Author Topic: Please, help processing NGC6960 (Solved/Closed)  (Read 12099 times)

Offline pfile

  • PTeam Member
  • PixInsight Jedi Grand Master
  • ********
  • Posts: 4691
    • View Profile
Re: Please, help processing NGC6960
« Reply #15 on: 2012 September 13 22:30:05 »
well, that does look a lot better. however, there are a lot of hot pixels visible in the integrated result. what integration method did you use and are your subs dithered? how many subs = 25 min?

also there seems to be something wrong with the alignment. or maybe the hot pixels are giving the impression of a mis-registered subs.

Hi Pfile,

It was 10 subs of 2 1/2 min. each, so 25 min. in total.

I use MaximDL for Image Adquisition, and I dither my subs.

Regarding integration, I used for:

Image Integration
   -Combination: Average
   - Normalization: Additive
   - Weights: Noise Evaluation
   - Generate Integrated image: Checked
   - Evaluate Noise:  Checked

Pixel rejection (1):
   - Rejection Algorithm: Winsorized Sigma clipping
   - Normalization: Scale + zero offsett

Pixel rejection (2):
   - Sigma Low: 4
   -Sigma High: 3

Are these parameters OK? Would you suggest something different?

Thanks for your help,

M.

well, i'd say those parameters are fine, but you will want to experiment with the Sigma low and Sigma high values to get values which properly reject your hot pixels. try making them progressively smaller and look at the rejection maps. if the rejection maps start looking like your DSO then you are rejecting too many pixels.

also i wonder if the hot pixels are messing up your alignment. you can use the "blink" process to run a movie of all your aligned subs. it will be easy to see if they don't actually line up as you step through the images.


Offline mcbbcn

  • PixInsight Enthusiast
  • **
  • Posts: 78
    • View Profile
Re: Please, help processing NGC6960
« Reply #16 on: 2012 September 14 21:53:46 »
Hi Pfile,

I went ahead and took 4+ second flats and no change.  Also, I recreated the master dark to be true 32 bit, and no change.

So, I'll experiment with some of the values that you're referring to and see what happens.

I assume the blink process is a Pixinsight process.  I know what you mean by blink, but I've never used it in PI or any other software, but I'll figure it out.

Thanks for all your help.  You've been great through this process.

M.

Offline mcbbcn

  • PixInsight Enthusiast
  • **
  • Posts: 78
    • View Profile
Re: Please, help processing NGC6960
« Reply #17 on: 2012 September 14 23:44:07 »
Hi Pfile,

I learned something new.  I noticed that all of my sub-flats had hot pixels, even when they were calibrated.  So, during my Master Flat creation, it was leaving all the hot pixels in the Master Flat even when I was calibrating it for substracting dark & bias noise.  But when I changed the 'CFA pattern detection' from "Detect CFA" to "Force CFA", then it was removing all the hot pixels. Have you heard about this behavior before"?  The only issue (and it's not a small one) is that when I integrate, all the colors are gone, so it seems that it has something to do with the CFA option.  If you have any thoughts, please let me know...

M.

Offline gvanhau

  • PixInsight Old Hand
  • ****
  • Posts: 345
    • View Profile
Re: Please, help processing NGC6960
« Reply #18 on: 2012 September 15 09:43:41 »
Hello

Here are initial the process icons I normally use when processing a set of images of QHY8L.
There is nothing really special
I don't use bias in light calibration, because bias is already included in dark's.
Dont check Dark frame optimization during light calibration, because this will also scale the amp glow.

Note: I replace the debayer icon with the debayer script, and  this icon is only there for doc purposes.

After color calibration, I usually continue with noise reduction, deconvolution, and stretch.

Regards
Geert
Geert Vanhauwaert

Offline pfile

  • PTeam Member
  • PixInsight Jedi Grand Master
  • ********
  • Posts: 4691
    • View Profile
Re: Please, help processing NGC6960
« Reply #19 on: 2012 September 15 10:04:04 »


I assume the blink process is a Pixinsight process.  I know what you mean by blink, but I've never used it in PI or any other software, but I'll figure it out.

M.

"blinking" is astronomy jargon - it means to swap 2 images quickly such that your eye/brain can easily see the changes between two images. in pixinsight you can blink more than 2 images with the Blink process. it will be completely obvious if the images are mis-registered. more commonly though people will use blink to see if there are any bad subs with bad tracking or bad focus or what have you.

Offline mcbbcn

  • PixInsight Enthusiast
  • **
  • Posts: 78
    • View Profile
Re: Please, help processing NGC6960
« Reply #20 on: 2012 September 15 11:44:20 »
Pfile,

Your insights have led me to discover that my flats (subs) are full of hot pixels.  For whatever reason, even when I'm calibrating them, they still have the hot pixels, which the end up in the master flat, which they end up eventually in the final image.  Any thoughts about how to remove the hot pixels from the subs?  I'm doing the usual calibration on the flat-subs for darks/bias,  but is not working.  I was able to get some success removing the hot pixels from the flats-subs by selecting 'force CFA', but that ended up messing my colors, and I ended up with an image that had no color, so something is wrong.  I've taken the flats two times, and still the same issue.  Is it normal to end up with a bunch of hot pixels in the flat-sub?

Here is an example of a flat-sub that has been calibrated, notice all the black dots...

http://dl.dropbox.com/u/8971416/NGC6960/scope__1x1_skyflat0_c.fit

I tried to remove them during ImageIntegration for creating the Master flat using very restrictive sigma high/low values, but they don't go away...

Thank you,

M.

Offline pfile

  • PTeam Member
  • PixInsight Jedi Grand Master
  • ********
  • Posts: 4691
    • View Profile
Re: Please, help processing NGC6960
« Reply #21 on: 2012 September 15 12:29:34 »
well the problem with flats is that you can't really dither them (they do not need to be aligned of course). so the hot pixels will line up right on top of one another and they don't look like statistical outliers, they look like signal. so rejection does not work.

i don't know much about that camera, but i do wonder if it's normal to have so many hot pixels.

at any rate, it sounds like you may need to investigate DefectMap clean your flats before you integrate them.

Offline mcbbcn

  • PixInsight Enthusiast
  • **
  • Posts: 78
    • View Profile
Re: Please, help processing NGC6960
« Reply #22 on: 2012 September 15 19:25:52 »
Hi Pfile,

I was working in parallel with another AP friend (Dave Raphael who sold me this CCD a while ago), and he suggested that smooth the flat, which I did using a Kernel custom filter in MaximDL, and it made quite a difference.

Here is the image before DBE:

http://dl.dropbox.com/u/8971416/NGC6960/afterkernelfilter/NGC6960_integration_before_DBE.fit

...and here is the image after DBE:

http://dl.dropbox.com/u/8971416/NGC6960/afterkernelfilter/NGC6960_integration_DBEv2.fit

Please, let me know what you think.  I think at maximum stretch, you can appreciate the big doughnut which I believe is the central mirror of my Maksutov, and then you can see pretty stubborn dust mote.

What would you recommend as next step?

Thanks for your help,

M.

Offline gvanhau

  • PixInsight Old Hand
  • ****
  • Posts: 345
    • View Profile
Re: Please, help processing NGC6960
« Reply #23 on: 2012 September 15 21:35:10 »
Hello I just saw the two images.

I think you can do a  better DBE, placing more samples near the corners. (There is a red cast towards the lower left and right corner. )

The donut seems to be a reflexion of the big star on the IR window of the camera. So i think it can not be removed using flats, may be it can not be removed at all.

After a better DBE i think the next step sould be noise reduction using a good star mask.

Regards
Geert
Geert Vanhauwaert

Offline mcbbcn

  • PixInsight Enthusiast
  • **
  • Posts: 78
    • View Profile
Re: Please, help processing NGC6960
« Reply #24 on: 2012 September 16 21:51:26 »
Hello I just saw the two images.

I think you can do a  better DBE, placing more samples near the corners. (There is a red cast towards the lower left and right corner. )

The donut seems to be a reflexion of the big star on the IR window of the camera. So i think it can not be removed using flats, may be it can not be removed at all.

After a better DBE i think the next step sould be noise reduction using a good star mask.

Regards
Geert

Hi Geert,

I went ahead, created a star mask, and applied SCDNR for noise reduction.  Also, I used the Histogram to lower the amount of stretching.  This is still a 25min. exposure from Bortle 8 sky, so noise is expected, but if the image is overstretched, then it's too much noise.  I tried to play around with mask and the histogram to star reducing the amount of the reflection in the big star but it's a bit of a trial and error...

Here is the result:

http://dl.dropbox.com/u/8971416/NGC6960/afterkernelfilter/NGC6960_integration_before_DBE_SM_SCNDR.fit

Any other thoughts from you or anybody else?

Thank you all for your help,

M.

Offline mcbbcn

  • PixInsight Enthusiast
  • **
  • Posts: 78
    • View Profile
Re: Please, help processing NGC6960
« Reply #25 on: 2012 September 19 20:50:51 »
Friends,

Now that you've helped me overcome the issue with the flats, I'm going to close this post and go back to create a new post on what type of tips you suggest when processing OSC pictures.

Thank you all for your help.

M.