Concepts High ISO and Grainy photos

To indicate that this thread is a discussion of theoretical concepts
The link doesn't go anywhere.

The examples show how uniformly the demosaicing algorithm works on a non-image. A blank piece of paper is useless as a subject for a photo comparison. You need to design some sort of a standard subject to do meaningful comparisons.
 
john.margetts said:
The link doesn't go anywhere.

The examples show how uniformly the demosaicing algorithm works on a non-image. A blank piece of paper is useless as a subject for a photo comparison. You need to design some sort of a standard subject to do meaningful comparisons.
Click to expand...
The link does work!

It's not an image of the paper! It's not focussed on the paper. The point is that I am looking for the noise regardless of the image.
 
Yeah, noise is created by not enough light which we compensate with higher isos, there are a few examples of this on the web knocking around and actually Tony Northrups is good
 
You cannot have noise or otherwise regardless of the image. Any noise is an integral part of the image. If you look at a real image, the noise varies according to the image density at any particular point.
 
john.margetts said:
You cannot have noise or otherwise regardless of the image. Any noise is an integral part of the image. If you look at a real image, the noise varies according to the image density at any particular point.
Click to expand...
Noise is related to signal strength, not the image content. It's physics, not photography!

So in this test there is intentionally no image so that we can see the noise that results from different levels of lighting.
 
The noise is caused by the way the camera's on-board computer processes the sensor output - even with Raw, there is a lot of processing going on. That processing depends on the image: the colour and intensity of each pixel is informed by the colour and intensity of surrounding pixels.
 
..and the link does work.
 
john.margetts said:
The noise is caused by the way the camera's on-board computer processes the sensor output - even with Raw, there is a lot of processing going on. That processing depends on the image: the colour and intensity of each pixel is informed by the colour and intensity of surrounding pixels.
Click to expand...
Have a look at the recommendation of

acrobatic_Citron​

Google 'Tony Northrup and iso'. I've just glanced at it but it looks like he explains it quite well.
 
"If you have taken a picture and under exposed it by even 5 stops so it appears almost black and you wish you had used a higher ISO then don’t because it makes no difference. Increasing the exposure by 5 stops in post processing should bring the picture back as it should have been….unless it’s a Canon where you might have used 100 ISO and 1600 ISO would have given you a better picture."

I remember my old Canon DSLR's. You couldn't lift the shadows without seeing horrible noise.

I remember I once accidentally under exposed some 20D pictures taken on a dull day and the results are nothing like what would happen if I did the same with my Sony A7.
 
woof woof said:
"If you have taken a picture and under exposed it by even 5 stops so it appears almost black and you wish you had used a higher ISO then don’t because it makes no difference. Increasing the exposure by 5 stops in post processing should bring the picture back as it should have been….unless it’s a Canon where you might have used 100 ISO and 1600 ISO would have given you a better picture."

I remember my old Canon DSLR's. You couldn't lift the shadows without seeing horrible noise.

I remember I once accidentally under exposed some 20D pictures taken on a dull day and the results are nothing like what would happen if I did the same with my Sony A7.
Click to expand...
Interesting. I use a Canon 6D because I like Canons and the low light performance is good. I bought it for astro so uniform low level noise is important.

For your Canon pictures where the dark depths were noisy I would say that's where you could increase the iso to 3200, 6400 so long as you don't blow out the highlights. It would increase the Signal to Noise in those areas.
 
miscat said:
Interesting. I use a Canon 6D because I like Canons and the low light performance is good. I bought it for astro so uniform low level noise is important.

For your Canon pictures where the dark depths were noisy I would say that's where you could increase the iso to 3200, 6400 so long as you don't blow out the highlights. It would increase the Signal to Noise in those areas.
Click to expand...

Canon were always behind the curve and I do remember when Nikon overtook them. Those earlier DSLR's, and I'd include their first FF camera, the 5D, did not perform well by todays standards. These days it can be possible to use very high ISO's and still end up with useable pictures. I think one factor in harming image quality is some artificial lighting, I've seen ISO 25,600 pictures which are useable and ISO 1,600 pictures which are horrible because of the awful artificial light.
 
woof woof said:
Canon were always behind the curve and I do remember when Nikon overtook them. Those earlier DSLR's, and I'd include their first FF camera, the 5D, did not perform well by todays standards. These days it can be possible to use very high ISO's and still end up with useable pictures. I think one factor in harming image quality is some artificial lighting, I've seen ISO 25,600 pictures which are useable and ISO 1,600 pictures which are horrible because of the awful artificial light.
Click to expand...
My link in the article to the astro stuff compares the Sony A7 to the Canon 6d as it happens!

https://www.lonelyspeck.com/how-to-find-the-best-iso-for-astrophotography-dynamic-range-and-noise/
 
woof woof said:
"If you have taken a picture and under exposed it by even 5 stops so it appears almost black and you wish you had used a higher ISO then don’t because it makes no difference. Increasing the exposure by 5 stops in post processing should bring the picture back as it should have been….unless it’s a Canon where you might have used 100 ISO and 1600 ISO would have given you a better picture."
Click to expand...

Why 5 stops? why not.. I don't know... 9 stops ...?

ISO 64 +9 EV versus ISO 25k, ok maybe a little too much (I had to use the adjustment brush for this)
 

Attachments

  • pushing (1 of 2) - B0001375.jpg
    pushing (1 of 2) - B0001375.jpg
    330.5 KB · Views: 8
  • pushing (2 of 2) - B0001376.jpg
    pushing (2 of 2) - B0001376.jpg
    466.8 KB · Views: 8
Last edited:
lindsay said:
..and the link does work.
Click to expand...

Strictly speaking the link works, but there may be no useful content when you get there, just a grey box, at least with Safari.

Opening the page in Chrome displays the article, which is a PDF object embedded in the page. It seems there's a bug in Safari that means the particular way this has been done by the web designer doesn't work with Safari.*

*other browsers might have other problems
 
dancook said:
Why 5 stops? why not.. I don't know... 9 stops ...?

ISO 64 +9 EV versus ISO 25k, ok maybe a little too much (I had to use the adjustment brush for this)
Click to expand...

With modern kit you have a good chance, now lets try with a Canon offering from a few years ago :D

I remember when I tried my MFT Panasonic G1 camera against my FF Canon 5D, that comparison made me realise that the tech was moving on.
 
Musicman said:
Strictly speaking the link works, but there may be no useful content when you get there, just a grey box, at least with Safari.

Opening the page in Chrome displays the article, which is a PDF object embedded in the page. It seems there's a bug in Safari that means the particular way this has been done by the web designer doesn't work with Safari.*

*other browsers might have other problems
Click to expand...
Ah! Safari! It's hard to test Safari with Windows! I'll look into that.
 
miscat said:
yes no reason for 5. I wonder why your iso 64 is all washed out. It could be due to poor performance at low ISO, like the canon example.
Click to expand...

It's from a Hasselblad X2D digital medium format camera shooting at 16-bit with a quoted 15 stops of dynamic range - it has some of the best highlight and shadow recovery a camera can get.

But the 15 stops applies for the full range from shadows to highlights, and so thinking about it if you under exposure shadows by 9 stops (more than half the dynamic range, I don't know how its weighted) then I assume no data will be recorded.
 
Last edited:
Musicman said:
Strictly speaking the link works, but there may be no useful content when you get there, just a grey box, at least with Safari.

Opening the page in Chrome displays the article, which is a PDF object embedded in the page. It seems there's a bug in Safari that means the particular way this has been done by the web designer doesn't work with Safari.*

*other browsers might have other problems
Click to expand...
I've updated it so that there is now a link above yor grey box so safari users can see it. I know if doesn't affect you 'cos you're got Chrome. Thanks for pointing it out.
 
john.margetts said:
The link doesn't go anywhere.

The examples show how uniformly the demosaicing algorithm works on a non-image. A blank piece of paper is useless as a subject for a photo comparison. You need to design some sort of a standard subject to do meaningful comparisons.
Click to expand...
The link doesn't work in safari apparently! If you look again you will see that I have added a direct link to the pdf for people who cannot see the document in the page.
 
Broadly, it seems like the article is discovering that a lot of modern cameras are more or less ISO Invariant*

If there is not a lot of light though, and you need to maintain a minimum shutter speed (such as to prevent motion blur) at a fixed aperture, you are still going to have to raise the ISO.

As with everything, there's a balance to be struck between various parameters depending on the requirements of a given situation. If you want less noise, and more dynamic range, then keep the ISO low and increase the amount of light you capture with a longer exposure, a wider aperture or by adding more light (e.g. flash). If you want to avoid clipping highlights, then under expose by shooting at lower ISO (with the same aperture and shutter speed) and push the gain in PP.

* Caveat: some sensors take advantage of dual gain, where dynamic range gets a bump up, such as at ISO 640 on my Sony A7III

www.dpreview.com

Sony a7 III dynamic range and high ISO improve over its predecessor

On paper, the Sony a7 III is a tempting option for photographers who've been considering a switch to full-frame mirrorless. But how does its image quality stack up? We compare it to the Mark II and a few of its other peers.
www.dpreview.com www.dpreview.com
 
P.S. There's a glaring error here, though

If ISO has no effect on noise why not always use the highest ISO we can
and have shorter exposures?
Because your photos will be underexposed and therefore be noisy. If you are using high ISO it is to
compensate for not having enough light.
Click to expand...

If you use high ISO (and don't adjust anything else) when there is plenty of light then your photos are going to be over exposed, not under exposed.

If you do adjust shutter speed or aperture to compensate, then you photos should be correctly exposed.
 
Musicman said:
Broadly, it seems like the article is discovering that a lot of modern cameras are more or less ISO Invariant*
Click to expand...
The linked photographylife article makes a very common mistake... ISO invariance is not caused by modern cameras having low "back end read noise". It's a very common/popular misconception because it is easy and logical.

What actually causes ISO invariance is that the read noise (primarily conversion errors by the ADC) overwhelms the signal noise generated (shot noise)... i.e. you can't boost the signal enough to make a significant difference. Many of the early digital cameras were essentially completely ISO invariant; but they were also very poor performers. What makes modern ISO invariance useful is a reduction in the camera generated noise levels up to, and including, the op-amp (primarily switching noise); AND a simultaneous reduction in noise levels after analog gain stage.

What makes a camera non-ISO invariant is that the ADC conversion/write process makes more errors when fed voltages of too low a level... so raising the ISO to boost the voltages also increases the accuracy of the stored/converted data... it is not "a peculiarity of the order in which the signal is processed with these sensors," as stated in the OP's linked article.
 
Last edited:
Musicman said:
P.S. There's a glaring error here, though



If you use high ISO (and don't adjust anything else) when there is plenty of light then your photos are going to be over exposed, not under exposed.

If you do adjust shutter speed or aperture to compensate, then you photos should be correctly exposed.
Click to expand...
Ah well, what do we mean by under exposed?

In your meaning I would say the final picture is "saturated"... that is blown out at max white because we have amplified the signal too much.

I would say that the if you are able to use a high ISO during normal conditions is is because the your shutter speed and aperture are insufficient so the pic is therefore underexposed!

I could take an indoor picture of the cat, who is not moving, at ISO 25600 and 1/1000 or I could choose to take the pic at ISO 400 and 1/50th. The latter will be well exposed and the former under exposed.
 
miscat said:
I could take an indoor picture of the cat, who is not moving, at ISO 25600 and 1/1000 or I could choose to take the pic at ISO 400 and 1/50th. The latter will be well exposed and the former under exposed.
Click to expand...
It's a basic change of concept in that ISO is not "exposure" with digital... (other than the dual gain caveat) ISO does not change the sensor's reactivity to light as it does with film; it is simply a brightness control.
 
sk66 said:
It's a basic change of concept in that ISO is not "exposure" with digital... (other than the dual gain caveat) ISO does not change the sensor's reactivity to light as it does with film; it is simply a brightness control.
Click to expand...
Yes. The exposure triangle idea is still good for getting the right final image brightness but it doesn't tell you about how noisy it will be.
 
miscat said:
Yes. The exposure triangle idea is still good for getting the right final image brightness but it doesn't tell you about how noisy it will be.
Click to expand...
The "exposure triangle" never really was ISO... it is, and always has been, available light, aperture, and shutter speed.
With film you choose the film speed based on the amount of light you anticipate being available; choosing a higher ISO to compensate for a lack of light/exposure, the same as with digital.

ISO is just "a stand in" for available light; but the ability to add /manipulate/choose the available light is a much more significant factor in creating an image...
 
You must log in or register to reply here.
Back
Top