›

Processing Narrowband and black spots...

Roger Smith
in PixInsight
Hi Adam,
I have been processing my duo narrowband images and have (had) many repetitive black spots on the integrated image (red channel mostly). After detail study of subs and dark frames I found my matching darks overcorrected the hot pixels. I could see the repetitive black pixel pattern (in the integrated image) from my dithering. I went back an re-ran WBPP2.3.2 with 300 added pedestal. A little better after integration, but still not good. It takes a lot of processing time to rerun WBPP, and then NSG with different added pedestal amounts.

I was wondering why image integration did not reject the 0.0000 value pixels. I have 46 images and used GESD.

To give myself more control, I used Windsorized Sigma Clipping and set the Low Range clipping up to be below slightly below those pixels that were above 0.0000. I had between 0.5% and 1.4% low rejects. Average was 1%. It is easier (compared to WBPP/NSG) to adjust Image Integration rejection parameters and try again.

Visually, even at 3:1 zoom, the GESD and Windsorized integrated images look exactly the same except for no black pixels in the Windsorized. For me, I have never really liked the lack of intuitive control GESD gives. Windsorized is very intuitive and my go to rejection tool. I like rejecting about 2% high, and 1% 

So my question is if adding pedestal in WBPP is equivalent to setting a higher 'clip low range' value in Windsorized Sigma Clipping in image integration?

Thanks,
    Roger

Comments

  • BlockHeadBlockHead
    First- you really should figure out from an acquisition standpoint why there is an oversubtraction of hot pixels. Typically this means you are using aged darks or the temperature is changing. 

    GESD does not handle high and low values symmetrically. See John Pane's post on the forum. The relaxation parameter will affect the rejection of high values more than low- and with GESD it is hard to get the big outliers like zeros rejected. 

    Sigma Rejection does actually do more rejection- which allows you to get rejected what you want.
    But I would encourage you to do the following experiment with the "normal value" rejection (not the zeros)-
    You will find that GESD rejects the smallest number of pixels that contains 100% of the outliers you expect to reject such as satellite trails, cosmic rays...etc. Sigma Rejection will reject many many more pixels in addition to the ones you *need* rejected to reach the same level of efficacy. In this sense GESD is much more efficient/effective... with exception to extreme outliers on the low end.  

    So... low end outliers (especially zeros) are really indicative of noisy data and should be looked at from an acquisition/equipment perspective. We really expect the noise distribution below the central values of the distributions to be dominated by the read noise and as along as the readnoise doesn't give zeros (with a pedestal if necessary)- GESD works well. So the efficiency comes with the price of no zeros. 

    -the Blockhead
  • Roger Smith
    Hi Adam,
    The flats are about 1 month different than the subs.
    Yes, ESD always has given me less than .5% rejects. Where Windsorized rejection is normally 2%+/- total pixels rejected.
    I know my image data is noisy, and background level higher (.10) than most Ha subs. 
    I checked several, different levels, hot pixels. In all cases the matching location master dark hot pixel was less ADU than the raw sub at that location. Only the fully saturated pixels has the master dark and raw sub equal (.9998). When subtracted it is zero.

    During registration, the black pixels defects move around due to dithering. During image integration with default setting (1.5 in ESD low relaxation) not all the black defects are corrected. Lowering ESD low relaxation to the minimum 1.0 value results in almost all black pixels being rejected. A few spots remain.
    Rejection percentage is .25% total.

    So I will stick with ESD for these 46 subs.

    This has been a good study for me to understand where the defects originated, and how to improve the rejection during integration.

    Thanks,
       Roger
Sign In or Register to comment.