Author Topic: False Color for Infrared conveted Cameras - Implemented  (Read 749 times)

Andreasb242

  • New to the forum
  • *
  • Posts: 10
False Color for Infrared conveted Cameras - Implemented
« on: July 22, 2018, 08:08:51 PM »
Hi

I have an infrared (full spectrum) converted 550D, and started to take false color pictures (before I did only monochrome, which is much easier).
It's really hard to imagine how an image would look after editing, as one of the main points is to exchange the blue and the red channel.
I use a 680nm Filter, where blue and green still gets light, for false color.

Therefore I decided I need a false color filter, which converts the picture for me.
I already implemented it, I'll create the pull request later and will link it here.
It's working, not perfect but I think it's OK, but has still room for improvements ;-)

Here a Screenshot:





I'm open for discussion - but probably there are not that many infrared converted cameras...
Hopfully my pull request gets merged anyway.


Andreas

garry23

  • Hero Member
  • *****
  • Posts: 1605
Re: False Color for Infrared conveted Cameras - Implemented
« Reply #1 on: July 22, 2018, 08:22:41 PM »
@Andreasb242

As an IR shooter, 50D and EOSM converted for 720nm, I personally set the LV to B&W as I’m looking for tonal variation mainly. But I understand you are different, ie full spectrum.

I personally use ETTR for exposure setting and my Focus Bar script, which does IR corrected diffraction, for focusing, especially infinity focusing and focus stacking.

And now, of course, I can channel swap in LR :-)

http://photography.grayheron.net/2018/05/infrared-post-processing-in-lightroom.html

http://photography.grayheron.net/2018/05/tent-rocks-in-infrared.html

http://photography.grayheron.net/2017/12/there-are-two-kinds-of-people-in-world.html








Andreasb242

  • New to the forum
  • *
  • Posts: 10
Re: False Color for Infrared conveted Cameras - Implemented
« Reply #2 on: July 22, 2018, 08:44:00 PM »
garry23, your images looking very good. Thank you for the Link!
I have to read your blog. I didn't that much infrared yet. Each picture needs to be edited, and mostly they don't look as good as your example ;-)

But with this filter you have at least a preview, if you have enough color information (I have also other filter, e.g with 950nm the whole blue is filtered out.)

The image is not processed with this setting, only the preview after an image is shot will be displayed in false color.
(One of the rules of magic lanter - it will help photographing, but not do any post processing)

Pull Request link:
https://bitbucket.org/hudson/magic-lantern/pull-requests/938/added-support-for-infrared-false-color/diff


Andreas

Andreasb242

  • New to the forum
  • *
  • Posts: 10
Re: False Color for Infrared conveted Cameras - Implemented
« Reply #3 on: July 23, 2018, 03:36:51 PM »
Tested with a 950nm Filter, where nearly no color information is available.

I tested the Filter. Yes, it's not perfect, but its better than the red image (in my opinion)

Orignal image:




With filter:





Andreas

a1ex

  • Administrator
  • Hero Member
  • *****
  • Posts: 11908
  • 5D Mark Free
Re: False Color for Infrared conveted Cameras - Implemented
« Reply #4 on: July 23, 2018, 03:42:38 PM »
Mind providing one or more CR2 files for including in the test suite?

QEMU can currently emulate the image capture process, to some extent, on a few models; ML overlays work as long as the preview image buffer is filled with valid data. This is currently done manually (i.e. by converting the JPEG preview to 720x480 YUV422 and loading it with GDB), but I'd like to emulate it as well.

Andreasb242

  • New to the forum
  • *
  • Posts: 10
Re: False Color for Infrared conveted Cameras - Implemented
« Reply #5 on: July 25, 2018, 03:06:10 PM »
Hi Alex

I don't know your testsuit, but would I think it would make a general Testcase for such things, e.g. a Testfolder structure like this:

Code: [Select]
TestIrFilter/
MAGIC.CFG
NameOfTheTest.ppm
NameOfTheTest.expected.ppm
NameOfThe2ndTest.ppm
NameOfThe2ndTest.expected.ppm
test.cfg
TestGreenFilter/
...
TestSpotmeter/
...
TestHistogramm/
...

While the contents of MAGIC.CFG is a working MAGIC.CFG, which needs to be loaded before the Testset starts.
Like:
Code: [Select]
# Magic Lantern Nightly.2018Jul23.650D104 (5f3946376d25+ (infrared) tip)
# Built on 2018-07-23 13:40:05 UTC by [email protected]
# Configuration saved on 2018/07/23 15:43:35
beta.warn = 23
menu.first = -3
falsecolor.palette = 6
falsecolor.draw = 1
disp.mode.x = 231

test.cfg is something like:
Code: [Select]
# What the test does
takePicture
sleep 5
takeScreenshot
compareScreenshot

The test would then do the following:
Iterate over all folders, load MAGIC.CFG as Magic Lanter Config, test.cfg as Test settings, and execute the Test for all Images within the Testfolder, and compare the result.

With something like this you can do much more than only testing filters.

I didn't look yet at your testing system, I saw only the views, which you provide on the Download page.

I can do something like this, this isn't that much work. But then I would start a new Thread for this, and also a new Pull request.

I read your thread about Testing, there was not much technical information.
If you wish something like this, I need the help from you, I need an API or something like this to call Qemu. The rest should be relative easy.

Andreas

a1ex

  • Administrator
  • Hero Member
  • *****
  • Posts: 11908
  • 5D Mark Free
Re: False Color for Infrared conveted Cameras - Implemented
« Reply #6 on: July 25, 2018, 03:28:00 PM »
Yeah, I need to document that API a bit. For now, I just wanted some infrared CR2 files to use as input images.

Andreasb242

  • New to the forum
  • *
  • Posts: 10
Re: False Color for Infrared conveted Cameras - Implemented
« Reply #7 on: July 25, 2018, 09:09:03 PM »
Hi A1ex

I created a Repo with three different .cr2 Infrared images:
https://bitbucket.org/andreasb242/ir-testdata/src/master/

As you can see, they are really different, because I used different filters.
Therefore the output result of my filter is also different.


Andreas