IRG with one shot using a DB850 + Tiffen#12

[Andy Perrin]

I have been noticing for some time now that when I take a photo with Tiffen#12 filter and try to process it by doing the (nominally) IRG-like transformation

new Red = Blue

new Green = Red - Blue

new Blue = Green - Blue

followed by white balance, which allegedly works because the blue channel consists entirely of IR and the Red and Green channels contain equal amounts of IR as the Blue channel, I do not get a photo that exactly matches what one would arrive at by combining infrared, red, and green photos shot separately.

 

LATE BREAKING NOTE!

When doing the above transformation, you MUST be using the unaltered RAW values for the channels. I did allow PhotoNinja to demosaic the image first, but there was no gamma, WB, exposure adjustments, etc. applied prior to the transformation above. ONLY demosaic. (It might be best to not even do that, but I don't have RAW Therapee or whatever.)

 

 

It occurred to me that one reason for this may be that the 700-800nm region contains light that slightly pollutes the blue channel and the others unequally. I happened to stumble on the following Midwest Optics dual-band visible+IR filter the other day and it hit me that this could be a remedy:

This filter is missing the region from 650nm to 800nm, exactly the troublesome region if my guess about the problem is correct.

 

Combined with the Tiffen#12 filter, one gets a combined spectrum that looks (very approximately) like the black line here:

(That graph is based on hand-digitizing the Midwest Optics curve and the Tiffen#12 curve and multiplying them. Some extrapolation was also done where I had no data. So this should not be taken as numerically accurate, rather, it is meant to convey the idea of what I was trying to do here.)

 

I bought one, and it came yesterday, so I did a quick experiment with a leaf where I first assembled an IRG image by hand, and then using the DB850+Tiffen#12 combination.

 

IRG "by hand," using an Omega 830DF30 for the infrared, and the red and green channels from a Hoya UVIR cut + BG38 2mm stack:

 

IRG by the DB850+Tiffen#12 stack (one shot):

 

The colors looked reasonable (in fact it is the "by hand" shot that's the weaker of the two, due to misalignments) so I went off to the reservoir and downtown to see what I could see. Here are some samples. I didn't do any processing outside of white balancing and exposure/contrast adjustments aside from the above transformation.

 

Waterworks

 

Gas station, showing some red->yellow color shifts

 

Street scene

[Cadmium]

Andy, Where did you get your DB850? I could not find anywhere to actually purchase the MIDOPT filters.

[Andy Perrin]

Cadmium, you have to call them up and order on the phone. A pain, but the woman on the other end was really nice and quite efficient. The filter comes in two sizes, one is 25.4mm and the other is 25.5mm (yes, just 0.1mm difference...) and you have to specify. There is a $20 difference for that 0.1mm! The cheaper one (25.5mm) is $60+tax etc. The woman said you can special order other diameters, but the price scales accordingly.

[Cadmium]

Thanks Andy, Any info on getting larger sizes?

[Andy Perrin]

No, I settled for what they had in stock. However you could inquire. These are dichroic of course.

[Cadmium]

Yep, will do, thanks Andy.

[Andrea B.]

Nice discovery, Andy!

[Cadmium]

I got a MIDOPT DB850 also, here I show the out of camera result from the DB850 + Schott OG550 (similar~equivalent to Tiffen #12).

This shot is just to point out the strong color shift of the DB850 from center to sides and corners.

With this shot you will see the color shift most strongly in the sky, and also lower right corner and up to almost the middle of the right side, where there is some gravel that should be gray, but has a greenish tint.

This would be a problem for me.

Andy, I don't know if you get the color shift with your filter and/or the way you have it mounted.

The filter does exhibit some strong color shift also when viewing it alone while shifting the angle of view.

The DB850 looks entirely clear when viewing it directly straight on, but shifts color at any slight angle, which is common for diachroic coated filters.

I have tested several Omega dual and triple band dichroic filters in the past that had the same color shift problems,

so I was expecting some amount of color shift with this filter.

This shot is not post processed, it is SOOC.

[Andy Perrin]

Cadmium put it on a longer focal length lens. It will show less shift I would think. I used 85mm before. You have to work within the limits of dichroic filters.

[Cadmium]

This of course is true, this is 35mm using a DX camera.

Let me see, woudn't that be even wider on an FX? Or am I thinking backwards again?

I would seldom use a longer lens that that for landscapes.

So for me, the color shift is going to happen.

[Cadmium]

Here are a few shots of a flower, just for fun, nothing here is really a good representation of this filter/stack.

First, I have not resolved the 'anti-white balance' setting that Andy advises for this use, where you strip the RAW image of all white balance by loading the RAW file into Photo Ninja,

turning off all the settings, and saving it to a Tiff file. This process is suppose to remove any in camera white balance. It is hard for me to imagine anything that has 'no' white balance,

but that is something I still have not tried.

Second, if I were to truly make a comparison, then I would need to shoot each shot with and without the DB850, one shot with the DB850 + OG550 and one shot with the OG550 alone.

Then I could get a comparison of what the DB850 is doing for the final result.

 

These images below are just plying around.

 

Straight out of camera, arbitrary in camera white balance.

 

Straight out of camera, Auto levels only.

 

Auto levels + Channel mixer: Red = Blue, Green = Red, Blue = Green

 

More complex multi swap EIR simulation processing, which I usually use for Tiffen #12/OG550 shots alone.

[Cadmium]

Andy, Here is the 35mm/DX landscape processed like you said: Load RAW intro Ninja, un-click all settings, save as TIFF.

It looks very dark and bland. Is this what you mean? Is this what yours look slike that this stage?

 

[Andy Perrin]

Yes, that is how it should look. The next step is to do the channel mixer on it and subtract the blue from the red and green channels, and swap. Then after that you adjust contrast and white balance as the last step.

[Andy Perrin]

Here are step-by-step directions. I did this on a Tiffen 12 pic without the DB850 filter by mistake but it doesn't matter, it is just to illustrate the steps.

 

1) In PhotoNinja, load the file from a RAW. Uncheck all boxes so it looks like this:

This is un-WB. Then save as a TIFF (to keep all 16 bits, which is crucial). Photo should look like this, with boring colors.

 

2) In Photoshop, open the TIFF, and do Channel Mixer as follows:

Obviously you could make an Action that would do all this for you.

 

Photo should now look like this:

 

3) In Photoshop, Auto Contrast the picture.

 

Photo should now look like this:

 

4) In PhotoNinja, do a white balance off whatever you want, or do the above steps on a PTFE piece and use that to get a WB. Adjust Exposure as needed in PN.

WB was taken off the glass table in this example.

 

Photo:

[KaJashey]

I had done some IRG photography just for aesthetics not for accuracy. I would up with a system similar to this though not exactly the same.

 

One thing that really really bothered me is that the Tiffen 12 filter leaks visible blue if the blue is strong enough. Artificial plastic blues were especially bad. Blue tarps, dasani blue, the food lion logo. These blues would cut through the tiffen 12 filter and be registered as IR even though they were not.

[Cadmium]

So the answer then is to use a higher nm longpass filter maybe? I just eye tested your idea, indeed my blue tarp outside and other blue things still look blue to my eyes through #12.

However, RG550 renders those blues black. Those filters are not far apart, #12 is about 420nm as i recall.