Yes, I'm doing lucky imaging.
Where possible, I'm actually already doing what you suggested. I just wanted to sketch the problem here as simple as possible. But the program already incorporates many tricks to speed things up. The software is actually meant for lucky imaging on wider targets: planets, or even the sun or moon. See my website for some example images http://www.astrokraai.nl if you are interested. But it could also work on starfields. The placement of the alignmentpoints (or rectangles in my story here), is an entirely different problem of it's own. Among one thing, it basically makes use of what you suggested. Areas without relevant information will be ignored (but sometimes it is really difficult to tell which are without relevant information. And often the field of view is all relevant information).
The problem I sketched here comes from doing lucky imaging on a larger surface, like the sun in a particular wavelength, where the target does not fit the FOV. So every portion of the image actually is relevant. There are no black rectangles. Everything portion of the image contains (distorted) views of the sun. Hence the rectangles all over the place (sometimes >1000!). If the FOV contains smaller targets, like the planet Mars for example, there will be far fewer rectangles (usually < 10). But the basic problem remains the same: comparing one image to another.
To give you an idea about the size of the project at the moment: it contains well over 12.000 lines of self written pascal source code. I started the project back in 2009, but right now it is being used by many amateur astronomers taking pictures (video sequences actually) of the planets, sun and moon.
Actually comparing one image to the other takes up most of the processing time, and I thought it would be very interesting to see if that could be made faster.
Reply With Quote
Bookmarks