›

Correcting Black Pixels in Star Cores

Reggie JonesReggie Jones
in PixInsight
Hi.  I have sub exposures of IC 4604 (Rho Ophiuchi Nebula) where the cores of about a half dozen stars have one pixel (in small stars) to several pixels in a couple of the larger stars that are black.  I think it is in all the sub exposures (LRGB) but it's most noticeable in the luminance subs.  I spent the time getting the x,y positions of all the affected pixels and then I tested a pixel math calculation to change the single pixel problem stars to what I calculated / guessed was the average grayscale of the neighboring pixels around it.  It seems to work.  The expression I used is:

iff (((x()=="number")&&(y()=="number"), 0.8, $T)

My problem is how to set up a similar calculation do this in the stars with multiple black pixels in the cores - they're not aligned specifically to any geometric shape but I think I could adapt a square, circle or rectangle and maybe calculate something that way.  I've gone over your Pixel Math video a few times and still trying to crunch through it to develop a solution.  Ultimately, I'd like to develop something I can use to process all the subs via Image Container to correct them.

Questions;

1. Do you think I'm on the right track?
2. Is there an easier way to solve this problem?  I don't mind the amount of work involved but I'm guessing someone has run into this issue before me and has worked something out that may be easier.
3. Is there any documentation you can suggest to use as a reference to improve competency in pixel math?  I haven't found any.  Warren Keller's book doesn't much mention it.

Thanks

Reggie Jones

Comments

  • BlockHeadBlockHead
    Reggie,

    This does not seem like an easy way to go about doing things. The manual measurement of position coupled with a non-algorithmic guess at substitute value I think are going to lead to less than desirable results and or at east the ratio of result/time is a small number. 

    1. I really think you need to find why the black pixels are there in the first place. The two obvious places are Drizzle and for OSC, Cosmetic Correction. Were any of these employed?
    2. i assume the pixels in the center of the stars are not *actually* zero valued. If they were, and you do not have zeros anywhere else... both Cosmetic correction or Defect Map could be used. Yes, you could make a Pixel math expression..but they do become difficult when gathering information on neighboring pixels.
    3. I assume you watched my Pixel Math sections in Fundamentals? 

    Gerald (OldWexi) on the Forum I know put together many many videos on PixelMath: http://www.werbeagentur.org/oldwexi/PixInsight/PixInsight.html

    I have not watched all of them...but it might be a good resource. 
    (scroll down to the English versions)

    One of the these, I will make a less that explains how I came up with the expression for the Elliptical shape with a gradient. That expression combines just about all facets of Pixel Math.

    -the Blockhead

  • Reggie JonesReggie Jones
    Thanks - no, I didn't think it would be easy to do but I was challenged to come up with anything at the time to work on the subs.

    These subs are what I received from my imaging request of this object from Telescope Live.  The equipment used was AUS-2.  I found the problem when I used Blink and reviewed the sub exposures for the first time.  No processing was done.

    Yes, I watched your Pixel Math sections more than a couple of times each as they're the only in depth resource I've seen that fully explains how it works.  I'll probably have to watch them a few more times as I'm using it before I'm completely comfortable (kind of how I learn things).  Thanks for the additional link, I'll check it out.

    What I've done for now is used the Clone Stamp tool to correct the main stars affected after I integrated the luminance subs so the problem isn't readily visible.  That seems to have worked so far.

    Thanks

    Reggie
  • BlockHeadBlockHead
    OK. I understand. Yes, they are calibrating on their end. 
    You should see if the black bits remain after Image Integration. I have seen them in Telescope Live data..and ignored them as they do reject out (usually). 

    It is late now..but I bet you could create a star mask... then apply a bias to the stars using MLT..and that will eradicate the black blips. OR you could , again with a star mask, apply a little convolution in combination with MAXimum. Now I am getting creative!! Maybe I should make a lesson on this! (just thinking out loud).

    -the Blockhead
  • Reggie JonesReggie Jones
    For the small stars with only one or 2 pixels with problems, all of the black bits did go away after integration. Only the large star with many (20 or so) pixels with problems still had a few pixels with black bits that I fixed with the Clone Stamp tool.  

    A lesson on how to correct these issues would absolutely be great!

    R 
  • BlockHeadBlockHead
    Hi Reggie,

    Could you make a handful of calibrated frames with the issue available?
    I will have a look at them.
    Thanks,
    -the Blockhead
  • Reggie JonesReggie Jones
    Yes, I can - how would you like me to send them to you?

    Reggie
  • BlockHeadBlockHead
    You will need to put the files cloud server somewhere (google drive, OneDrive, DropBox..etc) for me to download. You can reach me at ngc1535@caelumobservatory.com .
    -the Blockhead
  • Reggie JonesReggie Jones
    Ok - I just sent you a link to a shared DropBox folder labeled as "AdamBlockDownloads".  Let me know if you didn't receive it.

    Reggie
  • BlockHeadBlockHead
    HI Reggie,

    I looked at the data. The problem isn't simple. It isn't really black cores... and they are not small. The type of artifact in the centers of stars varies across the field. Interestingly, the stars in the center do not show the artifact (even though they are saturated as well). I wonder if it is because of flat fielding. 
    I sent a note to Telescope Live admin asking some questions. I also note this data is mostly cloudy..there were only two good frames here. I assume you know this?

    So... please let me see what they say. 
    Given this data.. if you want to manually do something...to hell with pixel math. Create a mask with the GAME script and then operate on all of the images (using an image container). I will thing about the best method... but there are no very good solutions here because the artifact varies in both brightness and degree- simple thresholds will not work.
    -the Blockhead

    -the Blockhead
  • Reggie JonesReggie Jones
    Thanks a lot for taking the time to look into this.  Yes, I saw the data was vignetted and cloudy but it is consistent with data I've received from this telescope previously (AUS-2; Takahashi FSQ-106EX & FLI PL16803 camera). There is even a warning posted for this equipment that the images from it suffer from vignetting caused by the image circle of this telescope.  I've found that this problem can mostly be solved through smart use of DBE (your videos on this really helped me understand how to use the process correctly) but this was a totally different problem I'd not seen before.

    Again, thanks for looking into this.  I haven't gotten to your videos on the GAME script yet; I'm still trying to get a good handle on masks.

    Reggie
  • BlockHeadBlockHead
    I was correct. The issue occurs during their calibration and the treatment of floating point values.
    I suspect this is an issue during flat fielding due to the extreme vignetting they are trying to compensate for.

    I would suggest you connect with them regarding the frames. Those are not usable frames... they are mostly cloudy and not worth sweat equity  to make (a good) image from.
    -the Blockhead
  • Reggie JonesReggie Jones
    Ok, thanks.  When I've contacted them on issues before, Marco has been very responsive.

    Reggie
Sign In or Register to comment.