›

Mure Denoise before integration

Juan Bernardo Torre Valle
in PixInsight
Hi Adam.-
I have read that some authors recommend using MureDenoise to each image before  Imageintegration and after normalizing them with NormalizedScaleGradient, instead of doing it with the integrated Master which is what you always do.
To do this, logically to speed up the process, processcontainer and imagecontainer are used to do batch processing.
I have verified that it really does a better noise reduction but I do not know if it may be affecting other parameters and it is preferable to do it after integrated.

Your opinion on this would be very interesting.

Thank you

Comments

  • BlockHeadBlockHead
    Hi Juan, 

    I have not done this experiment...so I do not know. The paper that I cited clearly states the algorithm is intended to work on the integrated image. This is why I include this information, the paper author is the ultimate authority. 

    For an answer, and probably better one, ask this question on the official PixInsight forum and Mike Schuster will respond.

    If you read the documentation for this script it says: "Warning: Do not combine denoised images." The resulting SNR of the combined images will be less. His algorithm is adapted for a single integrated image. 

    See the screenshot

    -the Blockhead
    Capture.JPG
    1571 x 1038 - 246K
  • Juan Bernardo Torre Valle
    edited November 2021
    ok,, thank you very much. 

    If the author says so clearly, then nothing to discuss
  • Roger Smith
    Interesting in the same screenshot warning, he also says "and be registered by projective transformation with no distortion correction." 
    I normally use (checked off) distortion correction during star alignment. But not sure if anything was really corrected. I like my Mure Denoise results, but maybe it could be better. 
    I am going to ask on PI forum.
        Roger
  • BlockHeadBlockHead
    If the distortions are small, likely it doesn't matter much. But the point is that distortion correction is another "warping" term that results in an interpolation of values that MD does not account for. It only models the uniform interpolation of the Alignment routine (defaul Lanczos).

    -the Blockhead
  • BlockHeadBlockHead
    oh...and I do think LPS will certainly also wreak some havoc with MD. It might still work...but probably not be optimal... depands on how many columns are affect I suspect.
  • BlockHeadBlockHead
    I guess I am wrong about LPS! I am very surprised!
  • Roger Smith
    Adam wrote: "If the distortions are small, likely it doesn't matter much"

    Juan on PI forum said distortion correction will not affect MD.
    Same for LPS.
Sign In or Register to comment.