Hi,
I would like to have software that takes an existing exposure and fills in the information that is present but unclear, but does not make the image brighter, just clearer.
A.I. Denoising is based on filling existing information with statistical experience.
However, this is partly an image-destroying process and also changes existing image content and depends on the training material.
My goal would be for the software to learn how to continue a virtual exposure as in ray tracing software.
Since I am currently working on astro imaging, I have encountered previously unknown problems that would also be useful for other areas of photography if they were solved.
Basically it is denoising, but as described above, A.I. denoising works with statistics based on images that may not match your own image processing style or image type, which is why there is a lot of criticism in this area.
So instead of learning how the target condition should look like, the software should learn how to get there without ever reaching a target condition.
You basically stop the exposure when you think it’s OK.
