My attempt at BGR

[Stefano]

Yesterday Stephan suggested me IrfanView as a software to edit images: https://www.ultravio...post__p__39238. I noticed it has a channel-swapping feature, and I really liked it. Below you see some images I have already posted in the forum in the past, but converted in BGR.

 

Image: https://www.ultravio...-1576877461.png

Post: https://www.ultravio...dpost__p__31485

 

Image: https://www.ultravio...-1581372961.jpg

Post: https://www.ultravio...dpost__p__33080

 

Image: https://www.ultravio...-1586374208.jpg

Post: https://www.ultravio...dpost__p__34983

 

Image: https://www.ultravio...-1585946815.jpg

Post: https://www.ultravio...dpost__p__34909

 

Image: https://www.ultravio...-1591277922.jpg

Post: https://www.ultravio...dpost__p__36235

 

Image: https://www.ultravio...-1602537628.jpg

Post: https://www.ultravio...dpost__p__39198

 

I remember this comment by Andy: https://www.ultravio...dpost__p__36801

My answer: BGR (both in IR and UV) can sort-of mimic true tri-color images, but the colors are not "exact" (reds are not exactly red, etc. You can see this in the first image in this post), but this can be fixed in post-processing. The biggest issue is that, in both IR and UV, you lack green. You have the extremes, red and blue, but not the middle green band.

 

This technique reminds me of IRG done like Andy did here: https://www.ultravio...chrome-rainbow. Yes, it works and it is quite accurate, but it is an approximation, that (in my opinion) can not replace true tri-color images.

[StephanN]

I say, I never noticed the channel-swap feature in Irfanview

[Stefano]

They already have all 6 presets (3! permutations of R, G and B). You find them under the "image" tab, if I remember correctly, and in "swap colors", almost at the bottom.

[StephanN]

They already have all 6 presets (3! permutations of R, G and B). You find them under the "image" tab, if I remember correctly, and in "swap colors", almost at the bottom.

 

Cool, thanks!

[Stefano]

I am thinking... there must be a way to obtain three channels (the missing green one) from a UV image (in a similar way, from an IR one). A camera registers each wavelength as a slightly different hue, more or less like I showed here: https://www.ultravioletphotography.com/content/index.php/topic/4124-micro-nikkor-p-auto-55mm-f35/page__view__findpost__p__39097. We could assign a new color to each hue and obtain three channels, or even 100. I am learning MATLAB, maybe this programming language has potential in this regard.

[OlDoinyo]

The color scheme you see is a function of what the reseau dyes do outside the spectral range in which they were intended to function. As such, the image characteristics will be somewhat capricious. If you take three exposures through narrow-band filters and stack them in different channels, you will get the kind of trichrome image of which you write. MaxMax already markets a set of filters for doing this in near IR. There is no corresponding ready-made filter set for the UV, but several others have attempted to do this with filters purchased separately (see here, for example.)

[Stefano]

I too agree that true tri-color images should be done with three bandpass filters, assigning one RGB channel to each filter. But I also think that we can extract a lot of spectral information from the color the sensor sees. I like a lot this UV spectrum by Andrea: https://www.ultravioletphotography.com/content/index.php/topic/3471-coastal-optics-10545-diffraction-grating/page__view__findpost__p__29706

 

As you can see, even in the RAW image, there are some colors. I could do something like this: assign a number, like from 0 to 255, based on the hue. Blue-violet may be 0 and orange (assuming you don’t have a quartz lens, I surely don’t), may be 255. Then assign each number a color from red to blue (or red to violet, as you like). Finally, changing the colors in the image. This way it should resemble a tri-color image, but without three separate filters. That’s only the basic idea, I don’t know it and how well this can work, but I will try it if I can.

[Andy Perrin]

As you can see, even in the RAW image, there are some colors. I could do something like this: assign a number, like from 0 to 255, based on the hue. Blue-violet may be 0 and orange (assuming you don’t have a quartz lens, I surely don’t), may be 255. Then assign each number a color from red to blue (or red to violet, as you like). Finally, changing the colors in the image. This way it should resemble a tri-color image, but without three separate filters. That’s only the basic idea, I don’t know it and how well this can work, but I will try it if I can.

 

Stefano, that's standard color mapping, and it's very easy (you can even do it in Photoshop, no need to mess with MATLAB!) but it's not anything like true color imaging by bandpass filters. In Photoshop, it's the "Gradient Map..." option in Image->Adjustments submenu. Back before I joined this forum, when I still only had a point-and-shoot that could see not much further than 375nm or so, that is how I colorized my images. You can get some pretty pictures that way but they do not contain any additional spectral information really. Basically you can never get something from nothing. Our UV photos really contain only two channels worth of information, as you can see by converting white balanced UV photos to L*a*b* space -- the "a" channel is nearly blank.

 

I have used the color mapping done in MATLAB on my thermal photos, to make grayscale into colors:

https://www.ultravio...urch-lwir-pano/

[Stefano]

Stefano, that's standard color mapping, and it's very easy (you can even do it in Photoshop, no need to mess with MATLAB!) but it's not anything like true color imaging by bandpass filters. In Photoshop, it's the "Gradient Map..." option in Image->Adjustments submenu. Back before I joined this forum, when I still only had a point-and-shoot that could see not much further than 375nm or so, that is how I colorized my images. You can get some pretty pictures that way but they do not contain any additional spectral information really. Basically you can never get something from nothing. Our UV photos really contain only two channels worth of information, as you can see by converting white balanced UV photos to L*a*b* space -- the "a" channel is nearly blank.

 

I have used the color mapping done in MATLAB on my thermal photos, to make grayscale into colors:

https://www.ultravio...urch-lwir-pano/

Thanks. So basically I can do it, but it wouldn't give me much, right?

[Stefano]

Well, at least I have two channels, as you said. In BGR images, if I isolate the blue one, I can isolate the shorter wavelengths. It would be the same of isolating the red one in normal RGB images, and probably similar to isolating the green one in both, as this channel is left untouched and both yellow and cyan contain ~50% green.

 

This are the red and blue channels of a BGR image above.

 

red

 

blue