Page 1 of 1

Vanderbei Deconvolution for coma and field curvature.

Posted: 09 Feb 2023, 06:05
by Dunmunro
I've been experimenting with Dr Vanderbei's plugin using AA8LAA SP3:
(image is centred on NGC1275)
(drizzled to simulate files from a 24MP camera)


The plugin seems to only work with monochrome images, so I split the master fits image into it's R-G-B channels, saved these as separate images and then used the plugin to deconvolve each image (simultaneously using 3 instances on AA8LAA on my I7 computer since the plugin seems to be single threaded). After the deconvolution was completed, I recombined the monochrome R-G-B images into a single colour image. I used the suggested settings as determined by the plugin. It took about 30 minutes to process these relatively small images. I replicated the process using a 2x drizzled version of the master image and the 3 instances took about 2.5 hours to complete (running simultaneously)!

The above images aren't all that pretty to look at but it has done a very good job of correcting the star aberrations across the image. I may tweak the settings to produce better results.

Re: Vanderbei Deconvolution for coma and field curvature.

Posted: 09 Feb 2023, 16:15
by fabdev
That's interesting. Yes, that plug-in is single-threaded.
It looks like most of the blurring is caused by image deformations from different wavelengths, not PSF. Maybe after having separated the image into R-G-B channel the new alignment with distortions could attenuate the problem (align R to G, then align B to G, then recombine).
Fabio.

Re: Vanderbei Deconvolution for coma and field curvature.

Posted: 09 Feb 2023, 20:05
by Dunmunro
fabdev wrote:
09 Feb 2023, 16:15
That's interesting. Yes, that plug-in is single-threaded.
It looks like most of the blurring is caused by image deformations from different wavelengths, not PSF. Maybe after having separated the image into R-G-B channel the new alignment with distortions could attenuate the problem (align R to G, then align B to G, then recombine).
Fabio.
Yes, I'll try correcting for deformations, and see how that looks.

Re: Vanderbei Deconvolution for coma and field curvature.

Posted: 10 Feb 2023, 00:19
by Dunmunro
fabdev wrote:
09 Feb 2023, 16:15
That's interesting. Yes, that plug-in is single-threaded.
It looks like most of the blurring is caused by image deformations from different wavelengths, not PSF. Maybe after having separated the image into R-G-B channel the new alignment with distortions could attenuate the problem (align R to G, then align B to G, then recombine).
Fabio.
Yes, that did help: