Wrong color output despite color profile

Software RawTherapee
1

Hi,

I am relatively new and have a problem with RawTherapee 5.8 regarding colors.

If I want to edit a DNG photo from my smartphone (Galaxy A52s 5G), I will get a result with not the correct colors and exposure - especially in the corners - even with a dcp color profile.
I don’t know why – with my Panasonic GX80 I never had such a problem.

Here is an example:

RT
RT1188×794 242 KB

RT Settings

However, if I edit it in Lightroom 6.14, I will get the right colors and exposure – so my dcp-file seems to work:

Lightroom
Lightroom994×728 248 KB

Here you can see the in-JPG photo output from the smartphone:

Original-JPG
Original-JPG1038×774 333 KB

Could you please give me a hint how I can get the right colors and exposure in RawTherapee?

To test it for your own – here you can download the DNG photo and dcp file.

Test-File.zip (12.3 MB)

Thank you!

2

The problem is that the image contains GainMaps: vignetting correction that is specific to each color channel. If the GainMaps are not applied, the colors will be incorrect, especially in the corners.

Currently the only open source raw developer that supports the GainMaps is ART. To apply the GainMaps in ART, go to the Raw section, enable Flat-Field module, and select ‘Embedded in metadata’.

I am working on adding support for the GainMaps to darktable, but I’m not sure when that will be ready for use.

1 Like
3

The only way to use RawTherapee for this image is to apply the GainMaps to the image before editing it. I wrote a script to do this using the Adobe DNG SDK, instructions for building and running it on Linux are posted in this thread. It takes the dng from the phone, applies the GainMap, and writes out a new dng that RawTherapee can use.

4

Hi,
playing with your file i realized that it cannot be found in Lensfun, thus there is no correction for vignette available. changing those parameters i got this:

Test_of_IMG_20211016_163723.dng
Test_of_IMG_20211016_163723.dng1920×1439 1.21 MB

Test_of_IMG_20211016_163723.dng.jpg.out.pp3 (14,8 KB)
(dev-version RT 5.8-3049)

But anyway, sth weird is going on with this file: applying the auto tone-curve, its being reset to linear…

1 Like
5

The difference between the vignetting correction and the GainMaps is that the vignetting correction applies to all pixels equally, while there are separate GainMaps for R G and B pixels. For this particular phone camera the differences between the R G and B are relatively small so the vignetting correction works well, but for other phones like my Pixel 4a the difference is much more obvious.

1 Like
6

Hi,

Thanks a lot to you all!

ART does work for me.

@ paolod: Thank you for the script. I also could found an app called raw2dng (https://play.google.com/store/apps/details?id=com.fimagena.raw2dng) which seems to be the same but for Android.
If I convert the dng to dng I get the same result in RawTherapee but if I convert it to TIFF or JPG I get the right colors.
Do I have to adjust something in RawTherapee or the app?

7

The original raw2dng app from https://github.com/Fimagena/raw2dng and the Android version does not do what we need here, only my dngpreprocess app applies the GainMap and writes out a dng with them applied.

To make the dngpreprocess app, I made a copy of the raw2dng project and added my script to it. I was just using that project as a sort of template since it was already set up as a command line utility that used the Adobe DNG SDK.

8

Seems to be a Samsung thing…acts like there is no embedded Jpg…

1 Like
9

Are you checking in RT what parts of the DCP are applied…I have seen often only the base table is applied…you can also apply the tone curve and look table…the look table will alter the colors consiserably…so just check what you have set in color management…

10

Thanks a lot to all for the hints and information :slight_smile:

So I will use ART for this and wait until Paolo has integrated the GainMaps in darktable.

11

I am working on gainmap support for RT…

7 Likes
12

Is there syntax for dngpreprocess to allow it to process a whole directory or do you have to write a looping script??

13

dngpreprocess only handles a single file, you need a script to process a whole directory. I have been using this script, I run it from within the directory containing the images and it converts only the Pixel images leaving other raws alone. It uses GNU Parallel to run more than one conversion at once, and moves the original dngs into a subdirectory. The only problem with this script is that when it adds _pp to the name, darktable no longer groups the PXL_xxx_pp.dng file with PXL_xxx.jpg, as it would if it was named PXL_xxx.dng.

#!/bin/bash

mkdir -p original

process_dng() {
    dngpreprocess $1
    mv $1 original
}
export -f process_dng
find . -maxdepth 1 -name 'PXL*.dng' ! -name 'PXL*_pp.dng' | parallel process_dng
1 Like
14

Thanks for sharing…

Got parallel installed but the script can’t find dngpreprocess. The script, dng and dngpreprocess are all in the same directory??

Error is environment: dngpreprocess: command not found. I must have missed something along the way…

15

You can either copy the dngpreprocess executable to somewhere in your PATH, or change the script to ./dngpreprocess $1 so that it finds the executable in the current directory.

1 Like
16

Thanks I did get it working…not as elegant as you have provided…thanks so much this is really nice to have…in a way I got around it for a while and surprisingly it worked well… I chose the lens profile for the huawei P10 and it had a pretty decent vignette correction and did not seem to distort the image…The vignette is so strong that without it the images were a pain to deal with…gain maps is really nice…