Reddish-purplish tint on output photos

Hi, I am new to editing and wanted to try out merging bracketed photos. I am using RawTherapee and decided to try HDRMerge-app to do that. After I feed HM the .cr2 files, it does the masking and I save it as HDR in .dng format. The photos I get always end up having a reddish-purplish tint and looking something like the photo I added to the post, both in windows preview and RawTherapee.
Is this okay and I have to somehow proccess the .dng in RT to get it to look correct or are there problems with HDRMerge?
I tried using stock settings (Preview size Full, mask blur radius 3, no mask image saved) with16, 24 and 32 bit. I also tried converting the .cr2 files to .dng with DNG converter before feeding them to HDRMerge but it doesn’t make a difference. I am using RT 5.10 and HDRMerge 0.5.0
HDRMerge output example: Imgur: The magic of the Internet

Your HDRMerge version happens to be 9 years old. Is there anyway possible you can try with the master branch?

where would a more modern version of HDRMerge be found? I am using a Windows computer and the latest version I have found is version V0.5.0 from

Thanks

Apparently, 0.5.0 is indeed the latest stable version; however, this was released way back in January 2015:

Isn’t this related to the known issue of RT using black/white levels and other metadata from camconst.json instead from the DNG itself?

Yeah, though it’s the latest one that has an installer. I’ll try installing v0.6 but it can take a while as I am totally not familiar with GitHub in that regard

Probably it is. But I have to idea, which numbers to put in the raw_crop, so I’ll probably roll with Luminance HDR for now and wait 'til I get my hands on Lightroom

Run exiftool on the DNG.

So I tried to run the tool and I am not sure which data I need exactly so I tried it with this:
photo_2024-03-07_17-32-09
And set the camconst.json to the following:


(EOS750d is the camera I use).
But I think that I didn’t do it correctly as the tint is still there :frowning:
P.S. Also, the image size in exiftools is the same for any of the input images and the HDRMerge one. The only other width/height I could find in the exif report was 6000x4000. Although if I check properties of the input file, its resolution is set at 6024x4022

A long time ago, I read about magenta color casts appearing in images:

http://kronometric.org/phot/iq/where%20the%20magenta%20comes%20from.htm

Probably not all relevant to your problem but examining the histograms (esp raw if poss) at the various stages might point to something significant …

Are you able to git clone https://github.com/jcelaya/hdrmerge.git \ \ \ ?
or are you also on windows? (If so, this is buildable and runnable in wsl2).

What makes you think size/crop is the problem here?

yeah, I’m on windows

As far as I could understand from the disscussion, a link to which you sent me earlier, this happens because RT percepts color arrays incorrectly because of differences in image sizes and as a solution to a similar problem another guy had, someone suggested him to tweak the .json file and add a custom raw_crop there

This usually doesn’t happen with HDRMerge… Might be an HDRmerge bug.

Can you upload an example output, or even better, maybe upload your input sequence to filebin.net ? (filebin is kinda popular with the HDRMerge/RT developers).

I’ve been poking at fixing various HDRMerge issues over the past few months, maybe you’ve got a good test case here.

Unfortunately, HDRMerge’s latest release is really old, plus there are now some unmerged patches that greatly help in certain scenarios. (Partly unmerged due to maintainers being absent, partly unmerged due to still being works in progress)

Hello! I stumbled upon this thread when searching how to fix the same problem with HDRMerge and raws from my Lumix camera (in both RawTherapee and Darktable).

I decided to give v0.6 a try as suggested. I don’t really understand how versions of HDRMerge work, because program compiled from GitHub is still titled as v0.5.

Anyways, this compiled version works well, without the purple tint!

Too bad I use Windows desktop and the version I compiled is on my 15 year old laptop with Linux Mint. But I guess it’s better than nothing and I can finally try playing around with bracketing. :slight_smile:

Just a little note, I compiled it using steps in the Install.md file. There’s a step to install a bunch of dependencies where I had to replace “qt5-default” with “qtbase5-dev qt5-qmake” to make it work.

Yes those will vary even between the three Ubuntu versions.