Jump to content


Photo

Eagle reprocessed with PixInsight

  • Please log in to reply
9 replies to this topic

#1 cuivienor

cuivienor

    Viking 1

  • -----
  • Posts: 562
  • Joined: 06 Oct 2010
  • Loc: Tokyo, Japan

Posted 08 July 2013 - 03:27 AM

Hello,

I have recently been doing my shopping for image processing software, "thanks" to super cloudy and rainy nights (and an impromptu thunderstorm that came in amazingly quickly one night that appeared to be clear). Up to now I have looked at free solutions (Gimp 2.9 with 16 bits processing, Fitswork, DSS), Startools, Pixinsight.

While I was a big big fan of Startools very quickly, I recently looked at just the newbie tutorials from Harry for PixInsight, and got the attached picture for the Eagle Nebula (very difficult to process image with loads of noise and no dithering).

Compared to my earlier Startools attempts this is much better. http://www.cloudynig...5921787/page...

Maybe it is my IT background, but Pixinsight seems like an amazing tool for the moment, and I guess I will need to give them my money by the time my trial expires!

Cheers,

Yannick

Attached Files



#2 Wouter D'hoye

Wouter D'hoye

    Surveyor 1

  • *****
  • Posts: 1519
  • Joined: 27 Jun 2003
  • Loc: Belgium

Posted 08 July 2013 - 05:33 AM

Hi,

A nice image. But imo the processing looks a bit overdone. I believe you went a bit overboard with noise reduction creating a somewhat blotchy effect. Sometimes it's more natural or pleasing to leave some of the noise in the image.

Kind regards,

Wouter.

#3 cuivienor

cuivienor

    Viking 1

  • -----
  • Posts: 562
  • Joined: 06 Oct 2010
  • Loc: Tokyo, Japan

Posted 08 July 2013 - 06:28 AM

Thanks Wouter for the nice comment - you are too generous, it's a recognizable picture but nowhere near nice :) you are probably right about the noise. This is actually my first DSLR picture, without cooling, dithering, and a few darks only, the noise is quite horrendous. Actually I checke against the lossless version of the file I have, and the background is much better there, so probably the jpg conversion for posting did not help.

By the time the skies clear, I should be able to use a cooled DSI III so hopefully will take much better (or at least less noisy) pictures.

Thanks,

Yannick

#4 Raginar

Raginar

    Fly Me to the Moon

  • *****
  • Posts: 6138
  • Joined: 19 Oct 2010
  • Loc: Rapid CIty, SD

Posted 09 July 2013 - 07:50 AM

Yannick,

After you do ACDNR, try redoing your HistogramTransformation. It tends to stretch your data and you need to move your black point. That should take care of the splotches. The other thing is to leave some room in HT for this... I find if I stretch it to where I would've in PS, that's too much for a final HT at the end.

That's just me though.

Good luck!

#5 jsines

jsines

    Mariner 2

  • *****
  • Posts: 256
  • Joined: 06 Sep 2011
  • Loc: Berkley. Michigan

Posted 09 July 2013 - 12:12 PM

A nice image. But imo the processing looks a bit overdone. I believe you went a bit overboard with noise reduction creating a somewhat blotchy effect. Sometimes it's more natural or pleasing to leave some of the noise in the image.



Yes, it looks like the noise reduction was a little too much. I get this all the time and I've been working really hard on trying to avoid excessive noise reduction. I've had PixInsight for about 9 months now, and still have this issue. I go overboard with all the noise reduction processes, then need to dial it back.

Lately I've been skipping AtrousWavelets in the pre-linear stage, and then dealing with NR in the post-linear stage. It seems to help with me, but YMMV. If you use AtrousWavelets, a mask is a must.

I would try a few different versions of the same image processing steps, but take out different noise reduction processes in each image. One of the nice things about PixInsight is that you can work on the same image in different ways at the same time, like a tree branching out to different branches, etc.

I think this image is also difficult because of the excessive nebula against the background, it was probably hard to find a spot that was true background for the background neutralization.

Good job, though. I like the potential here.

#6 Raginar

Raginar

    Fly Me to the Moon

  • *****
  • Posts: 6138
  • Joined: 19 Oct 2010
  • Loc: Rapid CIty, SD

Posted 09 July 2013 - 12:44 PM

Doing NR via deconvolution (L only) + TVGDenoise is pretty good.

Chris

#7 cuivienor

cuivienor

    Viking 1

  • -----
  • Posts: 562
  • Joined: 06 Oct 2010
  • Loc: Tokyo, Japan

Posted 09 July 2013 - 10:58 PM

Thank you very much for all the help!

Indeed, after the NR, I did do a HT again to take advantage of the space made for the dark point, but going further would have started to clip data. It actually looked better, but somehow I couldn't bring myself to clip data, thinking: yes I am removing noise/blotching here, but at what cost?

Otherwise, I will try to reprocess using the various techniques you have described - I may yet get an actual nice image out of it!

I in particular have not use AtrousWavelets, I'll try that as well. Also, I am having a hard time getting deconvolution right (and generating a good star mask with an environment like that one, full of stars), but again lots of things to test, and so little time!

Thanks again for all the recommendations!

Yannick

#8 Madratter

Madratter

    Fly Me to the Moon

  • *****
  • Posts: 6337
  • Joined: 14 Jan 2013

Posted 09 July 2013 - 11:13 PM

I am FAR from good with deconvolution. However, I will say that one thing that helped me a great deal was using Dynamic PSF to come up with the point spread function used by Deconvolution.

#9 cuivienor

cuivienor

    Viking 1

  • -----
  • Posts: 562
  • Joined: 06 Oct 2010
  • Loc: Tokyo, Japan

Posted 10 July 2013 - 11:45 PM

Thanks for all the tips, I have redone the processing, experimenting with ATrousWavelets, Star Masks, etc. Lots of fun.

I got the attached image. Better? I don't know. But it was fun to do :)

Attached Files



#10 cuivienor

cuivienor

    Viking 1

  • -----
  • Posts: 562
  • Joined: 06 Oct 2010
  • Loc: Tokyo, Japan

Posted 11 July 2013 - 06:53 PM

By now I'm talking to myself, but here it is reprocessed again. One thing I am missing in Pixinsight is the lasso tool: I would absolutely lasso that dust mote, feather the selection, and then apply a histogram separately there...

Attached Files








Cloudy Nights LLC
Cloudy Nights Sponsor: Astronomics