A challenge for James Ritson: IR lens hotspots

[Jörn Reppenhagen]

I guess almost every IR enthusiast knows that problem: That ugly lens hotspot in the middle of an image.
Common hint: "Use another lens". So I did. And each lens I tried (all Fujifilm, XF 18 - 55, XF 27, XC 50 - 230) showed a hotspot.
Search on the net: Lots of big-mouth "tutorials", none working.
Next common hint: "Convert to black and white." Works if I reduce the red contribution (red after channel swap, blue before channel swap) during B/W conversion - but of course the result is black and white, not what I wished for when using a 720 nm IR filter.
Another common hint: "Try different apertures." Helps a bit in changing the size of the hotspot, but doesn't get rid of it. Plus, I can't tell foliage not to move in the wind during a multiple seconds exposure (sample picture (un)clearly shows the result). Okay, I can tell - but it won't obey or just doesn't listen; not sure, yet.

Thought after channel swap: It's a reddish-orange circle with faded edges, so how about drawing a reddish-orange circle of the same size with a radial filling and faded edges, then substract that from the picture, maybe using a blend mode?
Tried and failed again and again, always ending up with a visual rim around the hotspot, getting extremely prominent with further image processing. Also tried other approaches, always happily hit the wall. Maybe I overlooked something obvious or just screwed up.

So now I am calling fellow IR lover @James Ritson to the rescue.

I guess this is an issue a lot of IR enthusiasts fight with and lose the fight, there's no working tutorial and/or sufficient funds for buying a non-hotspot lens just for IR purposes.
Maybe it could even be possible to implement a "IR lens hotspot removal" feature in future AP versions - IF there's a working solution.
Would love to see that.

---

Two small images attached, before and after channel swap, saturation cranked up for making the hotspot more prominent, plus original image in RAF format for playing around.

DSCF8979.RAF

[Komatös]

Hallo @Jörn Reppenhagen

Vielleicht hilft dir dieser Artikel weiter.

https://kolarivision.com/articles/problem-with-ir/

Edited September 6, 2021 by Komatös
change a link

[NotMyFault]

The lens flare seems to be not to complex.

1. Use channels panel to isolate one channel after the other (red / green / blue)
2. add levels adjustment (or curves) and select one channel
3. add a ellipse shape, filled in white, with Gaussian blur of reasonable radius (tens px or more), to cover the flare area of the selected channel
4. Nest shape to levels adjustment as clipping mask
5. adjust output level to reduce brightness until it matches the surrounding area
6. repeat for other channels.

Did this very quick on blue channel (before swap) only and assume this will lead to excellent results.

Bonus benefit: 

• You may save these adjustments as assets, and reuse them on other images. As long as the lens settings do not change (focus length, aperture) it could work without any further adjustments 

Alternative approach: 

Maybe the new frequency separation blur modes could help, similar to lens flare removal.

 

 

Edited September 6, 2021 by NotMyFault
Several updates

[Old Bruce]

Just some thoughts from my own frustrations with this...

If we had an IR light source we could probably make a series of masks by photographing a blank sheet of paper. But the hot spots are caused by internal reflections so we would need to deal with near and far focusing as well as all zoom lengths too. Be rather weird setting up the camera and focusing then turning off the lights to take the picture.

Perhaps a flash unit with an IR pass filter.

[Jörn Reppenhagen]

Thank you for your answers!

Komatös: Den Text kannte ich leider schon (hab schon wie ein Irrer geforscht), der hilft ein wenig, wenn man gerade anfängt, kann aber auch keine Lösung vermitteln.
Aber danke fürs Verlinken! 👍

NotMyFault: Thank you very much for trying and suggestion that approach. Looks a bit like my previous tries, unfortunately with the same issues - a distinct sharp-edged inner circle and circular red bleeding into the background. Doesn't look too dramatic, but becomes hair-raising if you try to process the picture further, e. g. try to recolor the sky to that deep dark blue we love with IR pictures, e. g. using HSL. I saw James' video about lens flare removal, will also try to use some of the techniques shown - but I fear it won't work. But let's see.

Old Bruce: I already tried that method by taking photos of a blank white sheet of paper in the sunlight to isolate the color of the hotspot. Didn't work too well, at least I wasn't able to achieve anything usable; but that could just be me. I also liked the idea of creating a kind of "removal template" that way. But the hotspot seems to look different according to lighting conditions. And of course different apertures and focal lenght also contribute to the drama.

So the challenge is still open.

[NotMyFault]

1 hour ago, Jörn Reppenhagen said:

a distinct sharp-edged inner circle

hmm das sollte nicht passieren wenn man gaussian blur auf den Maskierungs-Kreis anwendet. Harte Kanten ggf. einfach per Inpainting korrigieren, das sorgt normalerweise für perfekte Übergänge.

Das Ausbluten habe ich auch bemerkt, wird aufwändiger dies auch weg zu bekommen. Entweder einen weiteren Donut darüber legen, oder wieder brutal per Inpainting oder Kloning (nach Frequency Separation) behandeln.

[Jörn Reppenhagen]

NotMyFault: Ach, du bist ja auch Germane! - Ich antworte aber in English, damit die anderen Leute das auch mitverfolgen können.

I attached a super quick edit of your screenshot, of course it's low-res, but should illuminate the issue. It's just an HSL edit.
See the quite distinct "circle of nothing" between inner circle and bleeding rim, plus the results of that bleeding into the background?
You'll also notice there's more bleeding to the inside - invisible prior to the HSL adjustment.
That's also exactly my own results.

But I am quite sure there must be a real simple solution, I am just too blockheaded to find it.
Just have a look at the second image, a black and white conversion. That spot is still there, creating a quite bright misty spot in the center.

Now the third picture, same black and white conversion, but with the blue component set to 0. Spot almost completely gone! Also works after a channel swap, this time with the red component set to 0. There's still a faint discoloration, but I guess that's because the color of the blue (or red) component is not 100 % on the (hot)spot. Slight loss of contrast could be compensated by raising the cyan (yellow after channel swap) component accordingly.
Thus it seems to be possible to eliminate that spot completely just by reducing a single color component of the right color composition.
I'll experiment with a layer overlay of a single color (without any fading) and diverse blend mode settings for substracting that color from the background layer. Hopefully hitting the color on the spot does the trick. (Till now, it just didn't.)

Maybe this finding might help working out a solution. - Or maybe it's just elaborated BS.

[Jörn Reppenhagen]

An update ...

I managed to solve the problem halfway - it's still faulty, but bearable. Two sample pictures attached.

It's a sequence of white balance, levels, frequency separation, multiple HSLs on the low frequency layer and painting on the same layer, using a reversible mask for sky and foreground objects, then applying contrast and sharpness adjustments on the layers.

I am quite sure there is a quick, direct and flawless approach - but I just don't see it.