Light absorption test - Is it possible using Gimp?

[marco-gimp]

(11-27-2019, 11:18 PM)marco-gimp Wrote: Ha!
Well that didn't go very well.

The Adobe DNG converter was 440 MB download ... and then it wouldn't run due to a missing msvcp140.dll
... which I discovered is a very common problem.

No uninstall info, so had to simply delete it.

Found picosmos about 40Mb, but after installing, it wouldn't recognise dng files.
Perhaps it needs a plug in, however, I recognise these situations
... best to stop now, and try again later with a clear head

Finally found and installed a tiny DNG viewer called Raw Image Viewer.
I need to look at Gimp plugins but for now this was interesting.

I was surprised to find that the DNG image appears to have no pixel blocks - you can just zoom in for ever.

I also noticed lots of noise, particularly in the dark areas.
The 100% JPG produced better definition in these areas, and the colours were life-like.
Whereas the DNG image was a rusty yellow.

I then locked the camera settings, and took a picture with the lens covered - to gain a totally black image.

Here are the results in PNG :

Left JPG 100% ... Right DNG
       

I don't understand how the camera has added coloured noise to an image that is entirely dark.

I have read that for night sky images, when stacked, the raw images are better, as the software gradually eliminates the noise.
This, I thought that I understood, because some light enters the camera lens.
However, with the entirely black image being coloured noise; I realise that I simply don't understand the fundamentals.

As it stands, JPG 100% seems the way to go, though I note that JPG uses gamma correction.

The DNG image contains a colour profile sRGB, though logically it should be at zero colour (the uninitiated might think).

A quick read of some camera threads provided opinions that the problem may be due to :

How the light sensor chip works at a quantum level (apparently too complex to explain)
or 
A RAW image does not store colour values for every pixel - instead each value is either a red, green or blue value. However, you need each pixel to have all three colours - red, green and blue - for the final image. Therefore, a demosaicing algorithm has to guess the other two colour parts for each pixels, and it does this based on knowledge of surrounding pixels.

Either way ... it throws a big question mark over whether the camera can be trusted to accurately reproduce the amount of light entering the lens.
Yet the JPG software seems to be able to calculate from an image containing entirely coloured noise, the actual amount of light that has entered the lens.

Loaded in Gimp

I loaded a covered lens JPG 100 image into Gimp, and zoomed to 25600%
No pixel blocks can be seen - every pixel is black.

Using Brightness, increasing from zero, does lighten the black.

This confirms that the JPG, in this instance, has got it perfectly correct.

---

Just for interest, I zoomed in the Black DNG, to the basic image building blocks.

Here is what it looks like: