Samsung SM-G970F dng files: wrong colors raw

The metadata tags your camera as “SM-G970F”. I Googled it, and it appears to be a Samsung S10 cellphone, no? That’s somewhat new territory for us old folk… :smile:

Nominally, what it takes to insert minimal camera support is to shoot an image of a color target and use it to produces a camera profile. My granddaughter has a S10, next time she visits in the daylight, I can make such an image. Assuming the US version has the same sensor…

As cellphones incorporate the software to shoot raw, this’ll be new territory for support…

it good news but it looks like snapdragon and exynos versions have different sensors, and US have snap, i think.

also its no new territory is already 5 years here :slight_smile:

@dssdd I will answer later in the evening. Supper time now :slight_smile:

After reading this I loaded both examples into Darktable for an analysis of the material.

Im my opinion it is not an issue of the camera support, it is an issue of the massively overexposed image.

The squares in the upper image part represent overexposed areas within the RAW data. Which means the light was so bright that the sensor data is corrupt.
Programs like Rawtherapee or others can only try to ‘imagine’ what data could have been there.

Resulting problems often are flawed white balances:

  • To set the white balance I first tool a sample from the ‘grey’ sky in the top-right. The result is a greenish image, which you can see in the left half of the screenshot
  • On the right side you see the white balance taken from a sample from the stonewall in the center (where the small rectangle is). Here the white balance looks good.

What works different in various programmes are the algorithms for highlight reconstructions for overexposed areas.
Sometimes they look good, sometimes they create weird colors.

My recommendation: Take a properly exposed photo and load it into Rawtherapee and Photoshop. I don’t think that there will be bigger problems.

Edit: For highlight reconstruction programs look for colors around the overexposed areas and try to move them into them. Could be that some details like these fine lines are lost. In these areas it is not easy to detect valid data.

4 Likes

20190411_125459.jpg.out.pp3 (11.1 KB)

1 Like

I’m guessing that this, plus the fact that the Galaxy S10 has three back-facing cameras (plus 2 front facing, although it’s hard to imagine taking the selfie cameras seriously), will make things difficult unless each camera/sensor combination can be identified.

Cameras:

  • 12 Megapixel f/1.5-2.4 variable aperture camera with 960 FPS slow-mo, 4k, etc
  • 12 MP f/2.4 telephoto zoom lens
  • 16 MP super-wide angle camera.

is true, but not affect our case, bc only one camera can shoot in raw in all devices…
and different models have different code name SM-G9*** but only two mods.

@heckflosse
its ok, im wait, but please just dont forget about me compleatly :slight_smile:

Thanks, good info…I may be buying a S10+ soon.

@heckflosse

it looks like 5 supper times gone :cry:
can you light something?

Not yet. I would have done.

Too many irons in the fire, almost forgot to follow up: turns out she doesn’t have a S10.

looks like it was compleatly forgotten…
@heckflosse
please say atleast how it going or not? when comes or not comes?

you see, i looked not wrong way…
no support, no answer about plans, no timemap, soon 2 months gone and nothing.

Please don’t make your prejudice about open source projects into a self-fulfilling prophecy. You need to adjust your expectations for open source projects.

  1. There are usually less than a handful of active developers who need to divide their time over many different issues with the software.
  2. It’s entirely up to the developers to decide if they want to work with timelines and so forth, or just cherry-pick issues along the way. You might think this means a project is badly organised, but that’s the reality whenever something is done entirely by volunteers.
  3. Don’t assume your issue needs priority, simply because it affects you.
  4. Keep a constructive dialogue when the developers aren’t as quick as you would like, don’t get defensive.
  5. If the information necessary for a fix to your issue is actually not yet provided, not a lot can be done. @Morgan_Hardwood and @heckflosse, what is still needed?

The simple reality is that adding camera support is usually a bit of a tedious thing to do…

5 Likes

hey bro, with all respect, but what the hell self-fulfilling prophecy???
from the start i ask only for easy and short answer that clear situation -
“yes, we will take about for X-weeks”
or
“no, we will not take, until there is no possibility and we do not see it in the next six months, later - perhaps.”
like that, it’s so simple and it’s damn banal politeness.
I do NOT command “quickly add urgently, bastards,” NO.
I’m not aware of your situation, right? simple explanation of intentions, and approximate timelines …
why make me a villain?

http://rawpedia.rawtherapee.com/Adding_Support_for_New_Raw_Formats

1 Like

@Morgan_Hardwood
This link was needed from the beginning!
we would not have lost almost 2 months … then you answered me about the color target, I thought that it was necessary to have it …

So I took pictures, because the article does not contain words about white balance, I left the auto.
I hope I did everything right, shoot all in 1 sec, for all available modes of ISO, for each available aperture (there are only 2).
but there is one photo exception.
The fact is that the stock camera in manual mode has a restriction of only ISO800 when the module itself and third-party applications can choose FREE value up to ISO1250! but at the same time 1s, can only be selected in the stock application, and the rest are limited to 1/10.
so please let me know if i do something wrong or need something else, i do again.
@heckflosse

ps i have error upload for Filebin, so i put files here GalaxyS10.7z (268.6 KB)

@heckflosse
@Morgan_Hardwood
I have provided all the necessary files …
what is still needed?

My schedule is full for at least another week.

@Morgan_Hardwood
ok, thank for say it! no problem
but can you at least just look and say im do correct samples or no? so i can fix them for this weak.
or check this take lot of time too?