Star removal

[Clive]

Since the start of the new year, I've been playing about creating a star removal program to run on my IRIS images.  I have used the stand-alone version of StarNet in the past but it has what I consider to be two disadvantages, the need to stretch the image prior to removing the stars and its performance on removing saturated stars.  My program operates on either linear or stretched data and appears to remove the saturated stars a little better.   The following images (a crop of M31) are (top to bottom) the linear image, StarNet applied to the linear image, StarNet applied to a stretched image (Gamma = 2), StarNet applied to a stretched image (Gamma = 4) and the output of my program applied to the linear data.

 

 

The viewing black point on all images is set to 0 ADU and the white point set to 1000 ADU hence most of M31 is blown out but the faint stars show up clearly (the linear image has a ADU range 0 to 25000.  To allow direct comparison , the stretched StarNet++ images have all been restored back to linear by applying the inverse gamma value.

 

I'm not sure if the program has been 'tweaked' such that it only performs on images created using my DSLR, so if anyone could make available an image in 16 bit TIF format (the TIFF image straight out of DSS is ideal) I would be interested in trying it out.

 

[+Derbyshire Dave]

Looks really impressive Clive, I'd be interested know how you have developed it. Any particular sort of image you require?, I can let you have a 16 bit tiff for the flaming star which I've just posted if it's any help. It was stacked in PI, but I can post a 16 bit unstretched version on Flickr and you could download it. Can re-stack in DSS if you like..

[Clive]

47 minutes ago, Derbyshire Dave said:

Looks really impressive Clive, I'd be interested know how you have developed it. Any particular sort of image you require?, I can let you have a 16 bit tiff for the flaming star which I've just posted if it's any help. It was stacked in PI, but I can post a 16 bit unstretched version on Flickr and you could download it. Can re-stack in DSS if you like..

That looks a perfect test image Dave.  If you can post the unstretched image (pre-processing such as background removal is OK) I'll see what it will do.  IRIS (being so old) tends to be a bit finicky on TIFF files, it doesn't seem to like the latest 'flavours' and 32 bit are certainly out of the question.  If the PI image won't load then a DSS version should do.  I'll make available (public or private as you prefer) the starless image (good or bad!) without any additional processing so, you if you want, you can apply the same processing methods that you used.

 

It's based upon the assumption that imaged stars generally have a 3D Gaussian profile (even though the saturated stars are clipped).  The major problem I faced was finding a way to extract a reliable background level at each point in the image and your image certainly will test that!

[+Derbyshire Dave]

Ok, here we go Cllive.. Embeddedas a picture below, You should be able to download from Flickr..

 

In case you have any issues and have to put the address into your browser directly I've added three more w's to the address, which you can copy then remove..

 

https://wwwwww.flickr.com/photos/143213482@N03/51113687501/in/album-72157708765466056/

 

This is straight out of PixInsight stacking, and saved as a 16 bit tif

 

I'lll be interested to see how you get on,

Dave

 

 

 

 

 

 

 

[Clive]

I seem only to be able to download a JPEG image rather than a TIFF (although my experience with Flickr is none so any help to get to the TIFF file will be appreciated!)

[+Ibbo]

You might have to do it via PM and e-mail or dropbox.

[+Derbyshire Dave]

35 minutes ago, Ibbo said:

You might have to do it via PM and e-mail or dropbox.

Yeah, of course.. do you want to pm me you e-mail address, and I'll send you a DropBox link via e-mail

[Clive]

15 minutes ago, Derbyshire Dave said:

Yeah, of course.. do you want to pm me you e-mail address, and I'll send you a DropBox link via e-mail

PM sent.

[+Ibbo]

Dave do you want to send Clive the NGC3718 with cc I sent you over as well?

 

[+Derbyshire Dave]

15 minutes ago, Ibbo said:

Dave do you want to send Clive the NGC3718 with cc I sent you over as well?

 

Yeah, will do Steve

[+Derbyshire Dave]

Hi Clive

  Images from myself and Steve are in he shared Dropbox directory

Cheers

Dave

[Clive]

15 hours ago, Derbyshire Dave said:

Hi Clive

  Images from myself and Steve are in he shared Dropbox directory

Cheers

Dave

 

Hi Dave, thanks for you and Steve sharing you images, it was certainly a useful exercise.  Your image had a border down the left-hand side so I had to crop it by 10 pixels to remove it.  Your image also revealed a bug in my program that I've now corrected (trying to write a value > 32767 to a signed integer variable) so thanks for that.  Steve's image has diffraction spikes and my program is not designed to remove those (although it seems to have done as good a job as StarNet).

I've PM'ed you a OneDrive link to the starless files I've generated together with starless images created by StarNet for comparison.  I would certainly value your opinion.

 

[+Ibbo]

I did wonder but worth the exercise I think.

[+Derbyshire Dave]

2 hours ago, Clive said:

 

Hi Dave, thanks for you and Steve sharing you images, it was certainly a useful exercise.  Your image had a border down the left-hand side so I had to crop it by 10 pixels to remove it.  Your image also revealed a bug in my program that I've now corrected (trying to write a value > 32767 to a signed integer variable) so thanks for that.  Steve's image has diffraction spikes and my program is not designed to remove those (although it seems to have done as good a job as StarNet).

I've PM'ed you a OneDrive link to the starless files I've generated together with starless images created by StarNet for comparison.  I would certainly value your opinion.

 

Ah, the old signed/unsigned integer issue. Are you programming in C or C++? Interesting the border thing, I think that must be the PI stacking, I did crop in a bit in my original processing.

 

I didn't get the PM with the OneDrive link btw, and yesterdays PM now also gone.

 

Hats off to you Clive for tackling this! 👍

[Clive]

17 minutes ago, Derbyshire Dave said:

Ah, the old signed/unsigned integer issue. Are you programming in C or C++? Interesting the border thing, I think that must be the PI stacking, I did crop in a bit in my original processing.

 

I didn't get the PM with the OneDrive link btw, and yesterdays PM now also gone.

 

Hats off to you Clive for tackling this! 👍

 

Ok, I've tried creating a new PM (rather than replying to the old one as I did earlier) so I hope you get it (otherwise I could post the link on here?)

My initial intention was to use C but in the end I plumped for what (these days) I know the best - VB - and it actually currently runs from an Excel spreadsheet, although once I happy with it and all the other 'tools' I've created, I may use Visual Studio to create a standalone executable version.