White Balance Tint full-scale not enough for infrared photography

[Eosforon]

This should be a very easy thing to fix up. It is related to infrared photography WB.
The full-scale of the Tint adjustment was higher than 10 in a RT previous version (2.4 if I remember well) maybe 16 instead of 10 and it was useful to white balance nIR photography too.

This screenshot was taken with RT 5.5, since from RT 3.0 on the Tint full-scale has been reduced to 10, today using the "pick WB" tool a purplish cast still remains on the nIR photo because Tint value cannot increase higher than 10.

[Thanatomanic]

I think this may be the only use case where the tint needs to be this large.

For all practical purposes the tint value for my photographs are always between 0.9 and 1.1, or something around 1 at least. Even with my recent foray into UniWB photos. I actually find the wb tint one of the few sliders in RT that is VERY annoying to fine-tune (even with the click-long-and-drag mechanic).
So, my point is, I would not like a decision to simply extend the range again. Maybe we need a toggle for a large range or small range, or maybe we need a very non-linear scale.

[Thanatomanic]

@Eosforon, could you maybe share your raw file for some testing?

[Entropy512]

While I almost never bother to postprocess my IR images (they're only for rough engineering documentation at work), I do have access to an IR-converted camera so might be able to play around and do some investigations in late September.

Making the adjustment nonlinear might be the best solution (and might even improve a "typical" use case?) but the question then becomes what the best nonlinear function for this particular use case is. I'm GUESSING that a variation on EV shift (2^x) would likely do well? There is the challenge of handling the UI aspect of things - since users are used to seeing a linear value.

[Beep6581]

For most users it is unpractical having a slider with a gigantic range where a slight slider adjustment results in massive changes. See #2836

The temp slider is non-linear, the tint could be made non-linear too, outside of the practical range.

[sguyader]

Maybe adding some sort of "extend tint value" slider which could be toggled on (but being off by default)?
This slider could add to, or multiply the initially set tint value?

[Entropy512]

Hmm... Maybe. A checkbox seems to be not the ideal solution to me... But honestly after further thought, a nonlinear slider centered around "normal" use cases might also be problematic for people with IR photography needs...

I had a bit of a light bulb go off at the brewery tonight. (warning: Due to its origin, this light bulb might have very bad CRI... :) )

Is there any precedence for one image processing operator to have an effect on another one's settings?

The discussion on pixls recently regarding the automatic scaling of the film negative inversion work of @rom9 made me think - this is a VERY similar use case - a "not standard by any means" workflow that requires highly atypical white balance prescaling. Would it be possible to have a "coarse white balance prescale" operator that could be set by the film negative inversion operator on demand OR could be set manually by a user working on IR images? (or potentially eventually there might be some profile presets depending on the exact IR-pass filter used and the camera's CFA behaviors?)

[Thanatomanic]

@Entropy512 I am not sure what your argument is against a non linear slider. If it is necessary to set a proper wb for processing negatives, then surely having a slider that can be set accurately on a larger range of values is beneficial?

[Entropy512]

So a nonlinear slider will solve the issue of having insufficient adjustment range to handle "weird" behaviors, but may wind up resulting in the slider being too coarse in the new extended adjustment areas - which while better than not being able to hit that adjustment value at all, might still have the issue of being difficult to adjust properly in those use cases. e.g. I'm not sure if it COULD be set accurately across the entire range of values, or only accurately in a range near what is "typical".

To avoid these difficulties (insufficient adjustment range, white balance settings in nonintuitive positions), the negative inversion module in addition to the exponents calculated from the white/black point selection also attempts to do a prescaling of the values, which has been recently discussed as being a potential target for further behavioral optimization in https://discuss.pixls.us/t/any-interest-in-a-film-negative-feature-in-rt/12569/75

If those values were broken into a separate adjustment set, they could also be used for handling coarse prescaling of WB multipliers for IR images.

e.g instead of trying to have "one slider fits all" - have a coarse adjustment slider that's only used when it's really needed (negative inversion and IR images being the current use cases I can think of)

Or possibly there is some way to have one slider handle all situations (both typical and atypical) equally. possibly a slider with an EV-like behavior might handle both?

[Beep6581]

may wind up resulting in the slider being too coarse in the new extended adjustment areas

Can you demonstrate that? When extremes are needed, fine increments make very little difference. The non-linear slider is an elegant solution, it has my vote unless you can demonstrate that the same accuracy is required in the extreme range as in the usual range.

[Entropy512]

I won't have a chance to do that until after September 15 or so.

@Eosforon What sort of IR configuration are you shooting? Full-spectrum, or with a visible-block filter? If visible-block, what wavelength? If visible-block, is your primary goal getting it to appear white?

I can only really collect examples of 850nm visible-block.

[Eosforon]

Sorry for the long time no reply but I am in PRC at the moment and it is not easy to reply here. I agree a logaritmic scale could be a good compromise, I am not worried about fine increaments with slider as long as it would be like the Temperature adjustment. I could anyway write finely the exact number if I need for more little improvements.

My setup is a Powershot A590IS with DIY filter (unexposed film) so it is a visible blocking with half-cut more or less at 800 nm, I have succeded to upload a DNG sample at the following link:
CRW_1198.DNG

[Entropy512]

I'll take a look at your DNG when I get back from vacation.

Since an 800nm visible-blocking filter effectively eliminates color, have you tried desaturating the image as a workaround? You say "DIY filter" - but have you removed the hot mirror from the camera?

As I mentioned, I have access to an A6500 which was full-spectrum converted by Kolari, which can then have an 850nm visible-block added. I'll try an image or two from this when I get back.

[Eosforon]

Yes, obviously in a converted camera like this one the IR-cut standard filter is replaced with an IR-pass filter: so basically you block visible light and let nIR light pass throught.

With 800 nm half-cut some false colors could still be recovered: their amount could vary depending on the camera, since the bayer filters are not all exactly the same.
Anyway with that camera A590IS the problem with RT tint adjustement was the same, even with higher pass cut. I have used 680-720-760 nm in the past on that camera and the WB had the same full-scale limit. I have another standard A590IS (no filter modification) and I could provide shots if I find some time. In the meanwhile I show the same image as in the beginning but with the spot WB done by Darktable, with absolutely no purplish cast.

[Eosforon]

I have done some more testing with the standard Powershot A590IS (no filter modification) using a nIR band pass filter with 720 nm half-cut.


Even in this case Rawtherapee (first screenshot) is not able to properly adjust the WB since the full-scale is not enough, compared to the perfect adjustment done by Darktable (second screenshot).

I have found a result very similar to Powershot A590IS compact using Pentax K-30 DSRL so the problem is likely to be similar for most of infrared photographer no matter the sensor size, IR cut filter and brand. Rawtherapee at the moment is not the best choice for infrared photography. Here some examples with 720 nm and 760 nm filters compared to visible with both cameras (6 shots).

[Dumbphone1901]

Wondering if this thread has produced anything new since 2019? I am also looking at rawtherapee for IR processing, mostly as a replacement for PS (ACR). I am running into the same limitations with the green/magenta slider as the original poster. I use a full spectrum Nikon D200 with various filters. That particular camera has a very wide tolerance for in camera WB profiles so is ideal for IR.

ACR has a similar problem with WB as it ignores the in camera WB settings but ACR does have a convoluted method of creating a different camera profile which will bring the Green/magenta scale in to a usable range. A range extension as mentioned by others for rawtherapee should work well.

Nikon's Capture NX works pretty well at retaining the camera settings for WB but changing any WB settings throws the whole scale into confusion when trying to tweak settings and, of course, won't work for anything but Nikon files. Affinity carries a lot of the in camera settings but loses/gains things along the way. Switching to adjusting mode to tweek WB in Affinity brings up a similar mass confusion with scales and the same shortcomings in green/magenta scale as everyone else.

For now I can use NX Capture but I have to exit with 16bit tiffs and I don't trust Nikon for continuity (especially for anything that works in IR as that is not in Nikon's program). Rawtherapee looks to have much better raw fine tuning than any of the others and I think would be ideal for IR save for the range exclusion problems.

Any suggestions?

[Dumbphone1901]

Should have mentioned above that the problems detailed are with false color IR using filters at 720nm and below. For 830nm b&w there is enough range in both Blue/yellow and mag/green to get rid of any color cast.

[TechXavAL]

I haven't been able to download the raw samples (link expired), but as a guess: would the channel mixer give a solution to this problem?

In the first image of this thread (even not being a raw file), if you slide the green slider on the green channel to the right, the image gets almost b&w. And as stated in Rawpedia, you can play with the tint slider afterwards.

Just a suggestion

[shrx]

Similarly, the 1500 K minimum color temperature is not enough for some IR spectrum photos, 1000 K or even less is needed.