Noise vs color temp

[chauncey]

The same smoke image, virtually SOOC, with differing color temp.
The first one with proper temp, ala x-rite passport...6650, -18
2nd one with temp raised to 50,000, +150...why noise difference?

[Mike Buckley]

I'm not sure it's a difference in noise as much as a difference in display of the detail in the background. As an example, notice that the folds in the background are displayed more prominently in the second photo. It seems that the same is true of the texture in the background.

[DanK]

I think Mike has hit the nail on the head. When you change temperature, you are darkening some pixels and lightening others. That will change where this graininess appears.

[Manfred M]

This is a direct impact on the colour temperature you've chosen to "shoot" at and the chromatic noise you are seeing. Getting back to basics; your camera is sensitive to the three primary additive colours; red, green and blue. A lower the colour temperature, the more of the light will have a higher red component and the higher colour temperature light will have a higher blue component.

The blue channel tends to be the noisiest of the three colour channels. I'm not totally sure of what is going on, but suspect that the sensor may be slightly less sensitive to the blue end of the spectrum than the other two channels are and increased gain (amplification) will be required to match the output of the other two channels will account for this additional noise. The green channel has the least amount of noise in part because the Bayer sensor has two green receptors for every blue and red one. The sensor is fairly sensitive to red as well (hence the IR filter in front of your sensor).

When you tell the PP software that the colour temperature is 50,000K, it will be looking for a lot more contribution from the (noisy) blue channel. On top of that, because you are actually shooting closer to "normal" daylight 6650K, the amount of blue in the light will be significantly less than what the PP program would be looking for at a 50,000K WB (i.e. relatively speaking the blue channel will be underexposed). Of course this means that channel would get even more gain (amplification) which will bring out even more noise.

So the result you are getting can be explained by the wildly incorrect WB selection that you have made.


If you look at the examples of this CiC tutorial, run the your cursor over the individual colour channels; you will see that the green channel has the least noise, the red is in the middle and the blue channel has the worst noise characteristics.


https://www.cambridgeincolour.com/tu...ge-noise-2.htm


I hope my explanation makes sense to you...

[chauncey]

As a reminder, my body is a Canon 1Ds3, it has always had more noise than I care for.
In normal outdoor scenarios, I refuse to exceed ISO 400...indoors it's always ISO 100.
I don't generally use much noise reduction in PP...here there was none applied.

There must be something else going on here as I'm seeing the opposite happening Manfred...
red is the noisiest and blue is the least noisiest, regardless of the color temperature.

Of late I've been shooting this smoke in a brightly lit room with the smoke slightly backlit
with the strobe resulting in a background measuring 7-8% in Lightroom.

[Mike Buckley]

I don't generally use much noise reduction in PP...here there was none applied.

Was any applied in-camera?

[Manfred M]

There must be something else going on here as I'm seeing the opposite happening Manfred...
red is the noisiest and blue is the least noisiest, regardless of the color temperature.

Actually that is precisely what you are seeing. We are looking at two different, yet interrelated effects.

1. You've told your camera that you are shooting at a very high (blue end) colour temperature. Your software (camera or otherwise) will assume that the light it is receiving is biased towards the blue / violet end of the spectrum. Remember that each colour channel actually is a separate luminance channel and the data is really gray scale.

2. With WB you are going to try to neutralize the effects of this high colour temperature by adding what amounts to a warming filter (yellows and reds). Remember all WB will do is neutralize the excessive blue content in the light that was recorded as RAW data by your camera.

3. So the software will assume the capture will apply the compliment of blue, which is yellow and I expect there could be a cyan component as well (the compliment of red). So the end image has those red tones you see from the WB process, built up largely from the blue channel data, hence the high level of noise.

[Andrew1]

Noise comes from the electronics in your camera. The characteristics of each transition in the process from photon to photo is somewhat governed simply by the batch of parts that were used. Some are different but still fall within the specifications of the part. Also, red is double the wavelength of blue and is collected at half the rate. If the register of a red sensor has to be amplified at double the amount to get the same luminosity of blue then one would have to expect twice the noise. If that was a known problem though I expect the engineers would have compensated for it within the design? Will later versions of the same camera manufactured with advanced parts not have a noisy red channel? Are the electronics getting old? How many photos had you just taken? Is it within the jpg conversion?

What I'm trying to get to is there are lots of reasons for noise and that any finite explanation of noise at this level is at best a guess. You may be able to tell with some very detailed analysis of the electron signals but that may turn into an entirely different and more expensive hobby. In truth, the Engineers may have seen the source of some of the noise during the build but if it was within the specs of the day, it's fine. This stuff changes by the day.

Go shoot some more of your very nice pictures.

[Loose Canon]

Not to mention Chaunce, that on a more basic note you have significantly increased the overall exposure in post of an already relatively noisy shot, which seems a bit underexposed as it stands, by your white balance “correction”.

This alone will do it. And if you cropped the shot a lot that would also exacerbate the issue.

[Brev00]

I side with the idea that the second picture is brighter. If you raise the brightness on the first picture to create the same brightness as the second, you will probably see more noise. Letting dark images stay dark is an even more effective nr strategy than using specialized nr software. Why is the second image brighter, you ask? I am not learned in the physics of photography, but, in my experience, when I raise the white balance, the increase in the warmer tones tends to raise the brightness of the pic. The sensitive red tones will approach clipping sooner rather than later. You can try it out in almost any pic. Slide the white balance to the right and see what happens to the histogram. The reds will clip. Occasionally, I have a blue intensive pic that leads to the opposite effect: warming the pic will drop the blue brights. That is pretty rare. While the dark tones are masking the noise in the first pic, you may want to do some nr there before fiddling with the white balance.

By the way, I like both but prefer the first. Not just cleaner in terms of noise, but the contrast is true.

[chauncey]

FWIW...they are 100% crops, 1000 pixels square>there was exposure alteration due to the color
shift> RGB numbers went from 178, 192, 206, to 245, 213, 239, hadn't noticed until pointed out.

The most of the rest of the images shot had numbers ranging from RGB 10-253, sans adjustments,
with generally the same amount of noise. How might I change my technique to reduce noise if
I'm already at ISO 100?

[Manfred M]

FWIW...they are 100% crops, 1000 pixels square>there was exposure alteration due to the color
shift> RGB numbers went from 178, 192, 206, to 245, 213, 239, hadn't noticed until pointed out.

The most of the rest of the images shot had numbers ranging from RGB 10-253, sans adjustments,
with generally the same amount of noise. How might I change my technique to reduce noise if
I'm already at ISO 100?

Short of getting a new camera, I suspect you are stuck with the noise.

Noise is typically more visible in the dark end of the image data (just like you get in night shots) and comes from a number of different electronic sources (sensor noise, current leakage, etc.). You are shooting against a dark background, so the setup for smoke pictures is going to be inherently noisy. If you are as low as you can go with ISO, there is now way to reduce your gain.

ETTR is the common way of moving your exposure to a region that is less sensitive to showing up the noise. You are shooting and getting data from 10 (dark close to pure black) to 253 (for all intents and purposes, pure white), so there is no headroom for you. If you shoot to the right (you can try this) you will blow out highlights, and your smoke will lose detail as you darken up the background.

So the two most obvious ways of dealing with noise are not going to do anything for you.

[Mike Buckley]

FWIW...they are 100% crops...How might I change my technique to reduce noise

Noise reduction software.

[TonyW]

This has been tweaked in Photoshop. The noise has almost completely disappeared. There is bit left around the edges so that they do not get too blurred.



The reason for the increased noise with the red version is simply that in the original the red signal was very weak and when the colour balance was changed by such a large amount the red signal was amplified greatly and hence the noise in the red signal was increased. I don't think there is anything surprising about it.

An alternative way of changing the colour is to use the colour substitution in Photoshop, which does not increase the noise.

[KingsHockeyFan]

How might I change my technique to reduce noise if
I'm already at ISO 100?

Try this.....
Zoom to 100%
In the noise reduction panel of your raw processing software set your luminance noise reduction to max.
Set your color noise reduction to zero.
Slowly increase the color noise reduction to eliminate the red/green color noise
With color noise gone, slowly lower your luminance noise slider to the point where you cant lower anymore without showing luminance noise.

This method has worked for me on many occasions. May be worth a shot.

Here is where I learned this method..http://damiensymonds.blogspot.com/20...reduction.html

[Manfred M]

Noise reduction techniques all work by smearing pixel data (decreasing sharpness to hide the noise). I find that they tend to destroy the fine detail in this kind of work, so found it did not work for me when I did my smoke images. I also did not sharpen as much as I usually would, in order to minimize noise.

I tried using the D90 and the D800 when I shot smoke images and found the D90 produced noisy images whereas with the D800 I could clean up the objectionable noise with just the spot healing tool in Photoshop. I simply used the camera with the better noise characteristics



Conclusion - using noise reduction means trading of resolution for noise. I prefer the sharper looking image.


Just using a curves adjustment can also enhance the image without the need to apply noise reduction. It nicely gets rid of the background noise without affecting texture.

[chauncey]

My goal was to try to do the color alteration within LR...was an big error in judgment.
Better to do it in PS with layers and selections.

I prefer the sharper looking image.

As do I