minimising noise in digital cameras

[pellicle]

Hi

pardon me, but I just was prompted to write this for an explanation elsewhere

http://cjeastwd.blogspot.com/2010/03/noise-reduction.html

might be handy to remind digital folks that filtration works for digital colour as well as film

 

[donbga]

Hi

pardon me, but I just was prompted to write this for an explanation elsewhere

http://cjeastwd.blogspot.com/2010/03/noise-reduction.html

might be handy to remind digital folks that filtration works for digital colour as well as film

I'm not sure I agree with your assessment. A CWB is basically the equivalent to using a CC filter but a CWB for incandescent light (your example) will boost the blue channel.

A CC filter will suppress the R & G channel. Seems to me all other things being equal, under exposing the R & G channel will create noise.

 

[Loris Medici]

I kinda don't buy it too. If you're in a room with incandescent light (2700-2800K or so in case of mundane lightbulbs, right?), it's pretty normal/natural to have a weak blue channel. With digital all that you have to do (in this context) is to give as much as exposure without causing clipping (except for specular highlights) and you're OK.

Any filter that you use will affect image quality in a negative way. I mean there are trade-offs to consider...

 

[pellicle]

Don

I'm not sure I agree with your assessment. A CWB is basically the equivalent to using a CC filter but a CWB for incandescent light (your example) will boost the blue channel.

A CC filter will suppress the R & G channel. Seems to me all other things being equal, under exposing the R & G channel will create noise.

I confess I'm surprised to read that answer from you. The thing is one is not underexposing all the channels, one is then able to then better expose the channels allowing all three to have better signals (and not just over drive one to bring the other up).

For a film user I'd have thought this was intuitive. Are your slides under exposed when you use a 80A?

 

[nsouto]

Folks, allow me to interject here:

there is no such thing as "boosting a colour channel" in digital.

The amplification applied to the signal of any given pixel does *not* change: it's always the same!

What changes is the "curve" that is applied to the colour channel, by the camera firmware.

Or the amount of light hitting that channel, from your settings of aperture and shutter speed, through any filter/lens combo you might use.


Changing camera white balance does essentially nothing to the sensor or its read-amplifiers, it just changes how the sensel information is processed into a raw file.

Yes, there is processing of sensor information to make a raw file: it's what the camera firmware does. Plu of course processing to jpg, if that is the case.


The degradation of light caused by a filter will very likely be less than what is caused by noise increase when different colour curves are applied to a fixed amount of light.

To keep the amount of light constant when adding a filter, one increases the exposure by changing aperture/shutter. Just as with film.

Most digital sensors have the native equivalent of a 200ISO sensitivity. That does NOT change! What changes is the curve applied to the resulting analog-to-digital conversion.


The common saying that "with digital one doesn't need CC filters" is just about as bone-headed as so many others one hears from the digital crowd.


Of course one may use the white balance to obtain an effect similar to a CC filter.
But don't even imagine for a second that comes as a "boost" of anything electronic!

Either the camera has enough bits in the A/D converter to "stretch" the result or you get posterization. And the lower the bottom of the digital range to "stretch", the more it will be messed up by noise in the read amplifier before it even reaches the A/D converter. Simple as that.


Once again, to be very clear: changing ANY setting in a digital camera other than aperture or shutter speed does not change in any way the amount of light captured or the sensitivity of the sensor to that light.


Any other setting other than aperture or shutter speed just changes how the sensel information is transformed by the camera firmware into a raw image.


Which in turn one can manipulate further, or course.

 

[Loris Medici]

A side issue: Actually I don't get what's so objectionable with "noise". I mean, yes, noise is no good, but let's be reasonable; the advanced amateur digital SLR I have, when used at ISO 1600 gives me much much less noise than what I get (in form of graininess) from 35mm ISO 400 B&W film! Therefore, to me the subject actually borders in the zone of nitpicking/pixel-peeping; pretty much a non-productive activity...

Regards,
Loris.

...
The degradation of light caused by a filter will very likely be less than what is caused by noise increase when different colour curves are applied to a fixed amount of light.
...

 

[donbga]

Don



I confess I'm surprised to read that answer from you. The thing is one is not underexposing all the channels, one is then able to then better expose the channels allowing all three to have better signals (and not just over drive one to bring the other up).

For a film user I'd have thought this was intuitive. Are your slides under exposed when you use a 80A?

Which is why I wrote:

Seems to me all other things being equal,

Don

 

[donbga]

Folks, allow me to interject here:

there is no such thing as "boosting a colour channel" in digital.

The amplification applied to the signal of any given pixel does *not* change: it's always the same!

What changes is the "curve" that is applied to the colour channel, by the camera firmware.

Or the amount of light hitting that channel, from your settings of aperture and shutter speed, through any filter/lens combo you might use.


Changing camera white balance does essentially nothing to the sensor or its read-amplifiers, it just changes how the sensel information is processed into a raw file.

Yes, there is processing of sensor information to make a raw file: it's what the camera firmware does. Plu of course processing to jpg, if that is the case.


The degradation of light caused by a filter will very likely be less than what is caused by noise increase when different colour curves are applied to a fixed amount of light.

To keep the amount of light constant when adding a filter, one increases the exposure by changing aperture/shutter. Just as with film.

Most digital sensors have the native equivalent of a 200ISO sensitivity. That does NOT change! What changes is the curve applied to the resulting analog-to-digital conversion.


The common saying that "with digital one doesn't need CC filters" is just about as bone-headed as so many others one hears from the digital crowd.


Of course one may use the white balance to obtain an effect similar to a CC filter.
But don't even imagine for a second that comes as a "boost" of anything electronic!

Either the camera has enough bits in the A/D converter to "stretch" the result or you get posterization. And the lower the bottom of the digital range to "stretch", the more it will be messed up by noise in the read amplifier before it even reaches the A/D converter. Simple as that.


Once again, to be very clear: changing ANY setting in a digital camera other than aperture or shutter speed does not change in any way the amount of light captured or the sensitivity of the sensor to that light.


Any other setting other than aperture or shutter speed just changes how the sensel information is transformed by the camera firmware into a raw image.


Which in turn one can manipulate further, or course.

Honestly folks the key to this is just to shoot RAW or do a CWB. No CC filter needed. In some ways my JPGS come out 'cleaner' than converted RAW files (though that is a function of the RAW conversion software as well.)

I think we have a Tempest in a Teapot here.

As Loris said, film grain from fast film is often more significant than digital noise at higher ISOs.

BTW, has anyone looked at the high ISO output from the Nikon D700s and D3s? Real impressive. Same with the new Canon 5DMkII.

But heck who would have ever thought one could routinely shoot at ISOs like 3200 or 6400 and have something useable?

Don

 

[ann]

as an aside, i recently shot some photos in a tango bar so dark one had to use a flashlight to read the menu (true).

the dancers were lite from a couple of over head lights (this was not a las vegas type bar, just a neighbor hood bar in Buenos Aries)
Must seat 15 at the most.

Anyway, shooting at an ISO of 6400 with a d700 was shocking. At 1/10 sec. the place looked like it was lite professionally and noise? What noise? one really had to take out a loupe to find any.

 

[nsouto]

Just to follow on my post, I suggest folks read today's post by Ctein in The Online Photographer:

http://theonlinephotographer.typepa...grapher/2010/03/what-tests-dont-tell-you.html

Remember what I said about getting "posterization" effects when pushing the ISO in a camera with not enough internal bits to "stretch"? Where do you think that "banding" he talks about is coming from?

The fact that there are dslrs with usable ISO of 6400 and higher, changes absolutely nothing of what I or pellicle said.
In fact if anything, it opens up the possibility of using that extra ISO range to counter the effect of using a filter: you can use even a strong one and still get a usable image.

And please: high ISO is not a synonym for CWB or colour temperature, the two things are not related.

 

[Loris Medici]

ISO 1600 digital vs. ISO 400 B&W film example's purpose was to point out how the noise levels are low and negligible (unless you're pixel-peeping) with current digital cameras. OP's point is moot in case you're using a current digital SLR (because they don't give much noise to begin with), OTOH it could be an issue for older (about 2 generations back) models or compact cameras with tiny tiny sensors - which in fact aren't capable of giving anything good at and above ISO 400 - indeed...

P.S. BTW, Ctein's blog is as irrelevant as it can be...

 

[donbga]

Just to follow on my post, I suggest folks read today's post by Ctein in The Online Photographer:

http://theonlinephotographer.typepa...grapher/2010/03/what-tests-dont-tell-you.html

Remember what I said about getting "posterization" effects when pushing the ISO in a camera with not enough internal bits to "stretch"? Where do you think that "banding" he talks about is coming from?

The fact that there are dslrs with usable ISO of 6400 and higher, changes absolutely nothing of what I or pellicle said.
In fact if anything, it opens up the possibility of using that extra ISO range to counter the effect of using a filter: you can use even a strong one and still get a usable image.

What did you or Chris say regarding high ISO relative to his OP that has anything to do with posterization or banding as written about by Ctein. I would suggest that if you are having these kinds of issues you need to modernize your digital camera gear or learn how to expose properly.

And please: high ISO is not a synonym for CWB or colour temperature, the two things are not related.

Whoever suggested it was?

Your post lacks coherence and intellectual rigor.

Don

 

[nsouto]

@Loris:

The OP was talking about a very modern digital camera and the problem is common to almost all of them.

In case no one has noticed high-ISO capable cameras are not typical "current dslr cameras": they are less than a handful, compared to all other cameras in the market. Therefore they are the exception, not the rule in the digital market.

Ctein's blog is as relevant as can be: it is PRECISELY the high ISO banding he describes - in a very modern camera! - that is caused by the excessive stretching in one or more of the colour channels. The resulting image effect is that banding. Note that the effect disappears as soon as the ISO drops from maximum. I can't provide a better example of how appropriate that makes his post.

@Don:

When I am talking about CWB and the replies go the way of how high ISO works around all problems, the replies denote a complete mis-understanding of what high ISO is and what CWB is.

As for your other coments: sorry Don, but you got no clue what you are talking about.

I suggest you re-read carefully what I wrote in the context I wrote it. It's got nothing to do with film or film grain, nor is that ever in any question.

If that is too hard, then I strongly suggest you refrain from commenting in a derogatory manner.

As simple as that.

This is not the Usenet and I don't have to take that sort of verbal crap from you or anyone else here.

In case that is not clear: I did not make ANY derogatory statement.
YOU did.
Period.

 

[donbga]

@Don:

When I am talking about CWB and the replies go the way of how high ISO works around all problems, the replies denote a complete mis-understanding of what high ISO is and what CWB is.

As for your other coments: sorry Don, but you got no clue what you are talking about.

I suggest you re-read carefully what I wrote in the context I wrote it. It's got nothing to do with film or film grain, nor is that ever in any question.

If that is too hard, then I strongly suggest you refrain from commenting in a derogatory manner.

As simple as that.

This is not the Usenet and I don't have to take that sort of verbal crap from you or anyone else here.

In case that is not clear: I did not make ANY derogatory statement.
YOU did.
Period.

What ever!

 

[Loris Medici]

Ditto :rolleyes:

What ever!

 

[pellicle]

I have resisted adding anything further in this thread because it seemed that the replies were generally of a very low standard of argument, essentially not addressing my points in any way and attempting to say only:

• I don't think so (with zero supporting evidence)
• that's wrong (with again no supporting evidence)
• but my pictures are so good

clearly cameras such as the Nikon 3D with impressive high ISO 14bit signal processing and very high sensor area (low pixel density) may not benefit in a way which translates on a print. Indeed the reduction of noise may not in itself be any ones goal, but then my post was not entitled "do it my way" it was about minimising noise.

As it happens I use a Panasonic G1, which is a micro 4/3 camera. As it happens these cameras (and those with smaller sensors) do generate large amounts of noise in 800ISO and 1600ISO. Thus it may be of advantage to do anything to clean this up. I say may because in some situations it may not.

I am replying to this because I witnessed another tread which is an epic example of pathetic behavior over on another forum here.

It occurs to me that these threads are found by search engines (such as google) and remain as a resource for others (whom may only read and not participate) in the future. This is after all what the sponsors of this site seek to sponsor ... or do you think they are putting their money into making some sort of "its a knockout" mud slinging match?

Perhaps it is only me, but I regard HybridPhoto as a forum where things are discussed on an intellectual level, which ultimately attracts readers to see what is here and value it. Descent into did too did not argument does nothing to further the intellectual rigor of the participant who stoops to that nor does "I don't think so" replies with zero backing and zero attempt to demonstrate why the evidence offered in support of a theory was either wrong / flawed / mis-applied / irrelevent / an artifact of some conditions.

So before standing on a box and saying "your argument lacks "intellectual rigor", why not apply that to any response and lead by example.

naturally I expect did not from the peanut gallery at this point.

 

[Loris Medici]

pellicle,

First; nobody says one should definitely use high ISO values. That's mainly nsuoto's mistake; he was (somehow) stuck to the idea that I have suggested shooting at high ISO, whereas all I was trying to point out the fact that "noise levels are very low / negligible with current SLR digital cameras" (be it an entry level camera or something better)... Therefore there's nothing much to gain by following your suggestion. And please note the negative effects of using a USD 50 filter over a USD 1500 lens (exaggerated example for sake of clarity); you'll loose micro-contrast / sharpness / contrast.

A very important point (the main point if you like) is: Actually using a CC filter will force you to use higher ISO (depending on shooting conditions and requirements), therefore will result with noisier images, since "there's about 2 stops filter factor for a 80A filter", pretty much canceling out the "alleged" advantage / purpose of using a CC filter.

I have a 80A CC filter, I can take the trouble of supporting my claims above with visuals if you like / in case you're not convinced...

P.S. I still think in-line with Don; the main subject is kind of a tempest in a teapot. But thanks anyway for the thinking you have done, I appreciate that...

 

[donbga]

I have resisted adding anything further in this thread because it seemed that the replies were generally of a very low standard of argument, essentially not addressing my points in any way and attempting to say only:

• I don't think so (with zero supporting evidence)
• that's wrong (with again no supporting evidence)
• but my pictures are so good

clearly cameras such as the Nikon 3D with impressive high ISO 14bit signal processing and very high sensor area (low pixel density) may not benefit in a way which translates on a print. Indeed the reduction of noise may not in itself be any ones goal, but then my post was not entitled "do it my way" it was about minimising noise.

As it happens I use a Panasonic G1, which is a micro 4/3 camera. As it happens these cameras (and those with smaller sensors) do generate large amounts of noise in 800ISO and 1600ISO. Thus it may be of advantage to do anything to clean this up. I say may because in some situations it may not.

I am replying to this because I witnessed another tread which is an epic example of pathetic behavior over on another forum here.

It occurs to me that these threads are found by search engines (such as google) and remain as a resource for others (whom may only read and not participate) in the future. This is after all what the sponsors of this site seek to sponsor ... or do you think they are putting their money into making some sort of "its a knockout" mud slinging match?

Perhaps it is only me, but I regard HybridPhoto as a forum where things are discussed on an intellectual level, which ultimately attracts readers to see what is here and value it. Descent into did too did not argument does nothing to further the intellectual rigor of the participant who stoops to that nor does "I don't think so" replies with zero backing and zero attempt to demonstrate why the evidence offered in support of a theory was either wrong / flawed / mis-applied / irrelevent / an artifact of some conditions.

So before standing on a box and saying "your argument lacks "intellectual rigor", why not apply that to any response and lead by example.

naturally I expect did not from the peanut gallery at this point.

Chris,

A very measured and cogent response! Good work.

Don

 

[nsouto]

pellicle,

First; nobody says one should definitely use high ISO values. That's mainly nsuoto's mistake; he was (somehow) stuck to the idea that I have suggested shooting at high ISO, whereas all I was trying to point out the fact that "noise levels are very low / negligible with current SLR digital cameras" (be it an entry level camera or something better)... Therefore there's nothing much to gain by following your suggestion. And please note the negative effects of using a USD 50 filter over a USD 1500 lens (exaggerated example for sake of clarity); you'll loose micro-contrast / sharpness / contrast.

A very important point (the main point if you like) is: Actually using a CC filter will force you to use higher ISO (depending on shooting conditions and requirements), therefore will result with noisier images, since "there's about 2 stops filter factor for a 80A filter", pretty much canceling out the "alleged" advantage / purpose of using a CC filter.

I have a 80A CC filter, I can take the trouble of supporting my claims above with visuals if you like / in case you're not convinced...

P.S. I still think in-line with Don; the main subject is kind of a tempest in a teapot. But thanks anyway for the thinking you have done, I appreciate that...

Loris: I was very careful to quote your name properly. Is it too much to ask that you make the smallest possible effort to READ and quote my alias properly? Or is that too much of an intellectual effort for you? If so, then please simply cut and paste, I trust that is not outside your ability?


Once again you miss the point. When light levels drop - be it using a filter, a drop in the general solar output, whatever - one does NOT increase the ISO as an automatic reaction. The proper and correct action is to increase the exposure by means of aperture or speed changes. Increasing the ISO is the LAST thing one should do. Regardless of whatever "modern" dslr one is using and whatever ISO it can shoot at, which is totally irrelevant.

Therefore, your "important point" that using a CC "forces" one to use a higher ISO is completely bogus and incoherent.

I assume you are familiar with lens aperture and shutter speed? After all even on a "modern" dslr, those are still available...

 

[cupcake_ham]

What ever!

Don, the use of filters in digital capture in architectural photography is commonly accepted for the reasons the OP put forward. He is indeed correct about balancing of the channels in order to avoid underexposure in any.

As some of the best photographers out there in the architectural field have commented on this and have noticed the benefits....as have I for general real estate interior photography, I can tell you that it works.

And nsouto is correct in proposing a longer exposure than bumping up iso. I don't think anyone here changes their iso constantly outside instead of their aperture or exposure time.

 

[donbga]

Don, the use of filters in digital capture in architectural photography is commonly accepted for the reasons the OP put forward. He is indeed correct about balancing of the channels in order to avoid underexposure in any.

As some of the best photographers out there in the architectural field have commented on this and have noticed the benefits....as have I for general real estate interior photography, I can tell you that it works.

And nsouto is correct in proposing a longer exposure than bumping up iso. I don't think anyone here changes their iso constantly outside instead of their aperture or exposure time.

What ever! :rolleyes:

 

[Bob Carnie]

I seem to be at the back end of this discussion .
I do not use a digital camera, but I see a lot of captures coming my way from various sources, film scan, small capture size devices, right up to phase and betterlight scan backs.
One of my primary work functions is converting to black and white and then making archival prints via a lambda RGB laser printer onto fibre paper which we then process in trays.

The biggest problem that I encounter , is noise in a channel that basically causes problem's in the print.
I am exposing at 400ppi onto a high silver content black and white fibre wet paper with low Dmin /high Dmax values( A lower PPI will not work in my enviournment], that can show faults that my RA4 and Inkjet output do not show.
It has in the past perplexed me why this happened and I attribute it to lower ppi output of the other two medias as well the spreading of ink or how the dyes couplers action works. Basically the Higher the ppi and size stretches the file too much and shows flaws quicker.**kind of like the argument of editing in 16bit over 8 bit which gives you more headroom but only in reverse.***

I have been on my own on this for 6 - 7 years now as other Labs that offer this service , do not seem to like to talk with each other or are hesitant. To date I have made a few thousand prints from various file capture and this problem is there , most of the time. This has forced me to deal with noise as a very real problem.

To date I have resorted to channel blending , with apply image , picking up the good channels, getting rid of the bad, or blurring the bad and then applying a 60/30/10 kind of split, usually the blue is the ugly bride, but not always.
Once the apply image work is done I will pull up the strongest channel and curve shape the image to a decent looking black and white on screen then convert to greyscale from this point.

I stopped using the filters in the BW conversion that CS3 and CS4 have as this method does nothing to eliminate the noise that is hidden within the image.

To date I have not found the golden trick to eliminate noise from supplied files, but this thread should help me understand what my clients go through before they submit files to me . I would love to be able to give advice to them so lets keep this going and forget the bitching.

remember I look at this not as a photographer but one who gets the results of their capture and some interesting observations seem to be coming out from various participants.

 

[Loris Medici]

Loris: I was very careful to quote your name properly. Is it too much to ask that you make the smallest possible effort to READ and quote my alias properly? Or is that too much of an intellectual effort for you? If so, then please simply cut and paste, I trust that is not outside your ability?

It's more like I don't care actually. (Not in general, but very true in your case.) In any case sorry for the inconvenience...

Note: BTW, I don't like the fact that people prefer to be referred by their aliases instead of showing the least possible moral courage of using their real names in such forums...

Once again you miss the point. When light levels drop - be it using a filter, a drop in the general solar output, whatever - one does NOT increase the ISO as an automatic reaction. The proper and correct action is to increase the exposure by means of aperture or speed changes.

Very true, that's what I said in my first post, would you take the "intellectual trouble" to READ it again? See the specific part below:

"...With digital all that you have to do (in this context) is to give as much as exposure without causing clipping (except for specular highlights) and you're OK..."

Increasing the ISO is the LAST thing one should do. Regardless of whatever "modern" dslr one is using and whatever ISO it can shoot at, which is totally irrelevant.

Therefore, your "important point" that using a CC "forces" one to use a higher ISO is completely bogus and incoherent.

I assume you are familiar with lens aperture and shutter speed? After all even on a "modern" dslr, those are still available...

I strongly suggest that you take one little step above elementary aperture/shutter/exposure issues, and read something about "filter factor"...

P.S. I feel I'll can't stand another belch full of nonsense from you (sorry for being too frank, but that's what I am and unfortunately, I can't change that anymore at my age...), therefore I again strongly suggest to re-READ what I've said before, quoted below:

"...Actually using a CC filter will force you to use higher ISO (depending on shooting conditions and requirements)..."

N.B. the bold underlined section.

Adis...

 

[pellicle]

What ever! :rolleyes:

It's more like I don't care actually. (Not in general, but very true in your case.) In any case sorry for the inconvenience...

I think that both Don and Loris have made their positions clear on this.

Don, I personally would take it that as your role is moderator on this forum that you coul kindly refrain from making the sorts of comments which lend nothing to the discussion and lead to exactly the sort of argument which requires moderation.

Further it does nothing for the standard of this forum if people who are official on this forum act in a trite manner.

 

[Mike1234]

The OP is right. I know from experience that noise is generated when extreme color corrections are made to any image. This occurs even if the corrections are made in the camera. Correct filtration with appropriate exposure compensation and shooting at the lowest ISO possible minimizes post processing and noise. This is true for any camera regardless of bit depth.