›

A few questions about Pixinsight Fundamentals “Usage of Cosmetic Correction”

Michael Lewis
in PixInsight

First a big thank you to Adam for producing these videos. I’m the kind of person that is more successful knowing why things work and these videos do a great job of explaining that in PixInsight.

I’ve been watching the videos and taking notes to add to my workflow for processing LRGB data.  After watching only 9 of the videos my workflow has had numerous changes. Watching the Cosmetic Correction videos has filled in a lot of blanks, but I’m left with a few questions.

While discussing making a Hot Pixel Defect Map Adam opens and clones an image called tempdefectmap and uses it for the process. Is this a single dark sub or is this a Master Dark? Does it make a difference?

Later in the video Adam describes adjusting the Binarization Threshold to reduce the number of “Hot Pixels”. Any guidance on how to determine the sweet spot? In the second iteration Adam applies to the image, at 16:33, Adam says “That’s probably better”. Why is that better? I’m very sensitive to protecting any real data and want to avoid calibrating it out.

In the Defect Map process at 19:25 Adam uses Median for the Operation. The guidance offered by the flyout information in PI is a bit vague, it tells how the Operation will be applied to the process, but not where each Operation is appropriate. I’d really like to understand this better. Similarly the Structure option Adam chooses at 19:50 is Square, and he comments that “it’s appropriate in this particular case”. Why does Square apply to this case and what conditions would make other options more appropriate?

Again in the Defect List section Adam Suggests opening “a bias”. Is this a single bias frame, a Master Bias, or even a Super Bias? I think it would be a Master Bias, but not a Super Bias as the Super Bias is smoothed.

Thank you in advance for any advice offered. This forum is a great supplement to the video series.

Comments

  • BlockHeadBlockHead
    Hi Michael,

    A funny thing- I did this video a while ago so... I I need to re-watch it to see what in the world I said!

    1. With regards to making a Hot Pixel Map from a dark frame- I do not believe it matters too much whether it is a single dark or a master. I apparently was using a master (if you notice file name of "tempdefectmap", it has the same convention as masters created with BPP. All that is important, in the method I demonstrated, is finding a threshold that pretty much singles out the hot pixels. 

    2. This method of producing a hot pixel map from a dark frame isn't terribly rigorous or quantitative. My preference is to use the "Auto" method with statistical thresholds (sigma units based on neighboring values). Keep in mind, for all methods, you are replacing hot pixels you identify with a value based on a neighboring pixel. If you have undersampled stars... this *could* be problematic for this methods...(and rejection with proper dithering does the trick). Otherwise, critically or oversampled data means that these individual pixels will not "calibrate out" important data-  Let us say you were measuring the brightness of a star and you used a photometry aperture to make a measurement. This would combine/average all of the pixels of the star. If you how an 'adjusted' hot pixel in there...it would receive the value of its neighbors (also star values since over sampled, and no significant effect would be measured concerning the average brightness of the star. So the hot pixel filter is simply for cosmetic purposes. Furthermore- you really should be able to do everything with combination of dithering and rejection. However, I do like to hit the hot pixels early on... so I do employ CC.

    3. Regarding the method used for the defect map pixel substitution/calculation. My recent sections on statistics answer this question at some level. These answer is that median is a "robust" measure of the central tendency of a set of fluctuating values. Even in the presence of outliers... you will still get a good answer (basically the idea... please review the Image Integration Primer series under "Processes" that I recently uploaded.

    4. Square is appropriate because pixels or lines of pixels take that shape. This is analogous to the demonstration in Morphological Transformation. You match the shape of the operator (the tool) with the shape of the thing you are operating on.

    5. You are correct about the Bias.

    - the Blockhead


  • Michael Lewis
    Thank you Blockhead for the clarification. I've added the gist of these answers to my notes.
Sign In or Register to comment.