Deblooming, what am I doing wrong?

[andyc67]

I tried a few times now with no success.
I understand that to remove blooms I have to use HDR techniques. Here is an example of an HDR composition of a stack of 5min subs and a stack of 30s subs. The 30 s subs are not bloomed. This is the best I get. What is the correct way to get rid of blooms?

[msmythers]

If you use a stack of images where the camera has been rotated 90 degrees I would think that the bloom could be eliminated during Image Intergration much like plane or meteor trails are eliminated.


Mike

[sreilly]

Personally I gave up on de-blooming in PI. It's fine if you have a square chip and a guide star when you rotate the camera but way too much of a problem otherwise and you need to plan on doing so when planning your image. As much as I hate to say it, when I use the ST10XME I just calibrate in MaxIm, my acquisition software, and run Ron Wodaski's debloomer plugin. It's just plain simpler and efficient if you already use MaxIm. I love PI but I found it not so great on deblooming for myself.

Steve

[msmythers]

Steve

Since I don't have a CCD my question is does rotating the camera not with standing the other issues eliminate the bloom? If so then you have to decide about how to deal with the other issues which may turn out to be not to rotate and use other software. We rotate microwave antennas during testing in the same way to eliminate reflections and noise captured in certain quadrants of the antenna. It's not always easy especially with non-symmetrical antennas but it does the job.


Mike

[sreilly]

To be honest I never tried due to the fact I didn't have a square hip and would lose too much of the image by doing so. I guess due to the complexity of the blooms I just went with what I already had that I knew would work. The plug-in actually worked better than the built-in feature of that software and if I remember correctly there was an option to rotate the star in the plug-in.

[sreilly]

I'll probably get zapped for that but maybe someone could write a script that does the same thing as Ron's deBloomer. His deBloomer is a free plug-in these days.

[andyc67]

I don't have Maxim but I do have CCD Stack which also removes blooms. That is my fall back plan but due to FIT compatibility issues I have to resave my subs in TIFF format. Not a big deal as they are still 16 bit and then from there I pick up with Pixinsight.

[Alejandro Tombolini]

What is the correct way to get rid of blooms?

Hi, I would lower Binarizing Threshold, however a careful inspection of the generated mask will tell you how to move it.
This process example has information about HDRComposition tool that may help.

Saludos, Alejandro

[andyc67]

I will try that. I had the binarizing threshold already at 0.25. and also reduced the mask growth. My suspicion is that I need more intermediate exposures.

[msmythers]

One other thing you can use if you want to paint the blooms out is the TGVInpaint tool and the NoiseGenerator tool. This is very manual in it's use but effective.


Mike

[Alejandro Tombolini]

My suspicion is that I need more intermediate exposures.

Yes that is very likely

[vicent_peris]

Hi,

What do you need to do is simply to denoise the shorter exposure. That's why the composition masks are so useful. Once you have your HDR image, load the composition mask on it and apply a strong denoising (I recommend TGVDenoise) to the blooming area. The goal is to have the same noise level in this area than in the rest of the image.


Best regards,
Vicent.

[andyc67]

Thanks for the great advice!
I will try the masked denoise and if it all fails the inpaint technique.

[vicent_peris]

One tip more. Surely you'll need to brighten the composition mask before applying the denoising, otherwise some noise will still remain over the areas corresponding to the edges of the mask.

V.

[jkmorse]

This seems like such an obvious place for a PixInsight script.  If it can be done in CCDStack and MaximDL, I would be shocked if we can't do the same process with the myriad of tools available in PI.  That would eliminate the noise issue, though I recognize that Vincent's solution works in the interim.