›

Screen vs. Unscreen Stars

Ronald Clanton
in PixInsight
Adam,

In several of your videos, you discuss when to use "screen" vs. "unscreen" options in tools such as StarX, pixelmath, etc.  However, I can't find a segment where you explain what the difference is between the two.  Would you point me to a video that explains it?  Or, if there is not a specific video, would you consider doing a specific video on the subject?

Thanks!

Ron Clanton

Comments

  • Ronald Clanton
    Just to be clarify my question/request...

    I've watch the excellent video posted on 3/7/23 that explains the correct methods for using screened and unscreened methods... and their relationship to linear versus non-linear images.  So (as I said) I understand (robotically... lol) which one to use.  I guess my question is perhaps more around the terminology of the names "screen" versus "unscreen".  Why use this terminology?  Does it indicate something that is happening at a more fundamental level?

    I know this isn't likely to change my process, but I like to increase my understanding.

    Thanks,

    Ron Clanton
  • Mike Cranfield
    I believe the word "screen", used to describe the screen blending mode, comes from the analogy of projecting photographic slides onto a screen. The screen blend of two images is broadly similar to the effect of projecting the two images onto the same screen. I am not sure who first used the word "unscreen" but it is a more recent term intended to convey the inverse operation of screening - ie deducing one of the constituent images given the composite image and the other constituent image.

    I hope that helps a bit!

    CS, Mike
  • Ronald Clanton
    Thanks Mike!

    So (if I understand), that is why "screen" indicates additive/subtractive mathematical operations (e.g., projecting two images at equal weight)?  Then "unscreen" (or maybe more accurately "non-screen") is indicative of a multiply/divide operation?

    If so... I get it.  If not... perhaps I'm hopeless!

    Regardless, thanks for your response!

    Ron Clanton
  • BlockHeadBlockHead
    No... screen/unscreen is multiplicative in both directions. 
    I did actually explain how it works in a the video below. This deals with star reduction.. but it uses the screen math... so I explained it. Have you watched this?



    I think you will find it informative.
    -the Blockhead
  • Ronald Clanton
    edited November 2023
    Adam,

    Yes, I have watched that before and it does an excellent job of explaining star reduction using screen math.  I have also watched another video explaining when to check "unscreen stars" in StarX (linear vs. non-linear).  I guess what I'm missing (as stated in my original post) is what is the difference between screen and unscreen.  In other words, what does checking or non-checking the box in StarX actually do?

    BTW... this is a copy of the floating help for the box and where I got the idea of addition/subtraction for screening:

    Generates an "unscreened" stars-onlv image.


    Unscreening is the inverse operation to the commonly-used "screen" blending operation. If a starless image will receive additional processing after star removal, unscreened stars can be "re-screened" onto this image with much better results

    than simply adding the stars-ony image back to it.


    Unscreening is not recommended for linear (unstretched) images.


    With this setting unchecked, the stars-only image will be the simple arithmetic difference between the original and starless images, i.e., original - starless.


    Stars can be re-screened onto a starless image using PixelMath:

    ~((-starless)*(~stars)).


    where "starless" and "stars" are the names of the respective images.


    Thanks and Happy Thanksgiving!

    Ron Clanton
  • BlockHeadBlockHead
    edited November 2023
    Ron,

    You are mixing things here Ron.

    When SXT removes a star from one image... what is it doing?

    It takes the "star" (some values) and creates a second image of the star. This star that was removed has the TOTAL light (starlight plus background). What does it put in the place of the now starless image? It uses its AI logic to put *background values* there. 

    When I want to put these two images back together.. you can't just add them! You will have TWO backgrounds (AI background + Original background +star). So the stars only image is Original image minus Starless image. 

    So what does Russ do in SXT ? Well, when he creates the stars-only image it is the extracted star minus the AI background. In this way when you add the two images (stars_only+starless) you get back what you started with. 

    What I just described is what happens when you do not check the damn box. lol  It is what you highlighted in red.

    If you check the box and the "unscreening" formula is used- it is a multiplicative result and NOT the additive one I just described above. For small values (like in linear images) it does not really make a huge difference... but you do want to subtract that AI background before the multiplicative bit to  extract the star via the formula. This is what Bill Blanshan goes on and on about concerning screening and getting the total light (and color) correct. If the background is part of the multiplication... things get funky. 

    When you use the unscreen (muliplicative method) to extract stars... you must screen (inverse multiplicative) to put the images back together.

    Mike Cranfield can clean up my description above... but I think I am on the right track.

    -the Blockhead



  • Mike Cranfield
    Hi Ron

    So, when you run SXT on a starry image the first thing that happens is you generate a starless version of that image.

    Then, if you tick the "Generate Star Image" box, you are asking SXT to use the starry and starless images to generate an image with just the stars. This can be done in one of two ways - i.e. subtraction or "unscreening". Let's look at each.

    If you leave the "Unscreen Stars" box unchecked, SXT will simply subtract the starless image from the starry image. If you are dealing with linear data the background pixel values are very low so the issue of the AI background that Adam raises above is really not a problem - the background is so dark in comparison to the star pixel values it becomes immaterial.

    This is not the case if you are dealing with non-linear "stretched" data. So in that case you should check the "Unscreen Stars" box.

    If you check the "Unscreen Stars" box, SXT will start exactly the same way as before - ie it will subtract the starless image from the starry image. But it will then adjust this value to reflect the brightness of the background. In principle it does this by dividing by (1 - starless). (Although the actual formula used by SXT is a bit different from this, the effect is broadly similar.)

    OK, so you now have separate starless and stars images. At some point you will probably want to recombine them.

    Replacement of the extracted stars will generally be done at the non-linear stage. Therefore, as Adam says, you will "screen" the two images together using the PixelMath expression: combine(starless, stars, op_screen()). You will do this even if the stars were originally extracted by subtraction at the linear stage and subsequently stretched to non-linear.

    The only time you would recombine using addition is if you are recombining starless and stars images at the linear stage. In this case the background is likely to be negligible so straight addition will likely work fine.

    I hope this clarifies things and, between this and Adam's response, we have covered your question.

    CS, Mike



  • BlockHeadBlockHead
    Mike... just to be complete- Russ is doing what I am saying when he outputs his stars-only image.


  • Ronald Clanton
    edited November 2023
    Thanks gentlemen!

    I had to read your responses a few times to really understand the processes... but I think I get it now.  Hopefully, this discussion also provides some value to other members.

    Happy Thanksgiving!

    Ron Clanton
  • BlockHeadBlockHead
    So I talked with Mike... I stand corrected. Please refer to Mike's description of things as a most accurate reference.
    Thanks!
    -the Blockhead
Sign In or Register to comment.