›

Poor Results After DBE Applied to Mure'd OSC Integration (PI V. 1.8.9 - 1)

Michael Lehv
in PixInsight
Hi Adam, 

I'm seeing horrible results with DBE following MURE. Here's an example of manually preprocessing 46 X 300" OSC subs of M65/66 using the following workflow:  Calibration > Debayer with Separate RGB Channels > Star Alignment with a green channel reference sub applied to all the subs of the three channels > Separate integrations of the registered subs of each channel > MURE applied to each of the three mono integrated images with appropriate Variances > Channel Combination of the three MURE'd mono integrated images > DBE applied to the combined image.

MURE works fine with expected de-speckling compared to an image without MURE. DBE is reasonable on the image without MURE. However, DBE used on the MURE'd image results in a dreadful pink mess with overwhelming blotches. 

I've uploaded four Google Drive .xisf images and two Process Icons. Image 1 has the problem - the other images and the Process Icons below provide supplemental information you might need.


I've spent an inordinate amount of time trying just about every possible combination of Image Integration and DBE options with no real improvement in the MURE-DBE result. I didn't use Local Normalization because the raw subs were all very similar and also to keep things simple. 

The problem I'm seeing is not unique to these images although usually instead of pink, green is the main color. Where have I have gone astray? 

Would really appreciate your help on this. Thanks.

Michael

Michael Lehv
msl@lehv.com

Supplemental Data 






Comments

  • BlockHeadBlockHead
    HI Michael,

    Sometime ago I reached the same conclusion you did. When using some CMOS sensor there is a correlation between pixels that MUREDenoise does not accommodate. These days tools such as Russ Croman's NoiseXTerminator are a good alternative. 

    On a teaching note- I would say MUREDenoise is NOT working as you would like on the image before DBE. It is just that DBE allows you to better see the non-optimal result. Again, it was not working to begin with. 

    -the Blockhead
  • Michael Lehv
    Thanks Adam,

    I also suspected that DBE was merely exposing problems that MURE had created. So, you are saying that that the real issue is that MURE doesn't work properly on integrated mono subs derived from Debayer's Separate RGB Channels option? 

    So I guess that the conclusion is that MURE Denoise cannot be used on OSC images themselves nor on their mono channels. The MURE script tooltips seem to suggest this but they're not entirely clear.    

    I will give NoiseXTerminator a try.

    Thanks again.

    Michael 

    Michael Lehv
    msl@lehv.com 
  • BlockHeadBlockHead
    I did not say "cannot" ... I said that I found for some CMOS sensors the pixel-to-pixel pattern (noise?) is correlated and this is not something that MUREDenoise can accommodate. I do not think MUREDenoise is creating the issue- the correlation of pixel values is a hardware issue... not an algorithm. Correlations are not noise...that is the point. 

    -the Blockhead
  • Michael Lehv
    Hi Adam,

    Now I understand what you are saying. So, for my current OSCs (QHY128Pro and QHY410C with otherwise identical FSI and BSI sensors) MUREDenoise isn't a workable option. However, I took your advice and gave NoiseXTerminator a try. The result seems dramatically better. 


    On this linear combination of mono images I applied NoiseXTerminator after DBE and PCC. To my somewhat naive eye, using the default settings seemed to provide excellent, if not oversmoothed, results.

    I'd appreciate hearing what you think?

    Thanks.

    Michael
Sign In or Register to comment.