Pages: [1]   Go Down

Author Topic: Digital Noise  (Read 2016 times)

John R Smith

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1357
  • Still crazy, after all these years
Digital Noise
« on: July 08, 2010, 06:04:58 am »

Good Morning from Cornwall (where the weather is fine and bright with great skies, ideal for photography, but I am stuck in the Office).

This is a general question about digital noise. Now that I have been working with my Hasselblad 3FR files in Lightroom for a while, I am puzzled by why, exactly, some areas of the image seem to be more affected by noise than others. In general, I can understand that noise should be apparent in under-exposed areas of the image, and not visible in the correctly exposed areas. But my pictures do not always seem to follow this logic.

Colour Noise – (the magenta and cyan speckles). This does seem to be worst in the shadowed areas of the image, particularly where the contrast is very low. But sometimes it is very noticeable where I would not expect it, for example on the shaded side of someone’s face, which is no darker than the side of a cream boat – the boat being noise-free, but the face badly affected. And particularly in skies, which are not exactly under-exposed, but seem to always be the worst area of the picture for colour noise. But correctly-lit dark objects (like a black-painted wooden shed) show no noise, even though they print at more or less the same luminance as the shadows, which are noisy.

Luminance Noise – (non-coloured speckles more like film grain). I very rarely seem to get this in the 3FR files, but when I do see it is almost always in a white area (notice-board, white boat) which is not in full shadow but semi-shade. Or again, in areas of very low contrast which are not in full shadow. What causes this, and how does the mechanism differ from colour noise?

Essentially, I would like to improve my technique so that I did not need to use any noise removal at all, although I appreciate that this is probably impossible. At present I need to use very little – in LR I usually need 0 luminance removal and between 5 to 15 for colour noise. For my ISO 50 and 100 frames this amount will completely smooth them, except for the skies which are always more tricky. In LR there there does not seem to be any way of selectively applying NR only to certain parts of the image, so far as I can see. Whereas using masking, it is possible to restrict the effects of sharpening. What would be nice would be to apply NR only to those parts of the picture which require it.

Having only worked with digital files for the past six months (previously I shot film and scanned it) I realise how little I understand this whole topic of digital noise. I can never entirely predict exactly where the noise will be, and because one can only see the effects of NR at 100%, it is completely possible to miss the parts of the image that are noisy when you can only see a tiny section of it on the screen at any one time. Or you spot the colour noise and deal with that, not realising that a completely different part of the frame is actually affected more by luminance noise.

Incidentally, as I increase the ISO to 200, 400 and 800, not only does the noise increase as I would expect, but DR shrinks also. So the higher the ISO, the sooner the highlights are clipped, which did throw me a bit at first, as you need to be more careful with your metering at ISO 400 and up.

John
« Last Edit: July 08, 2010, 09:02:05 am by John R Smith »
Logged
Hasselblad 500 C/M, SWC and CFV-39 DB
an

digitaldog

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 20652
  • Andrew Rodney
    • http://www.digitaldog.net/
Digital Noise
« Reply #1 on: July 08, 2010, 09:50:31 am »

What you can do in LR is bracket some images at various ISOs (I’d use something like a Macbeth Color Checker or a scene with a decent range of colors and tones). You can then adjust each ISO bracket with the best possible default slider settings for NR and build just that into a default rendering for each specific ISO setting. The trick is to make sure you start with an image that has no other adjustments you do not want “baked into” this ISO specific default. Go into the LR preferences and set the “Make defaults specific to camera ISO Settings” for on. This will produce a default setting that is applied by differing ISO. Then set the NR (or other settings you want for your ISO specific default). In Develop, Alt/Option key toggles Reset to “Set Default”. The resulting dialog shows the ISO and camera serial for that new default setting you will build. Now you have ISO specific NR settings which is kind of useful.
Logged
http://www.digitaldog.net/
Author "Color Management for Photographers".

John R Smith

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1357
  • Still crazy, after all these years
Digital Noise
« Reply #2 on: July 08, 2010, 10:58:12 am »

Thank you, Andrew. I will have a look at that strategy.

One thing that I did wonder was - are the R, G and B channels on a CCD sensor prone to differering levels of noise? Remember that I work exclusively in B/W, but of course I do use different levels of each channel to do the monochrome conversion (using the Grayscale sliders in Lightroom). So sometimes my picture will have more of the red channel, other times more green, and so on. Could this be the reason that I see these varying results?

John
Logged
Hasselblad 500 C/M, SWC and CFV-39 DB
an

digitaldog

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 20652
  • Andrew Rodney
    • http://www.digitaldog.net/
Digital Noise
« Reply #3 on: July 08, 2010, 11:49:49 am »

Quote from: John R Smith
Could this be the reason that I see these varying results?

I suppose its possible.
Logged
http://www.digitaldog.net/
Author "Color Management for Photographers".

BobFisher

  • Guest
Digital Noise
« Reply #4 on: July 08, 2010, 01:01:26 pm »

Channels can have varying noise levels.  I remember some years ago in a PSE book I had (maybe for PSE 3) a section on noise reduction and targeting the adjustment to a specific channel that may be noisier than the others.

To your other point, there is a correlation between increasing ISO and a reduction in dynamic range.  Look at any camera test that shows DRange and ISO together and it'll show the relationship.  I believe it's due to the degradation in S/N as ISO is increased and typically is evident more in the shadows.
Logged

David Saffir

  • Full Member
  • ***
  • Offline Offline
  • Posts: 172
    • http://davidsaffir.wordpress.com
Digital Noise
« Reply #5 on: July 08, 2010, 06:39:35 pm »

I think you said that you are using Lightroom to develop your RAW files. What has been the result when using Hasselblad's Phocus software?

David
Logged
David Saffir
[url=http://davidsaffir.wor

deejjjaaaa

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1170
Digital Noise
« Reply #6 on: July 08, 2010, 11:14:19 pm »

you might want to check what specialized NR software offers during a postprocessing stage

Imagenomic Noiseware = www.imagenomic.com
Nik Software DFine  = www.niksoftware.com
NeatImage = www.neatimage.com
Noise Ninja = www.picturecode.com
Topaz Denoise = www.topazlabs.com

Logged

John R Smith

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1357
  • Still crazy, after all these years
Digital Noise
« Reply #7 on: July 09, 2010, 03:11:35 am »

Quote from: David Saffir
I think you said that you are using Lightroom to develop your RAW files. What has been the result when using Hasselblad's Phocus software?

David

David

In the experiments that I have made, Phocus seems to do quite a lot of NR by default (and other stuff too, like CA removal). If you were working in colour, Phocus gives a very nice result. But the B/W tools in Phocus are just too primitive, so for me LR was the software of choice.

I don't really have a noise problem, as such - I would just like to understand more about the nature of noise and be able to control it more effectively. For example - why do we get both luminance and colour noise? Is luminance noise spread evenly across all three channels (I assume it would be). Why are skies so noisy when they are in fact correctly exposed? When darkening skies, is it better from a noise point of view to darken the blue channel or to reduce luminance? Remembering that we are only considering B/W here, not colour printing.

John
Logged
Hasselblad 500 C/M, SWC and CFV-39 DB
an
Pages: [1]   Go Up