deconvolution mask subtraction

[jamesRC]

Hi,

I'm trying to learn to do deconvolution. There's a point where you subtract a star mask from a range mask, using PixelMath. I would expect that the result would have only my target galaxy showing, so that applied as a mask this result would allow action such as sharpening to only the galaxy target. Instead, the result of the mask subtraction has both stars and galaxy showing, it's the same as the range mask alone. I feel something is not happening here. Or else I got the wrong idea :-) Can anybody please clarify?

Thank you,

James

[aworonow]

My guess is that the range mask is binary, hence the values are only zero or one, whereas the star mask has most values in proportion to the brightness of the stars. For instance, a star-mask value might be 0.3. Hence 1-0.2 = 0.7, and there appears to be a star remaining. Try binarizing your star mask.

Also, there is a script called GAME that allows you to draw masks in the shapes of ellipses. That might be useful for your galaxies too. (Scroll to the bottom of https://pixinsight.com/forum/index.php?topic=11847.0)

alex

[jamesRC]

Alex,

Thank you!
I'll get right to checking that out. I guess there's no mention of t hat in  the procedure I'm following, or else I'm "supposed to know". :-/

James

[jamesRC]

Alex,

I just tried binarizing the star mask. It doesn't work at giving me a target-galaxy-only mask, and besides selecting "binarize" in the mask options gives me a VERY noisy range mask. Drat! I'll go check the GAME app.

Thanks anyway. Still this is an important step in the deconvo process and it apparently doesn't work, what the heck!!?

James

[pfile]

it's not strictly necessary to mask the image before deconvolution. if you set up the Wavelet Regularization area properly you can avoid sharpening background noise. for my part i usually set the wavelet scale to 3 and then increase the noise thresholds to (on the order of, you may have to experiment) ~12 for scale 1, ~8 for scale 2 and 5 for scale 3.

having said that it's not unheard of to mask the background and the central cores of the brightest stars in the image. if you upload your masks to dropbox or google drive or whatever and post the links, i can see why the subtraction isn't working right. most likely it's because the values in one mask are too small compared to the others, so the subtraction is essentially a no-op.

rob

[jamesRC]

Rob,

Thank you very much.

I'll play with your suggestions.
BTW I just discovered the Advanced Sharpening Script. From the intermediate images I see during execution of Advanced Sharpening, this may be a pretty good replacement for Deconvolution. Not perfect of course! but pretty good.

Thank you again Rob!

James

[STEVE333]

Hi James - There is a nice tutorial on Deconvolution that you might find helpful. I've followed this tutorial with good results.

https://www.youtube.com/watch?v=97OyeSR76Fs

I usually mask out the larger stars because Deconvolution often causes funny artifacts with them.

Just FYI.

Steve

[jamesRC]

Hey, Steve.

I've been trying to use deconvolution in the way described in your video reference. But the mask subtraction, as I said earlier, is not working for me. Dunno whythough there have been suggestions.

Thank you,
James