K3ii NOISE


Anstonian

Link Posted 25/06/2020 - 18:16
pschlute wrote:
Andrew, thanks for sending the DNG files. I processed them in ACR using an auto setting for exposure etc. I didnt apply any output sharpening but I made a noise reduction edit to one of them. These are full size jpegs so click for the larger versions. They are listed on my Flickr page as private so only viewers of this forum will be able to see them. I will delete them in a week or so, so download them if you want to keep them for a reference.

The one with the bird has a great deal of out of focus areas but I cannot see noise being particularly troublesome. OOF areas will usually show more noise.



Would be interested in anyone else's thoughts ?












and now the NR image



The one of the garden is more interesting. There is visible noise in particular on the brown bin and also the shaded part of the shed. I applied a genera noise reduction in Photoshop which reduces it a bit. You can compare the noise reduced one with the other. The original shot is underexposed and I suggest that the brightening of these areas in raw conversion has led to the noise.

Cheers for this Peter. I really appreciate your input and anybody else wishing to advise would be greatly welcomed. I have been so freaked out about this noise issue I went and took a series of shots with mine and an Olympus OM-D-E M1 Mark II just to do as straight comparison. I have yet to sit down and do a this so I am looking forward to that (when I get the time). The lens on the OM-D was a bog standard lens so I would think that would compare to my glass.
Andrew
Andrew Goble

Anstonian

Link Posted 26/06/2020 - 14:48
Hi, Just for reference I have done the comparison of pictures from my K3II and the OM-D M1 MkII and can report that the K3II has blown away the pictures from the Olympus. It seems there must be a problem with the way I set the camera up that is giving me the issues. Although I must report that the picture of the green house and the shed were taken with the camera on a sturdy tripod, SR turned off, mirror up and using a wireless shutter release. Surely under those conditions the picture should be crystal clear even when zoomed in so that brings me back to the settings I have used. Does wisdom out there believe that a better lens would make a significant difference or is it (I can take it) for a large part, purely down to me.
Andrew
Andrew Goble

Lubbyman

Link Posted 26/06/2020 - 17:09
A different lens isn't likely to make a difference to noise, noise comes from the sensor and electronics. It's always present in the electronic output from the sensor (thermal noise etc.) and the amount of noise is pretty well independent of the light level. If the signal to noise ratio (i.e. light level to noise level) is big enough, the noise is not noticeable. If the signal level (i.e. light level) is low, then the noise is proportionally bigger relative to the light level so is more noticeable. Low light level might be because of overall low exposure or because of a darker part of the image (e.g. dark colour or shadow).

Noise can be reduced or even eliminated in processing the output from the sensor - either in the camera or by the user - but something else will be lost (e.g. resolution). It's a balancing act. If you're post-processing RAW, you do noise reduction with your post-processing software (although I may have read somewhere that the camera may apply some noise reduction to RAW). Try different options and see what works best for you. If you're concerned with jpegs from the camera, adjust the camera noise reduction settings to what works best for you. And 'what works best' may depend on the type of subject (e.g. dark overall or bright + shadows).

Noise is an issue for all signal processing, the same sort of thing applies to audio, radar, radio telescopes, particle physics experiments etc. etc. etc. It's always there, the balancing act is to get rid of as much as you need without affecting the useful stuff in the process.

Not sure if that helps!

Steve
Last Edited by Lubbyman on 26/06/2020 - 17:10

Anstonian

Link Posted 26/06/2020 - 17:42
Lubbyman wrote:
A different lens isn't likely to make a difference to noise, noise comes from the sensor and electronics. It's always present in the electronic output from the sensor (thermal noise etc.) and the amount of noise is pretty well independent of the light level. If the signal to noise ratio (i.e. light level to noise level) is big enough, the noise is not noticeable. If the signal level (i.e. light level) is low, then the noise is proportionally bigger relative to the light level so is more noticeable. Low light level might be because of overall low exposure or because of a darker part of the image (e.g. dark colour or shadow).

Noise can be reduced or even eliminated in processing the output from the sensor - either in the camera or by the user - but something else will be lost (e.g. resolution). It's a balancing act. If you're post-processing RAW, you do noise reduction with your post-processing software (although I may have read somewhere that the camera may apply some noise reduction to RAW). Try different options and see what works best for you. If you're concerned with jpegs from the camera, adjust the camera noise reduction settings to what works best for you. And 'what works best' may depend on the type of subject (e.g. dark overall or bright + shadows).

Noise is an issue for all signal processing, the same sort of thing applies to audio, radar, radio telescopes, particle physics experiments etc. etc. etc. It's always there, the balancing act is to get rid of as much as you need without affecting the useful stuff in the process.

Not sure if that helps!

Steve

Thanks for this Steve. I was starting to lean towards the problem was (as I saw it) with the camera and not with me. Having compared the shots from the K3II to the Olympus M1 MkII I am satisfied the problem is me not using the tech correctly. Now that is something I will be able to rectify. So, for sure, your remarks are really helpful and welcomed.
Cheers
Andrew
Andrew Goble

Lubbyman

Link Posted 26/06/2020 - 23:35
Here's the results of a test done this evening with the K3 + 16-85mm (a very good lens) to show signal to noise ratio in practice. Taken outside, bright but overcast, uniform light. Subject was a neutral (very dark grey) board. All at ISO 200 and f8, defocussed to minimise the effect of imperfections of the matt painted surface. All images are 800x800 pixel crops so the forum images should show exact pixels. RAW (DNG) files, no noise reduction in camera or in post processing (with Affinity Photo), converted to highest quality jpegs for the forum.

Here's the base image. Camera chose 1/25sec to give a mid-grey (as it should). Some noise apparent but not much and easy to remove in post processing (the default noise reduction in Affinity Photo does a good job).




Now an image underexposed by 4 stops (at 1/400sec) with + 4 stops added in post processing to bring the average brightness up to base level. Although the overall brightness is the same as the base image, there's a lot more noise. That's because the noise from the sensor has been increased by 4 stops in post processing. More difficult to remove in a 'real' picture because there's more of it competing with the detail that you'd want to retain (the software wouldn't know what is noise and what is 'real' detail).



Finally, an image overexposed by 4 stops (0.6 sec) with -4 stops added in post processing to bring the average brightness up to base level. Although the overall brightness is the same as the base image, there's less noise. That's because the noise from the sensor has been decreased by 4 stops in post processing. You've got to look very hard to see any noise, and that's without any noise reduction at all.



In a real picture, the same level of noise will be present everywhere. However, it will be more intrusive in darker areas because there is less light to mask it. It's up to you how much noise is acceptable and to ensure that it is below that level in the final image. Noise reduction in camera or post processing software is good, but it can never know whether the random specks in a shadow or dark area are sensor noise or vital detail that you need to see.

I hope that makes sense!

Steve

pschlute

Link Posted 27/06/2020 - 00:41
Anstonian wrote:
Hi, Just for reference I have done the comparison of pictures from my K3II and the OM-D M1 MkII and can report that the K3II has blown away the pictures from the Olympus. It seems there must be a problem with the way I set the camera up that is giving me the issues. Although I must report that the picture of the green house and the shed were taken with the camera on a sturdy tripod, SR turned off, mirror up and using a wireless shutter release. Surely under those conditions the picture should be crystal clear even when zoomed in so that brings me back to the settings I have used.

What did you focus on ? It appears that the shed and wooden arbour are in sharpest focus with everything else a bit soft, especially further into the image. You used f16 which will give you a wide depth of field, but that does not mean that at 100% zoom everything will be in sharp focus. Your plane of focus should be very sharp however. f16 is also where diffraction will start to kick in causing softening. I added no "output sharpening" to these images but when applied the shed and arbour sharpen up quite nicely. The edges at 18mm are always going to be much softer than the centre.

But in your shoes I would do some further testing. Use a medium aperture like f8 and be clear where you are focussing. Then examine the results. Post on here or email me some images if you would like me to have a look.
Peter



My Flickr page

MrB

Link Posted 27/06/2020 - 06:54
Steve, thank you for your post - those tests are very helpful examples. To help clarify further the effect of noise and noise reduction, do you have image 2 (at 1/400sec) but with the +4 stops added in camera (ISO 3200) for comparison?

Philip

Lubbyman

Link Posted 27/06/2020 - 21:35
MrB wrote:
To help clarify further the effect of noise and noise reduction, do you have image 2 (at 1/400sec) but with the +4 stops added in camera (ISO 3200) for comparison?

Not done yesterday. But just done it this evening. Same set up, same process, light similar but got a bit brighter after test shots to choose today's base exposure (1/13 sec) so images are lighter than yesterday, also some variation between subsequent shots (why do clouds have to move??? ).

Here is the base image. Noise much as in yesterday's base (allowing for overall brighter image).



Now underexposed by 4 stops with 4 stops added in post processing to bring the average brightnesss up to base level. Noise much as in yesterday's equivalent (allowing for overall brighter image).



Now increasing the in-camera sensitivity from ISO200 to ISO3200 to compensate for the reduced exposure time. And the result is (wait for it ... hushed expectancy ... big drum roll ...)



Yes, allowing for slight change in overall brightness due to slight change in light between shots, noise is essentially the same as when compensating for exposure in post processing. And that's as it should be, if my understanding of RAW and ISO is correct. Changing ISO doesn't change the output from the sensor so the ratio of noise to signal (light level) is unchanged. Increasing ISO affects downstream processing by amplifying the combination of noise and signal from the sensor, so the result should be the same whether done in camera or on computer. In fact, I'm not sure whether changing ISO actually changes RAW data or whether ISA is merely a number in the RAW file that is then used by post processing software to get the intended brightness. If the latter, it would be equivalent to me telling the software to increase exposure by 4 stops. Either way, the noise looks the same whether ISO is changed in camera or exposure is changed in post processing. Of course, this only applies if using RAW files, jpegs straight from camera will have lots of processing done already, including accounting for ISO.

As the saying goes, a picture is worth a thousand words.

Steve

MrB

Link Posted 28/06/2020 - 00:44
That's very interesting - thank you for doing another set of comparison images, Steve.

All the best.
Philip

Lubbyman

Link Posted 28/06/2020 - 10:59
A pleasure. Particularly in comparison with the shifting and repairing of the compost heap that I'd been doing earlier in the day.

Steve
Add a Comment
You must be registered or logged-in to comment.