My telescope equipment (iOptron CEM70 mount) has not been working since I returned from my January vacation. Fixing it in the winter has been difficult, so I have not been able to observe for more than a month. On Sunday (March 6, 2022) I brought everything into the cottage to prepare it for shipping back to the manufacturer and suddenly it started working again... It doesn't like the winter I guess. It is rated to operate at temperatures above -10C and I was pushing it to operate at slightly colder temperatures. -10 is a heat wave!
While I was waiting for that -10C heat wave, one thing I did do was purchase some inexpensive software that does deconvolution and I then did some simple tests with this software.
This software (Astra Image) has a huge list of different types of deconvolution. I tried each of these to see which worked better. I had to manipulate several parameters, so it was quite subjective in the end. I show here what I consider the best one from that test/comparison.
Deconvolution is an image processing technique that is capable of sharpening up an image.
The algorithms have to measure or assume a certain type of blurring has occurred. The "blurring" can be considered a convolution of the blurring function with the true image.
If we know that the blurring was, we can estimate what the convolution was, and we can then deconvolve the image.
Consider a single star. Ideally it would be perfect point source (unresolved). In practice it has a multi-pixel width and an intensity pattern. That pattern can be used to estimate what blurring pattern was convolved with the true image.
It is then possible to deconvolve the image to try to restore the original image. Of course it isn't magic and can't create information that doesn't exist in the original image.
The top image is of Messier 2, a globular cluster of stars, after deconvolution. This is probably the best type of object for deconvolution since the stars can be brought back to point sources from blurs.
The lower image is the original image, which I had previously considered to be quite good. The deconvolution routine has done great job here of sharpening up the image. The "winner" among the algorithms in my simple test was the "Van Cittert" method.
Note how all the stars became sharper. Some very faint ones suddenly jump off the screen. The light energy was distributed over several pixels and is now much more concentrated, making them stand out from the dark background much better.
Often software improves the astro-images in one way but make a real mess of them in another. I will be using the deconvolution process carefully, as one more tool in the toolbox.
Comments