minimising noise in digital cameras

[Loris Medici]

According to my knowledge, noise isn't "generated" by the "white balance correction" done with the firmware/software, but "by the sensor itself", unless you add it yourself later deliberately - by means of noise generating tools/filters provided by the image editing program... (E.g. the filter in PS under Filters > Noise > Add Noise...) BTW, I wouldn't qualify 3200K to 5500K white balance correction as an "extreme color correction"...

OTOH, I agree with giving good exposure (max. possible exposure - w/o causing clipping in non-specular highlights which also conforms with shooting conditions) and using the "sensor's native ISO" (Not "lowest possible ISO value"; sometimes these two aren't the same...), again, wherever possible.

I'd really like to see comparative tests proving the assertion in "practical / real life conditions", exhibiting a meaningful / tangible benefit - which also will cancel out the negative effects of using an unneeded filter... Anyone willing to take the trouble? I promise to prepare my own "objective" visuals and supporting text in return, even if it ends up in disgracing myself. "I don't fear to stand corrected" (like some...), I never had. Anyone?

P.S. To me the issue is rooted in the fact that some people still haven't completely finished their transition / conversion from analogue to digital (or just adaptation if you like), therefore are still stuck in (some) concepts which are meaningful only in the context of analogue photography...

 

[Mike1234]

According to my knowledge, noise isn't "generated" by the "white balance correction" done with the firmware/software, but "by the sensor itself"...

Yes, and noise can be generated more in one or two channels than the other(s). Yes, the sensor's lack of sensitivity is the cause. The cure? Pre-exposure color correction and the proper exposure compensation with said filter. Sensors do have three sets of sensors, right? If one set (lets say blue) is significantly underexposed then noise is generated in the blue channel whether color correction is done in-camera or in PS.

 

[pellicle]

Loris

According to my knowledge, noise isn't "generated" by the "white balance correction" done with the firmware/software, but "by the sensor itself",

that is correct, however if there is subtle but acceptable noise in each channel, and you combine them (as one does to have R G and B become a proper colour image) then the noise in each channel combines to additively make more noise.

some years ago I wrote this article exploring my coolpix 5000 and its raw capacity.

try hovering your mouse over the images and see the results of the different noises ...

I can't do the javascript for the mouse over, but here are the two images

If nothing else, note the funny splotchy bits in the grass beside the footpath? I encourage you to go to the page and try it to see this better.

It does indeed translate to a print for those who think this is just pixel peeping.

Also it is my understanding that much of what is attributed to grain on 35mm negative scans is infact channel noise, and the blue channel is commonly the culprit. The Blue channel is buried in the most filtration by a combination of factors and also is the most dense in examination of the specification of film density:

BTW, I wouldn't qualify 3200K to 5500K white balance correction as an "extreme color correction"...

nor would I ... but as sensors are linear in nature it can be.

Taken with a 4/3 sensor camera (not an uncommon size, although for the many D3s and 1DMkIII users out there such is the punishment for daring to use such a poor format)

you will note that even though I set white balance to tungsten that the blue channel is still a little down ... Now ... red

looks ok ... green

looks fine too.. then lets look at the blue

there is a saying ... that the chain is only as strong as its weakest link

(further details on my blog here)

Lastly, I would suggest that we lack the appropriate tools as photographers in determining exposures of our images, as light meters are not really tuned to cope with sensors. If you don't play round in raw much you will never really grasp just what a good exposure is. Tools such as LR and ACR mask much of what is happening. While they are good to work with in a production environment for exploring and understanding what is happening on your camera I encorage you to play with dcraw.

For instance I took this image, which seemed quite reasonable on the display of the histogram:

of course we all know that the histograms are only based on the JPG which the camera would generate with the existing camera settings.

I was surprised just how much lower the levels were in the RAW

of course that is RAW and LINAR, so as gamma will be applied plenty of room for posterization to occur to your image.

(discussed in this page)

HTH

 

[cupcake_ham]

What ever! :rolleyes:

Wow, and you're a moderator? Maybe you could post your samples showing the OP, plus leading architectural photographers are incorrect in this assumption. I think I just lost interest in this forum altogether when a moderator displays an attitude like that. What childish, unprofessional behavior!

 

[Bob Carnie]

One thing that I have found with BW Conversions that may apply to this discussion.
1. If you do not isolate the noisy channel , like the blue in Pellicles example above, and convert without getting rid of the nasty bugger .

2. Any curve shaping will only enhance the noise.
3. Any sharpening will enhance the noise.

In Pellicles example if I was converting this I would replace the blue channel with the red, add the red channel to the green in Apply Image at about 40% opacity and look at the blending modes for the best look. Then open the green channel in curves and shape to taste then convert to Greyscale.


With colour , and IMO it is harder to gut out a channel as in channel blending method above , but a few observations after working on a lot of images.

If you do not get the three colour channels nuetralized , (white balance will get you close but no cigar)

any moves in curves, levels, sharpening , dodge/burn , tonal enhancements will only create unwanted artifacts that can look like unwanted noise.


I would be interested in methods of reducing noise, that go beyond bluring a given channel. I am not sure if there are any ways of correcting this problem after the fact. If there is a way to correct noise before or after capture other than blur I am all ears.

 

[Loris Medici]

pellicle, both of your examples (with histograms) are underexposed by my standard. (First one about 1/2 stop, second one way more - something like 1 1/2 to 2 stops...)

To minimize noise, I give as much as exposure w/o causing clipping in non-specular highlights (it's the third time I mention this BTW...), later pull back the exposure slider to my liking in the RAW processing software.

You'll be amazed how better your images will get that way; creamy highlights/skin tones and clear shadows...

P.S. You're right on the fact that usually the Blue channel is the worst one...

 

[pellicle]

Hi

One thing that I have found with BW Conversions that may apply to this discussion.

In Pellicles example if I was converting this I would replace the blue channel with the red, ...
I would be interested in methods of reducing noise, that go beyond bluring a given channel.

I split the channels, dump the red, copy the green past it back into the (now greyscale) red channel as a layer and blend the layers to get a look I like (particularly if I want to tend towards a slight filtration look).

merge visible and flatten

about 2 min an image as long as I'm not having my machine go spakko handling a 300Meg image

 

[Bob Carnie]

I am a little daft here in Canada, could you step this off for me. Do you keep the blue?and why dump red, it usually has good info???
I am using CS3 on a Mac .

thanks

Hi



I split the channels, dump the red, copy the green past it back into the (now greyscale) red channel as a layer and blend the layers to get a look I like (particularly if I want to tend towards a slight filtration look).

merge visible and flatten

about 2 min an image as long as I'm not having my machine go spakko handling a 300Meg image

 

[pellicle]

I am a little daft here in Canada

no bob, its me ... replied without reading ... dump the blue

heh

 

[Loris Medici]

No-one!???

...
I'd really like to see comparative tests proving the assertion in "practical / real life conditions", exhibiting a meaningful / tangible benefit - which also will cancel out the negative effects of using an unneeded filter... Anyone willing to take the trouble? I promise to prepare my own "objective" visuals and supporting text in return, even if it ends up in disgracing myself. "I don't fear to stand corrected" (like some...), I never had. Anyone?
...

 

[Mike1234]

Loris, not all images can be edged toward the upper end of exposure as you describe. Many scenes simply contain too much dynamic range than the sensor can handle and poor color balance resulting in underexposure of one or two channels just makes things worse.

Why don't you show us YOUR tests.

 

[Loris Medici]

In principle, with indoor scenes (we're talking about 80A filter, remember?) we don't deal with and/or we don't need that much dynamic range...

My tests? Right after the OP or any other supporter of the "idea". Don't worry, I'm a man of his word...

 

[Mike1234]

Loris,

I believe you're a man of your word but I still see no tests of yours... only pellicle's.

Mike

 

[pellicle]

Loris

No-one!???

its just that I don't happen to have the time to undertake what seems like walking into an argument.

 

[Loris Medici]

I think it's more like you're pretty much aware of the consequence, therefore avoid it like hell... OTOH, I would expect that you had already tested that before publishing the article!?

Ok. I'll post results of my test made with two different digital SLR cameras; Sony A100 (released on June 2006) and A700 (released September 2007) later today or early tomorrow... (Notice that both aren't state of the art cameras.)

Loris

Its just that I don't happen to have the time to undertake what seems like walking into an argument.

 

[Loris Medici]

Mike1234, nobody has posted "comparative" tests including our assertor... Do you often have comprehension problems?

Loris,

I believe you're a man of your word but I still see no tests of yours... only pellicle's.

Mike

 

[pellicle]

I think it's more like you're pretty much aware of the consequence, therefore avoid it like hell...

Loris you can think what ever you like, of course there is nothing I can do to suggest otherwise. I do what testing I do and charge noone any fees for it. I publish what I publish and if somone seeks to extend that research, well that is what the basis of science has been for thousands of years, perhaps since Aristotle.

I have nothing to hide and nothing to gain.

you on the other hand (in my perception) seem to be always after a fight and love to counter everything I say. That is of course your prerogative.

tell me ... what motive could I have for distorting the facts? I don't own Hoya or a CC filter company, nor do I advertise any single product.

You are essentially accusing me of being biased, so I would seek the premise of innocence unless you can establish anything like a motive then the rest is your personal fancy.

I believe your personal words were:

I have a 80A CC filter, I can take the trouble of supporting my claims above with visuals if you like

so stop huffing and puffing and produce something. Not that I need you to, but I haven't asked you to ... you have been asking me to produce something which shows what I found to be wrong or flawed.

The floor is yours my friend ...

 

[Loris Medici]

Loris you can think what ever you like, of course there is nothing I can do to suggest otherwise. I do what testing I do and charge noone any fees for it. I publish what I publish and if somone seeks to extend that research, well that is what the basis of science has been for thousands of years, perhaps since Aristotle.
...

Thank you for your immense generosity.

1. Your scientific method is flawed. See: http://en.wikipedia.org/wiki/Scientific_method for further information - don't get me weary, would you?

2. Why do you assert others would only be extending your research / hypothesis? I mean, can't things work the other way? (See #1...)

...
You on the other hand (in my perception) seem to be always after a fight and love to counter everything I say.
...

Nope, I happen to have a strong allergy to misinformation and ignorance or more correctly insistence in ignorance, since ignorance itself isn't a guilt. If you happen to exhibit both that's your problem, not mine. (See the paragraph below beginning with "BTW", for further clarification...)

...
tell me ... what motive could I have for distorting the facts? I don't own Hoya or a CC filter company, nor do I advertise any single product.
...

I don't know? I'm not a psychologist.

BTW, please note that I'm not much interested about "you and your motives and/or perceptions"; I'm interested in "facts". I don't have any problems with you personally, OTOH, I definitely have a problem with the claims you've made in this context. Both original claims and supportive(!) claims against counterclaims from others (including me)...

Do you often take everything centered in/around you?

...
The floor is yours my friend

Thank you! Finally an end to off-topic / personal stuff (artful dodging?) and coming back to the original subject...

Will return ASAP. (1-2 days at max.)

 

[pellicle]

BTW, please note that I'm not much interested about "you and your motives and/or perceptions"; I'm interested in "facts".

so why did you bring them up, why not just dispassionately post information and supporting evidence rather than jumping up and down?

I look forward to your post, it will be nice to see you provide more than criticism.

 

[pellicle]

Thank you for your immense generosity.

I notice you can't go past your own need to make a sarcastic poke ...

1. Your scientific method is flawed. See: http://en.wikipedia.org/wiki/Scientific_method for further information - don't get me weary, would you?

while I said that I agree with the principles of science, I did not say that my initial post was a fully peer reviewable piece ... its nice how you twist everything ... it will be very interesting to see the rigor of your piece.

2. Why do you assert others would only be extending your research / hypothesis? I mean, can't things work the other way? (See #1...)

are you sane? do you not read what you posted? do you know the first thing about science? If you take my work and pull it apart finding faults in it, then provide an alternative direction that is extending my work.

I'm so glad I don't go to the "education" facility you {claim to} teach at

Nope, I happen to have a strong allergy to misinformation and ignorance or more correctly insistence in ignorance, since ignorance itself isn't a guilt.

then you must be having some sort of auto-immune response to yourself right now. Try anti-histamines....

Thank you! Finally an end to off-topic / personal stuff (artful dodging?)

good ... and it was poor form for you to start it

Will return ASAP. (1-2 days at max.)

splendid

 

[Loris Medici]

so why did you bring them up, why not just dispassionately post information and supporting evidence rather than jumping up and down?

That's rather something that you do way better than me; would you really like me to quote each and every instance you did that in this thread?

The evidence "boomerang" hits right back to you; would you please kind enough to expand (or better complete) your "scientific" article with some "evidence"?
(Hint: Experiment(s) including control sample(s) - the very reason in my insistence on seeing comparative tests. See: http://en.wikipedia.org/wiki/Experimental_control for further information...)

I look forward to your post, it will be nice to see you provide more than criticism.

Sure I can do (and will do). I hope you'll take it as an example for your future tests/articles...

 

[Loris Medici]

while I said that I agree with the principles of science, I did not say that my initial post was a fully peer reviewable piece ... its nice how you twist everything
[/QOUTE]

Oh, I must have been confused by your format (earlier) and descriptive words (later) then:

"...I publish what I publish and if somone seeks to extend that 'research', well that is what the basis of 'science' has been for thousands of years, perhaps since Aristotle..."

(Note: All emphasis above belongs to me...)

...
are you sane? do you not read what you posted? do you know the first thing about science? If you take my work and pull it apart finding faults in it, then provide an alternative direction that is extending my work.
...
[/QOUTE]

Your work isn't extendable, because it's totally wrong/faulty. (There's another term - right on the spot - for it actually, which I refrain to use in public...) To refresh your mind, I provided the alternative (actually, correct) direction in the third sentence of my very first message to the thread. Someone else "extended" it )) by mentioning "shooting RAW" - which I forgot to say that myself, since never shoot JPEG...

...
I'm so glad I don't go to the "education" facility you {claim to} teach at
...
[/QOUTE]

It's not an educational facility actually, and I'm not a teacher. I do occasional alternative photography workshops in the Istanbul Photography Center. (Which happens to be Leica Gallery Istanbul at the same time.)

My degree is in business administration and my (16 years) IT career is in the fields of database programming / business analysis / system integration / ERP systems and implementation of ERP systems.

Later...

 

[pellicle]

Loris

...
I'm so glad I don't go to the "education" facility you {claim to} teach at
...
[/QOUTE]

It's not an educational facility actually, and I'm not a teacher. I do occasional alternative photography workshops in the Istanbul Photography Center. (Which happens to be Leica Gallery Istanbul at the same time.)

My degree is in business administration and my (16 years) IT career is in the fields of database programming / business analysis / system integration / ERP systems and implementation of ERP systems.

Later...

thanks for clearing that up, previously in another thread you'd insinuated that you teach photography.

So now we know your fields of expertise .. thanks for that.

While you are preparing your comeback experiment I've put this quick experiment together to cover that topic you alluded to above.

This is a white toilet paper roll, photographed (of course) in tungsten light (not a fluro or energy saver). Its not a photoflood, so the colour temperature will be below the proper rating of Tungsten ... still, its what most of us work with



I left the camera on Av to determine the exposure, took the one on the left with it set to tungsten, and the one on the right with it set to daylight and an 80A filter in front of the lens.

RED


GREEN


BLUE


I don't know about you, but I'm seeing more noise in that left hand side, particularly obviously in the shadows but if you look around its even in the areas which are up around the level of 200 on the white of the roll.

Your work isn't extendable, because it's totally wrong/faulty. (There's another term - right on the spot - for it actually, which I refrain to use in public...)

I agree that this is, hardly a rigorous scientific test, but it does further lend evidence to support my initial suggestion that using a filter helps minimize noise in digital cameras. But let me know if you want to borrow that toilet roll up there ... it could be helpful for you

 

[Loris Medici]

pellicle, I think you're underexposing again; indicated by the fact that you're getting noise in the highlights, which is highly improbable in case of adequate exposure...

Can you please correct your exposure and try again? (In case you don't know: We don't photograph white objects w/o exposure compensation; something like +2 exposure compensation should do well in case you're metering from the illuminated bottom left corner...)

BTW, I do 6 week photo courses at the same institution too... So, as a matter of fact I teach photography.

 

[pellicle]

pellicle, I think you're underexposing again; indicated by the fact that you're getting noise in the highlights, which is highly improbable in case of adequate exposure...

but as the noise is mainly appearing in the blue channel and my assertion is that without filtration of the light source or blowing the reds that the blue will be under exposed and streched to fit. The streching will of course exacerbate noise.

You know, if the evidence matches the theory ... it makes that theory look more solid.

its possible, that I could have increased exposure. These were JPG not RAW. But then I didn't meter off the white, and the scene did have a large amount of black, so the metering should be ok. I also was getting values in the white like 200 so its right up there for a textured exposure of white ... 255 is blowout This was also 100ISO ... where I'd expect more tolerance to noise.

Can you please correct your exposure and try again?

well I was initially going to wait till you publish something ... I've already provided supplemental support for my case ...

couple of days right?

Easter is coming up (not sure if there are any festivals or religious feasts in your part of the world) so I'll be away after Thursday.