Filter choice for Infrared/red/green images. From yellow to orange : how to make sure my Blue channel is empty of visible light.

[Fedia]

I've been doing Infrared/red/green images lately. I'm using and orange fitler (550nm cutoff) + the DB850 right now. I get good results but I am constantly wondering if I shouldn't let more Green light in to get richer colors.

 

So I made a little experiment with my unconverted Canon 600D (which i think has the same sensor as my converted 1200D) to test wether or not my blue channel was empty depending on the filters I placed in front of the lens.

 

But before I was able to compare any filters there was this strange phenomenon occuring with the histogram : It showed completely different wether I looked at it in the live view of once the photo was taken in the review menu :

 

Here the camera is pointed at a neutral grey surface with the 550nm filter on. White balance is . This is live view.

 

 

On the histogram the blue channel doesn't appear to be empty at all, I repeat this is an unconoverted camera, so it should be.

 

And this is what the histogram looks like once the photo is taken and is reviewed in the menu :

 

 

Here the blue channel is empty as expected.

 

So if someone has an explanation I would be glad to hear it.

 

Is it that the live view histogram represents the unprocessed channels and that the blue channel has in fact a very wide range of sensitivity and is therefore very difficult to empty ?

 

This would mean that the correct blue primary as seen in the review menu histogram is obtained through a certain amount of substraction of the two other colors ?

 

 

[nfoto]

The histogram is based upon the set w/b from the internal jpg. Even if you are shooting RAW.

 

To see what the RAW file contains *before w/b*, specialised software like RAWDigger has to be used.

[Alaun]

Also the blue channel has some sensitivity in the green and perhaps even into the red.

[Andrea B.]

If you upload a raw file, I will run it through Raw Digger to see what is actually being recorded before white balance is applied.

[Fedia]

4 hours ago, Andrea B. said:

If you upload a raw file, I will run it through Raw Digger to see what is actually being recorded before white balance is applied.

Thanks ! Here is the raw :

_MG_0501.CR2

[Andrea B.]

 

The raw composite of Fedia's file shows the data recorded by the camera before white balance has been applied. The data has otherwise been demosaiced, auto-scaled and gamma-ed by Raw Digger. This raw composite is almost spot on yellow. That is, Red and Green are almost equal and Blue is minimal. Remember that when yellow is darkened it begins to appear to the human eye as a kind of olive green, quite obvious here.

I selected a center portion, averaged the color, and got some color wheel data from PS. I added a dot of showing what this raw color would be if brought up to full saturation and full brightness. It is yellow. (Pure yellow = 60°).

Color analysis like this is always subject to minor error because it is done in 8-bits. Averaging, resizing, saving, sRGB vs. aRGB, monitor color spaces and so on definitely do affect what you see here. But it does give us a very good approximation of what the camera actually recorded.

RAW COMPOSITE:  No white balance applied.

 

 

 

Here is the JPG extracted from the raw CR2 file together with its color analysis. The camera settings produced an Orange color. (Pure orange is usually considered to be 30°). The same cautions apply about color accuracy as noted above. (I think I forgot to deal with the color profile. This has aRGB. I hope it appears properly in everyone's browser. I can redo this if needed.) Note that the camera settings have affected the raw data by pushing the average color towards red. Camera WB was set to daylight. The JPG extraction made in Photo Mechanic.

JPG EXTRACTED from RAW FILE

 

 

 

Here are screen shot JPGs from Raw Digger showing each channel. As expected the blue channel is very very dark, almost black.

RED CHANNEL

GREEN CHANNEL

BLUE CHANNEL

 

 

Here is the raw data histogram. (The two green channels were averaged.) Note that the pixels in the raw blue channel seems to be just "noise".

RAW HISTOGRAM

 

 

I screen-shotted Raw Digger's statistical analysis of the pixels for easier viewing. 

STATS

 

 

Curiosity:  I wondered how different the average raw color would be if there were no blue pixels. You can see that the saturation bumps up a little bit.

So I was thinking that if you boost the saturation setting on the camera when making this photo, you might get fewer blue pixels in the raw file? Try that and let us all know if that works.

ADDED LATER:  I mis-applied the process! Increasing saturation in a photo app can eventually produce a 0 indicating 100% saturation. But this trick ain't gonna work in-camera on the raw data, only maybe on the JPG. 

 

With (left) and without (right) the blue pixels.

 

 

 

Please let me know of any errors or typos. I am really bad at editing myself. 

 

 

[Andrea B.]

I should add that your external 550 filter might not be narrow enough or precise enough to completely exclude blue all by itself. All our filters do seem subject to leakages. But maybe the saturation suggestion I made will reduce them.

 

Here is something interesting. The wavelength to RGB color converter shows 550 nm to be a kind of lime green. It translates 580 nm to yellow. But I have no idea how accurate this converter is.

[Andrea B.]

If looking at RGB coordinates, a 0 anywhere indicates 100% saturation and a 255 anywhere indicates 100% brightness.

[ulf]

Rawdigger is an excellent analysis program, but possibly a bit costly for casual usage.

The FastRawViewer from the same company give this result:

 

If viewing and culling of raw images is useful that is a good program.

I use it all the time before converting my images from RAW with RPP64.

https://www.fastrawviewer.com/

[Fedia]

Thanks Andrea, I'm learning a lot with your reply.

 

So you revealed that the raw file showed yellow instead of orange before it is processed but once it is opened in a raw processing software the correct color is recovered automaticaly.

 

My first interrogation about this is : is it frequent for an unprocessed raw file to be off like this in terms of color ? It almost looks like the sensor was natively balanced for much warmer light than sunlight. 

 

My second interrogation is about the histogram on the live view of my camera, which looks nothing like the unprocessed raw file. Compared to the latter, it seems to be just wrong :

 

2 hours ago, Andrea B. said:

RAW HISTOGRAM

 

 

Live view histogram :

On 7/17/2023 at 3:19 PM, Fedia said:

 

 

 

Technically the color it shows is red... I don't know to what this histogram corresponds to

Well maybe it doesn't matter so much, at least now I know from the Raw data that my orange filter is able to empty my blue channel enough so I can isolate IR in it on my full spectrum camera.

 

3 hours ago, Andrea B. said:

I was thinking that if you boost the saturation setting on the camera when making this photo, you might get fewer blue pixels in the raw file?

I didn't know the saturation you set in camera can influence the Raw data. I thought everything you set in the picture profile could only influence the JPEG or be recorded as metadata attached to the raw file.

[Fedia]

2 hours ago, ulf said:

I use it all the time before converting my images from RAW with RPP64.

I'm interested, is it usable for free or do you necessarily have or buy it ?

[Andrea B.]

You are of course correct about my saturation comment as it pertains to a raw file! I will amend that. It's way too hot here & I wasn't thinking. I crossed out the original stoopy comment and added something more reasonable. Q.v.

[ulf]

6 hours ago, Fedia said:

I'm interested, is it usable for free or do you necessarily have or buy it ?

It is clearly stated on the website I linked to you: "fully functional 30-day trial"

 

That means that you if you want to continue to use it after 30 days you have to buy it. 

I do not know how it blocks usage after 30 days.

It might be with ugly watermarks or just that it stops working until you enter a license code.

 

IMHO the ca 30 Euro is well spent money if you like the viewer.

The Raw digger, in the most simple of three modes is even cheaper, but is mostly for analysing histograms in detail. It also have a "fully functional 30-day trial"

There are also package deals with both:

https://www.fastrawviewer.com/purchase

 

For me the functions in FRV are more valuable than just the histogram analysis in RawDigger, as you can change and investigate very many different aspects of the raw file to get an idea of the potential for raw conversion and post processing.

 

If your image processing matures away from the jpegs directly out of the camera, into something more complex, you might eventually also think so.

 

[Andrea B.]

So you revealed that the raw file showed yellow instead of orange before it is processed but once it is opened in a raw processing software the correct color is recovered automaticaly.

 

My first interrogation about this is : is it frequent for an unprocessed raw file to be off like this in terms of color ? It almost looks like the sensor was natively balanced for much warmer light than sunlight. 

 

My second interrogation is about the histogram on the live view of my camera, which looks nothing like the unprocessed raw file. Compared to the latter, it seems to be just wrong.

 

*******************

 

The image you see on the LCD of your camera (either in Live View or when reviewing your shot) has been produced by the camera's JPG engine (internal camera software). The JPG engine first demosaics and scales the raw data, then applies any camera settings you have made such as white balance, saturation, sharpness, sRGB/aRGB etc. The camera histogram does not present the raw data, it presents the JPG data.

 

The raw composite image I showed you from Raw Digger represents a demosaiced/scaled version of the data recorded by the sensor before the JPG engine has applied camera settings. So any raw composite will never look like its corresponding JPG on the camera LCD.

 

When you open a raw file in a photo app such as Photoshop, that app is capable of reading the camera settings stored in the file and applying them to produce a working image which should be very close to the image produced by the camera's JPG engine. However the original settings can now be changed, refined, pushed/pulled, dropped/added, etc. and then re-applied to the original raw data. It is much more difficult to make extensive changes to a JPG because any altered settings are being applied "on top of" already altered data.

 

To re-phrase: Once the dark yellow (olive green) raw file is opened in a photo app, the app immediately reads and applies the settings stored in the file  to produce an orange JPG image which is the same as or almost the same as the one on the camera LCD. 

 

I hope that what I just wrote is neither over-simplified nor too confusing!

[Andrea B.]

Forgot this.

 

The raw data does show that some blue pixels were recorded. The JPG engine probably clips them so you only see a blue line on the left-had wall.(Maybe beccause a contrast setting was applied? Maybe the blue was lost when aRGB or sRGB was applied? There could be several possible explanations.)

[Fedia]

6 hours ago, Andrea B. said:

I hope that what I just wrote is neither over-simplified nor too confusing!

not at all, It is very clear ! Plus it makes sense since changes I make in the picture profile do influence the live view histogram.

[Andy Perrin]

I think Andrea nailed this pretty well for you. The upshot of it all is that we can't trust the in-camera histogram for this kind of work!

 

Every out-of-visible light photographer learns that lesson sooner or later, haha. Along with the fact that It Always Takes More Blocking Than You Expect to see things on the tails of your sensor's sensitivity curve (UV-B, UV-C, 1000nm+ IR...)

[KaJashey]

Andrea is right that the spectral color of the 550nm filter is a sort of green. It is an IR filter and they have somewhat cliff like curves on their frequency cutoffs. The the left the low frequencies pass and to the right the filter is more opaque. The bottom and tops of the cliff is rounded off and not a great place to measure. Where the cliff is at 50% cutoff is where they are measured. This means the filter is 50% opaque at that funky green and some higher frequency light will make it through.

 

The filter is visibly orange because of all the reds and yellow that make it through. the greens and slight blues don't contribute much.

 

I have not looked at histograms or done much tabletop study. I know when I was trying 1 shot IRG pictures I noticed i would get strong blues from plastics registered as blue/IR in my images. Blues from tarps, Dasani water bottle blue, some other ad stuff. Using a yellow plus orange filter stack cut it down but didn't eliminate it.

[Fedia]

2 hours ago, KaJashey said:

the greens and slight blues don't contribute much

The greens contribute, you can still see that the foliage is green when looking through the filter with the naked eye, even with the strong orange cast. But you are right a part of the greens are cut out. That's why I wanted to see if I could use a filter for which the cuttof frequency would be shorter than 550nm. So more a yellow filter than an orange.

 

The test I was trying to conduct was aimed at verifying if the blue channel was still empty of visible light with yellow filters, but I had this histogram problem that came in the way...

 

3 hours ago, KaJashey said:

I noticed i would get strong blues from plastics registered as blue/IR in my images

I have deffinetly observed that ! From different things painted blue, from tarps as you did. But in my opinion these objects reflect Infrared as well and that's why they dont loose their blue color even with an orange filter on. A reliable blue source is actually the sky. If the tarp is still blue but the sky has completely shifted to green it probably means the tarp reflects infrared.

 

Even with a 720nm filter I observed blue objects to appear slightly blue. That's because these objects absorb the shorter wavelength infrared (right next to red) and pass the longer wavelength to which the blue channel is more sensitive.

[KaJashey]

2 hours ago, Fedia said:

I have deffinetly observed that ! From different things painted blue, from tarps as you did. But in my opinion these objects reflect Infrared as well and that's why they dont loose their blue color even with an orange filter on. A reliable blue source is actually the sky. If the tarp is still blue but the sky has completely shifted to green it probably means the tarp reflects infrared.

 

Even with a 720nm filter I observed blue objects to appear slightly blue. That's because these objects absorb the shorter wavelength infrared (right next to red) and pass the longer wavelength to which the blue channel is more sensitive.

 

Smart idea using the sky. One thing i did was test visually with my eye not the camera to see if visual blue passed. I'm a little hazy remembering the results but I think there was a problem.