Alright, once again I’ve come across this issue of having artefacts with my drone DNG files in Darktable. The first time it was Inspire 1 v2 and HDRHerge creating some OpcodeList1 FixBadPixels metadata (more on that here)
Now there are plain DNG photos just as the drone shot them. So no processing of any kind has been done in any external software.
Those artefacts are all over the image. Expand the photos so you can see what I’m talking about.
This is what the issue looks like in Darktable and RawTherapee:
These are not hot pixels. Hot pixels don’t look like this (I have a camera that has hot pixels so I know these pluses aren’t it). The drone is brand new. Unpacked today. These are literally the first few photos taken with it.
Lightroom doesn’t show these artefacts.
In darktable, change demoasoc to amaze, turn color smoothing to 5x. Enable the hot pixel module, check detect by 3 neighbors, threshold to .0750, strength to 0.3
I indeed may had that confused. Then again, it’s a brand new drone. This doesn’t happen on my years old Inspire 1 files nor on Mavic Pro 2. I can’t see a single hot pixel there. So I am still a bit scheptic at the moment.
Also it’s ISO 100, and those hot pixels don’t show up in the drone JPEG. I just can’t find the metadata if they have this mapped.
I know how to do hot pixels removal. That’s not the issue. I just think these shouldn’t be visible in the first place. Like what was the case with the github issue I’ve linked in the original post. I gotta test with Lightroom to see if it’ll show up there (or maybe they remove hot pixels in the background idk).
After disabling white balance and demosaic this clearly looks like a hotpixel to me. In longtime exposures the sensor gets hot and single pixels drop out until the heat is gone.
Are you sure Windows Photo isn’t showing the JPEG preview, or if it is doing a full RAW conversion, isn’t somehow using a library/driver/codec that was installed by Lightroom to do it?
At least your embedded DNG color profile is correct, unlike the Phantom 2 Vision+ which had a completely broken embedded DNG color profile.
That’s exactly what I asked my friend. He told me it’s not, that he can zoom all the way etc.
I don’t know that but I would expect then that the image from the Lightroom and Windows photo screenshots would at least have the same colors which they clearly don’t.
Windows Photo looks more like the drone’s JPEG file. So I’m very suspicious if he actually saw the real rendered DNG or just the preview.
Hm, I have found 1 blue hot pixel on Mavic 2 Pro. I guess Mavic Mini 2 just might have a shitty sensor. I’ll test Mavic 2 Pro in low light with longer exposure tho.
The raw file contains a preview image. However, it is smaller (960x540) than the full image (4000x2250), so we can easily tell them apart.
I agree, these are hot pixels. By this, I mean isolated sensor elements that have clearly and falsely saturated, and this “falseness” is spread to adjacent pixels by the demosaicing process.
I sometimes get them with with longish night exposures on my Nikon D800 DSLR. When they occur, they are at random places in the image, so are not fixable by using a constant mask.
I prefer to fix them before demosaicing. See my Hot pixels page.
