Newbie pains, need help


#1

Good day.

I had the fortune of beingunder dark skies during an auroral outburst.
I was ready with a tripod and a camera with a 16 to 35 mm wide angle lens. I tried a few shots with the lens set to infinity. Digital camera allows instant results but low resolution on playback screen.
The results were quite horrid. Seems one can not use the infinity mark on the lens to reliably focus the lens (at various zoom settings) , so I had to guess where to manual focus.
Using auto focus on the stars would not work , too dim and tiny at wide angle, and I did not have the idea to put on my headlights and focus on a fairly distant road sign (at 16mm wide angle!!).

Anyway I got a few images that are a little blurry, with little donut stars, and I have been trying to find software to “correct” the image. I came across ImageJ, installed it, Installed the deconvolutionlab plugin, troubleshot the java issue I had, got it working, loaded 2 images,
: one TIFF 24 by 27 section of image to use as a Kernel. It was a white colored star close to the center of the image where lens distortion was less.
And the scaled down TIFF 5000 x 3337 pixel aurora image.

I let it run overnight using Richardson-Lucy and 10 or 15 iterations.
It ran all night, and was still running this morning. The progress bar had hardly moved ( less than 1/4 the way to total).
I gave up.

This morning I cut out a 700 by 500 area from the 5000by 3337 image (TIFF) and tried processing that with another method… Fast Itterative shrinkage. I guessed that I wanted to “shrink the donuts” down to stars so I would give that a try. Using the same Kernel.
It has been running for over an hour now and again, the progress bar is maybe 1/10th of the way.
The deconvolutionlab window has the option to stop (I used it a few times last night before letting it run overnight, so I know SOMETHING is running), and memory usage keeps changing as seen in the lower mid-right of the window. But nothing is happening. The log window is visible and the last thing it said was:

Starting algorithm “Fast iterative shrinkage-thresholding”. 000.00s
260380 KB 000.01 s Loading input image: “_DSC8738test-S.tif” Done 001.77 s
258915 KB 001.77 s Loading PSF: “KERNEL.tif” Done 001.77 s
Processing started… 001.77s

I am not sure I am doing this right
I am hoping someone can suggest the best way for me to correct the image. Maybe I am using the wrong plugins, maybe I am not using the settings properly…
I have no experience with this sort of processing at all.

Alp
here are links…

https://www.flickr.com/photos/82664576@N03/36725899114/in/dateposted-public/

27 by 24 white star, for kernel

https://www.flickr.com/photos/82664576@N03/37435487551/in/dateposted-public/
scaled down from over 7000 pixels wide to 5000 wide by 3337


#2

One of the links does not seem to be working…

https://www.flickr.com/photos/82664576@N03/36725899114/in/dateposted-public/

.I inserted it here again, the image does not show up.

here are the 2 links again, in an effort to get the link text to show up, andnot the images I did some trickery that seems to work.
just remove the quote and then add https//www. to the front.

"flickr.com/photos/82664576@N03/36725899114/in/dateposted-public/

27 by 24 white star, for kernel

"flickr.com/photos/82664576@N03/37435487551/in/dateposted-public/

scaled down from over 7000 pixels wide to 5000 wide by 3337

Sorry for this half @ssed method of sharing links.


#3

Good day Andy,

inverse filtering (if taken seriously) is a highly complicated kind of signal processing.

Three of the main problems before the actual filtering are:

  1. Gamma correction

  2. Noise estimation

  3. Kernel determination

Here is an article that deals with 1. and 2.:
http://www.gluender.de/Writings/WritingsTexts/HardText.html#1980

Regards

Herbie