fixing elongated stars

[CarlCC]

I'm new to PI and working with the latest trial version.

Is there a process in PI  for fixing star elongation (egg shaped stars)?

I can do it in Photoshop with a process involving the min filter.

Comments greatly appreciated
Thanks
Carl

[Nocturnal]

Hi Carl, welcome to PI. There isn't a specific tool to do that but you could try an erosion filter with a star mask.  Go to Harry's Video page and find the star size reduction video that shows how to do that. Try using an oval pattern instead of a round one.

I haven't actually done this myself so I don't know how well it works. My advise is to find the source of your oval stars and fix it There's only so much you can do to remove them in post processing. If a star is spread out over say 2 extra pixels this also happens to all your other light. Effectively you're slicing your exposure length in 1/3 and losing resolution as the light gets smeared. I always limit my exposure length to where I get no more than 0.5 pixels of drift per image.

Anyway, let us know how the erosion filter works for your image.

[Carlos Milovic]

Hi Carl

You have two (main) options to achieve that:
1) Deconvolution
2) Morphological Filters

1) Not the easiest way, but it should give the better results. First you'll have to build the PSF. You may try a rotated and asymmetrical gaussian profile (See at the Software Development subforum for a process I created to read that kind of PSF from your stars, ReadPSF). Then, you'll need to carefully set all the parameters... Juan published a tutorial some time ago. It would be best if you read it before playing around. I recommend to you the Regularized Richardson Lucy algorithm (in the Deconvolution process), since it will generate less artifacts than the others (also those included in the RestorationFilter process).
2) This is much easier, but may destroy a bit your data. I'd recommend to use masks to protect all the background stuff, or you'll end with a "painted like" look. Use as smaller structural elements as you can (3x3 or 5x5), and make use of several ways (again, there is a tutorial somewhere about this tool...). Finally, try using a selection operand, around 0.1 or 0.2 instead of the minimum operator, to avoid dark pixels to grow too much.

Also, a combination of both methods work pretty well. Of course you'll never end with a image as good a one without that problem to star width, but even so it will significantly better than the current state.

[Jack Harvey]

I have had very good luck using the motion blur tool of deconvolution.  First make a preview of an area with some good examples of your ovoid stars.  Then you have to play with the two sliders to get the correct angle (I cannot remember if you set the angle 90 degrees to the angle of elongation or with the angle of elongation but you will see the results quickly and set it right).  The it is only a matter of trial and error with the length slider to correctly reduce the amount of elongation.  Once you have a result you like then apply to the whole image.

I have only had to use this a couple of few times and it has worked like a charm.  Good Luck and report back your success/failure.

[CarlCC]

Thanks everyone for the input and suggestions.
 I presently use CCDStack and Photo-Shop for most of my processing. In getting to understand the features of PI , I am attempting to compare how PI does the things that I do in CCDStk/PShop.

Although I don't have many elongated stars to deal with , when necessary, PShop has a quick and simple process of selecting the area of interest pasting in a separate layer, blending with the darken mode and fixing the  elongation with the offset filter.

So my basic question  really  is , can I use PI to do all the thing that I do with PShop , or do I still need PShop for some specific fixes.

Any further comments greatly appreciated.

Thanks

Carl

[Jack Harvey]

I quit using PS a couple of years ago.

[Carlos Milovic]

I've not used PS for any astrophoto processing since 2003 or 2004

[Niall Saunders]

I'm just glad I didn't waste any money on PS in the first place

Cheers,

[RBA]

You guys are not quite answering Carl's question...  The behind-the-lines message is "we don't need PS so probably you won't either" but that's not telling Carl how to resolve the problems he's used to solve in a particular way with another software.

Carl, there's something you should keep in mind. Sometimes a particular workflow in PS doesn't have a parallel process in PI. What I mean is, yes you can probably solve that problem in PI, but the tools and the workflow will likely be different. And sometimes, trying to replicate the steps will lead you to frustration because there are things you can do with PS that you cannot do with PI, and viceversa, if what you try is to replicate step-by-step the workflow. That however is different from saying that the problem cannot be solved using either software. Often times PI offers an alternative way to do it, and often times it gives much better results.

For example, I could try to describe how to do that "selecting the area of interest", creating a new image, blending it with the "darken mode" (which is basically a min(img1,img2) operation in PixelMath), etc. but... would that be the best way to do it in PI? Probably not.

And people have pointed you to the right tools: Deconvolution and the MorphologicalTransformation (MT) tool.

Now the "trick" is learning how to use those tools to achieve the goals of fixing the problems these tools can solve, and again, usually delivering superior results. For example you mentioned the Minimum filter... Once you get to know how the MT tool works, you'll realize it's like a maximum/minimum filter on steroids, and LOTS of them! In fact, saying that probably doesn't do justice to the MT tool.

But new tools take time to learn, practice, etc. It's not difficult, but I see many people getting into PI coming from PS and because after a few hours they cannot do in PI what they already know well how to do in PS, they give up. That's unfortunate but it's their loss. Anyway I don't want to turn this into a lecture (I hate to lecture almost as much as being lectured ), since you just want a straight answer, so I'd suggest you to experiment with these two tools, always use a star mask with them for this purpose like Carlos pointed out, protecting everything but the stars (easy to do with the StarMask tool), and if you get lost with all the parameters, since there's still no official page with the PI documentation, in the meantime you can use this "Unofficial PI Reference Guide" http://blog.deepskycolors.com/PixInsight/ and of course, Harry's videos for several good introductory (and not so introductory) video tutorials: http://www.harrysastroshed.com/pixinsighthome.html

[Jack Harvey]

Rogelio  I beg to differ on some of your discussion.  The initial question was how to deal with elongated stars.  I not only suggested which tool but what part of that tool to use and how to adjust the settings of the tool to achieve the desired result.  I would suggest your ......is like min/max on steroids is certainly vague and does not address the many options on that tool to deal with star rounding.

I agree my comment on not using PS does not address his question and your suggestion that similar results can be obtained with either software package, but in a different manner is right on.  I suspect many will use both tools and pick and choose among them as to what works for them.  That was my initial approach and with time I discovered more and more the ways to perform things with PI - so the transition although complete was gradual.

At any rate welcome aboard Carl and hope you find the forum useful.

And finally Congratulations to Rogelio for todays APOD.  Done mostly with PI I assume<G>.

[RBA]

Rogelio  I beg to differ on some of your discussion.  The initial question was how to deal with elongated stars.  I not only suggested which tool but what part of that tool to use and how to adjust the settings of the tool to achieve the desired result.

Yes, absolutely. That first line in my comment was only referring to the last few one-liners about not using PS. I did later make the point that you all had indicated him the tools in PI Carl could use to work on that problem.

I would suggest your ......is like min/max on steroids is certainly vague and does not address the many options on that tool to deal with star rounding.

I also agree, and that's why I added that saying that doesn't do justice to the MT tool. It was just a poor way to indicate that if you look for a minimum filter, you have that on the MT tool "and a lot more".

And finally Congratulations to Rogelio for todays APOD.  Done mostly with PI I assume<G>.

Thanks! The processing once all frames were put together was done all with PI, but frame adaptation was done entirely with other software. Maybe later I post a message about that, it might be an interesting discussion.