›

LinearFit Producing Weird Results on Some Images

Zak’s Astrophotography
in PixInsight
Hi Adam,

I have been having issues at times with LinearFit behaving "oddly". I have tried following your NB FastTrack workflow videos, but it's not producing the results that I am expecting.

For example, when I try to use LF on HA data using OIII as my reference, it's not producing the expected results. From my observations, it only seems to be a problem on objects that have stronger OIII signal than HA where LF seems to "invert" the images. I've also noticed it may introduce more green/teal colour noise depending on the image, but I suspect this is because I am using the OIII as my reference. There may also be some occasions where it over "saturates" the colours even with an unlinked STF stretch. I am not sure what the cause(s) for this are.

Would there be some other ways I can both reduce this colour noise as I've tried using your technique to reduce the colour noise in my images, but even then it doesn't seem to work properly, and also some techniques to try and tame the colours in my HOO images from ChannelCombination?

Attached is a link to some screenshots of the problem I am experiencing. I know that the images don't look "great" due to insufficent exposure times, but I am hoping to overcome this with more integration time. Please let me know if you want links for the master files of these images.

Zak

Comments

  • BlockHeadBlockHead
    Zak,

    I fear those screenshots do not help. What are the values in the images before you apply linear fit?

    1. Linear Fit of data is not a requirement. You do not have to do it at all...and you can choose whatever scaling factor you want. 
    2. NB Normalization (without Linear Fit) will give you approximately the same results.
    3. Expect with respect to #2... Linear Fit could care less about what kind of data you are giving it. If your data is dominated by the SKY and not the nebula you will definitely get wacky results. Watch my videos in Linear Fit to understand how to work with it. Basically you select an area of some bright nebulosity that is the same in both images. Make previews to do this. Then you perform the linear fit. You will get the scaling factors you are looking for.
    4. But let me repeat... Linear FIt is not a requirement...it is just helpful. If your data is too noisy after Linear Fit... it means you are scaling up (multiplying) by a very large factor. Linear Fit does not "make noise". It will just make the noise "bright" by scaling everything up... if you number is large this is what happens. It does help to have data that is not that noisy in general.

    So other ways:
    1. Do not do Linear Fit at all. Just stretch the data as you like.
    2. Use only NB Normalization (which will be similar)

    I do not understand the idea of "oversaturating" colors. One color may dominate...but this depends on your data and object. 

    -the Blockhead
  • Zak’s Astrophotography
    Thanks Adam. I shared those screenshots in case it would have helped. All good.

    In the use of NarrowbandNormalization instead of LinearFit, would I do it after DBE, and do I just choose the settings that I think are good?

    I will ensure to watch the videos on LinearFit so I have a better understanding of how it works.

    One thing that I need to improve is my integration times on projects. I tend to have an urge to switch targets during my imaging sessions instead of focusing on one object over several nights.
Sign In or Register to comment.