How to correct this?

[Emiel Kempen]

Hello all,
I am using an ASA 0.73 reducer / corrector on my 254mm F/4 Newtonian telescope.
After colimating the scope and getting everything as square as possible I still got misformed stars in the topleft corner.
Does anyone know if there is a method to correct these stars (semi) automatically with a tool or script in PixInsight?
Any help is much appreciated.
Regards,
Emiel.

[Hans Pleijsier]

Maybe you could try and play with ATrousWaveletTransform and a mask

[Emiel Kempen]

Hello Hans,
Thanks for the reply.
Could you please give some step by step directions how to do that?
I am not so very deep into PixInsight.

Emiel.

[zvrastil]

Hi Emiel,

I'm pretty sure there's no automatic or semiautomatic way in PixInsight to fix such problem (well, there's DynamicCrop  ). Probably, an image processing guru could find some very clever way using wavelets, masks and morphological transformation to remove those tails, but I doubt the result will be plausible. I'm afraid the only way to improve the situation (except cropping your images) is to fix the problem in optics. Have you tried to contact ASA about this? I'm pretty sure it should not behave like this (if there are limitations, the effects should be symmetrical at least).

regards, Zbynek

[Lex]

Emiel,

How did you collimate your Newtonian?
The reason could be that your actual collimation is not in harmony with your reducer and that you will have to optimize your collimation for/with this reducer.
As zvrastil said, I also do not think that several taken action with PI could iron out this misshaped stars.

Regards

[zvrastil]

Emiel,

one more thought - try to use CCDInspector software (if you did not yet) to collimate the scope while your camera is on.

cheers, Zbynek

[Hans Pleijsier]

I am not a guru either but there is nothing against some trial and error, I hope ...

So give me access to a fit- or a tiff-file and I will sit on it some time this weekend.
And provide me with the RA and DEC of a designated star in the frame because I might use PI's StarGenerator to produce a mask.

I will come back with a result, good or bad. I think that will trigger the guru's ...

HP

PS: And off course I would advise you too to fix the mechanics before your next shoot but for now we will try to rescue what can be rescued.

[Emiel Kempen]

Hi all,
Thanks for the replies!
You are all right in that the problem with the misformed stars has to be solved on the scope first.
The reason I am looking for a solution in PI is that I have 3 nights of images with these misformed stars.....

one more thought - try to use CCDInspector software (if you did not yet) to collimate the scope while your camera is on.
cheers, Zbynek

How did you collimate your Newtonian?
The reason could be that your actual collimation is not in harmony with your reducer and that you will have to optimize your collimation for/with this reducer.
As zvrastil said, I also do not think that several taken action with PI could iron out this misshaped stars.

I am using this method to set the correct offset of the secundairy mirror: http://www.astrosysteme.at/images/Collimating_ASA-Astrographen_E.pdf see the "Adjusting the Offset (only for advanced users)" at the end of the document.

Then I use a combination of my Hotech SCA lasercolimator and my Catseye sighttube, cheshire and autocolimator to colimate and finetune.

After that I carefully focus with a Bahtinov mask and shoot 3 images of a starfield without bright stars.
These 3 images I run through CCDInspector and see what it looks like and try to correct it. I have attached the fieldcurvaturemap and the aspectmap of these images.

The fieldcurvaturemap looks ok to me and the aspectmap shows exactely where the misformed stars are, but how to fix this with colimation......

I am not a guru either but there is nothing against some trial and error, I hope ...

So give me access to a fit- or a tiff-file and I will sit on it some time this weekend.
And provide me with the RA and DEC of a designated star in the frame because I might use PI's StarGenerator to produce a mask.

I will come back with a result, good or bad. I think that will trigger the guru's ...

HP
PS: And off course I would advise you too to fix the mechanics before your next shoot but for now we will try to rescue what can be rescued.

I have put a RAW file from my Canon 40D on my website. http://www.ekempen.com/raw.zip
It is a 120sec ISO 1600 image of SH2-124, shot from Hoogeveen, The Netherlands.

Regards,
Emiel.

[Hans Pleijsier]

Emiel,

The aberration in your frame is very special. To my surprise the maximum distortion is not in the ultimate NE corner but it is located just below the corner. My experience with Newtonians is zero so I can not help you with this. What follows is a quick fix and I must say the procedure requires more time than I hav put into it ... but maybe this workflow gives you some perspective to find a better solution.

---

First I made a crop of the affected problemarea. (file kempentestbegin is the jpg of a fit).

Then I made a duplicate and work from there to make a mask.

Apply ATrousWaveletTransform b3-spline 5 layers but cancel out the first four layers.
The result is the basis for a mask that leaves a <hole> for the stars with the biggest aberration. (I am sorry, you will miss the small ones).

Use Histogram and rangeselection to optimize your mask.

Apply the mask to the original frame. Now your picture is protected and you can drill, hit and play with the problemstars without effecting the rest of the frame.

Under the menu MASK-show mask you can select/deselect the mask. Deselect it. The mask is still there but you wont see it.

Now you could apply the tool <<convolution>>. Select library. Pick a filter. If you have the time you can build your own filter tailormade for the aberration at hand. I made one very quickly but here you have a world to explore.
Verify that the mask is effective and the other stars are not effected by the passfilter.

Apply histogram and curves.

Use pixel math to past this area back into you original frame.

---
Meanwhile I would ask the javascript-gurus to design a script with which - after a mouseclick you can paste backgroundpixelvalues to unwanted starpixels outside a certain user-defined radius of the PSF-centre.

Zbynek, I know that you can build this in 15 minutes ;-)

HP

[Emiel Kempen]

Hello Hans,
Thanks very much for your efforts!
That is looking much better, with some fine tuning it could be even better.
I will surely try this when I can find the time for it.
I think the problem causing the distorted stars may be a wrong distance between the corrector and the camera.
Again, thanks for your time.
Regards,
Emiel.

[zvrastil]

---
Meanwhile I would ask the javascript-gurus to design a script with which - after a mouseclick you can paste backgroundpixelvalues to unwanted starpixels outside a certain user-defined radius of the PSF-centre.

Zbynek, I know that you can build this in 15 minutes ;-)

I missed this one... Well, that's really an overstatement 8). First, I'm not javascript guru. I am indeed software developer and I have like 12 years of praxis with C++ and C#. So, my natural environment is PCL. Frankly, before PixInsight I wrote no single line of JS code, nor code in any other dynamically typed language like Python etc. I found JS useful for automating some of my processing steps (or just to experiment), it would take me a long time to create complete script with user interface etc.
This does not mean your idea is not worth considering. But to be honest, I don't really understand what you mean. Could you be more clear?

For that particular case, I can imagine (just wild thinking) something like
 - take a PSF model as an input,
 - detect stars,
 - for each star, scale PSF to match the star as as possible
   - do the difference between actual star image and scaled psf
   - add the difference to the mask image
 - select the generated mask for the image
 - do morphological erosion with mask applied

However, there's a lot of catches - close stars whose PSFs partially overlap, diffraction spikes, star-like features in nebulae or galaxies to avoid... So I'm not sure if all this is worth the effort.

regards, Zbynek

[Hans Pleijsier]

Zbynek, your algorithm is the super gt-version of what I had in mind.
Because the main application is/should be limited to correction of not-so-well-shaped stars I think that the process could be like this:
- place/centre a preview-window on the star at hand;
- replace all starpixels outside radius r by new pixels that has the same pixel value as the local background of the preview
- where r is defined by the PSF generated for this frame.

OK ...  30 minutes