Deconvolution Issues on Images with Higher Contrast / High Dynamic Range

[brent1123]

Hello,

How does everyone deal with Deconvolution on images with high dynamic range? Specifically, I am having issues with ringing of small stars surrounding larger bright stars as well as bright areas like the core of M31, where the background quickly approaches the value of small stars in the foreground. The image is Linear and has had no other processes applied to it aside from Background Extraction.

I am running a Decon process using (per Inside PixInsight by Warren Keller):

 - An external PSF rendered using the PSF Image Script

 - A non-linear image clone as an image mask for application

 - A star mask for deringing support. This star mask was made from 3 different star masks targeting various star sizes combined using Pixel Math with a Screen blend: 1-(1-A)*(1-B)*(1-C...

Decon looks fine when running on any number of previews on most of the image, but these bright areas seem to require different settings altogether. I have tried creating a Range Mask of brighter areas and subtracting it from the Luminance Mask, but this sidesteps the issue since I do want to run Decon on the brighter core of M31 (as I see a lot of detail improvement in the dust lanes at the cost of ringed stars). My second and latest attempt (used for the attached screenshot) was to make another star mask targeting small yet bright stars, convolute and brighten it to increase the star sizes, then combine it with the other star masks, but it does not seem to provide sufficient protection.

Any suggestions would be welcome, I'm stuck on multiple images due to this apparent impasse





https://imgur.com/a/SNsZSw4

[ngc1535]

Hi,

I will be speaking on this very topic at the Advanced Imaging Conference on Friday as part of my presentation.
The two parameters you need to adjust are the global dark and the perhaps you may need to adjust the synthetic
PSF if the tails are too large. You did not mention your global dark setting.
The "trick" is to find the right parameter that minimizes the ringing where stars are embedded in nebulosity *AND* not have
the global dark brighten stars that are against the dark sky (which will be your next complaint!).

The complex local support image you indicate may not be necessary? Usually you just want to not sharpen only the very brightest stars.

-adam

[stevek]

I've tried to come up with a "theorem" for the global light and dark settings, thinking there could be some sort of correlation with a given set of LP, exposure and other parameters and build an excel macro to spit out the light and dark settings.  I gave up though.  It seems to be totally unique to the image in question and the set of prevailing circumstances regarding the capture of that image's data.

Steve