminates very dim stars along with noise! Sensors that do less NR on the imaging chip itself require more NR in post-processing. If you turn NR completely off when processing an image from such a camera, you will be able to see the type of noise that is in your examples. This will be particularly the case when you view them so magnified that you're looking at the equivalent of a small section of something like a 60x40" version of the full picture.

There's nothing defective about your sensor. The images you show are about what is to be expected when very little or no noise reduction is applied to a slightly underexposed image taken using that sensor and viewed at such high magnifications.

Applying appropriate amounts of noise reduction when the image is processed should remove most of the chroma (color) noise in your image. If you are saving JPEGs straight out of the camera, turn the noise reduction setting in the camera's menu back on to the default "medium" setting. If you are processing the raw image data externally using a raw processing application, use that program's NR tool to apply the needed amount of NR.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.