›

Oiii Drizzle looks like it has less bit depth

Ian Hudson
in PixInsight
I have a lot of dithered full frame NB images of Heart Nebula.

HA and Sii drizzle integrations produced great images.

Oiii produced acceptable integrations but drizzle integrations which appear blocky / less bit depth.

To improve I have:

Blinked all images and discarded the ones which are visibly substandard - no improvement
Ran images through subframe selector and discarded those with an SNR of less than 1.5 - no improvement

Siii and HA were taken on similar nights, same exposure settings and same pixinsight settings, so I don't understand why I am getting such different results.

Links to samples below 

Comments

  • BlockHeadBlockHead
    This looks like a pedestal, offset, oversubtraction issue to me.
    I am not even convinced either of these integrated images is correct.

    The first thing you need to do is look at a single calibrated frame and check for Zeros.
    I would put money on the fact there are some... like a gazillion of them in your data.
    Then you need to figure out why. 

    Use the method I show in this video:
    (around 13 minutes)

    The two most common reasons (there are around 10)

    1. Did you use a pedestal for NB images? This is automated in WBPP... but you can defeat it I guess.
    2. Used a dark of a different gain than the data was taken at.

    A third one is that you used different software to acquire the lights and calibration data.
    So many ways.

    Look at the calibrated frame... demonstrate there is nothing wrong with one.

    -the Blockhead
  • Ian Hudson
    Thank you, Adam.

    I agree even the non-drizzle image is not great.

    I used a manual pedestal of 300 for Oiii (same as HA and Sii)
    Same master dark as HA and Sii
    Same software to acquire lights, flats and darks.  (NINA) and same across HA and Sii

    I will check a single calibrated frame

    Thanks once again
  • BlockHeadBlockHead
    Yes... and 300 may or may not be enough.
    (The Auto setting in PI does just fine.)

    -the Blockhead
  • Ian Hudson
    Adam

    I have Looked at a sample of calibrated frames none of them have zero pixels.

    I have uploaded some calibrated frames (Oiii at 3 different exposures with an HA for comparison) all with the pixelmath calculation showing no red pixels.

    Following your video I experimented with pedestal and settled on 300 as a figure which with my set up would never give zero pixels.  It's too high but given it's a 16 bit camera I figured + or - a few hundred wouldn't make much difference providing none were zero.    
  • Ian Hudson
    PS

    ZWO ASI6200 MM Pro - 16 bit 9576x6388 pixels
    Takahashi FSQ-106EDX4. 530mm F5
  • Ian Hudson
    very light polluted skies (surburban UK, 7 miles from a major airport)
  • BlockHeadBlockHead
    Hmmm... 
    OK. Well... that leads to some other ideas that are more nuanced. 
    I am thinking about rejection now. 

    So... how many exposures? What is the exposure length?
    Did you use WBPP?  (Did you set it to Winsorized Sigma Clipping instead of Linear)
    Did you look at the number of pixels rejected if you used the AutoCC method?
    (I don't think this explains your issue...but I was recently bit by an issue with this.)

    -adam

  • Ian Hudson
    368 exposures
    3 different lengths 25s, 70s and 360s
    Yes I used WBPP but only for calibration
    I didn't check rejected pixels with AutoCC......  I've got lazy as WBPP 'always' works !!!!
    ........... ah hah (I think)
    I originally used Normalised Scale Gradient but some of the frames were absolute rubbish and the results were not good.  To save time in tests I didn't use NSG but went straight to image integration without checking options......no pixel rejection!!!!

    Tomorrow (its 9pm here now) I'm going to take a few steps back, pause and move forward !!! 

    Thank you once again Adam, I really appreciate your help.
  • BlockHeadBlockHead
    Ooooh.
    There is something in the mix. Those 25s exposures.
    In addition to proper normalization and rejection you need to make certain the images are getting the proper weights. (Personally I am not convinced the 25 second exposures will add very much. NB exposures of anything at 25s is just noise I would think.)

    So yeah, now we are tracking in understanding with your results.

    -the Blockhead
  • Ian Hudson
    Reran NSG and left transmission / weight at default values.  Moved rejected images....... and ..........  All of the 25s and most of the 70s images were rejected. Let NSG provide values for integration and drizzle integration.

    Final drizzle integration looks better but still not good.

    I have updated the files in the link above.
    1.  made them starless with stareXteminator
    2.  cropped them down to reduce the size

    The images are
    1. An HA integration for comparison - showing lovely detail
    2. The Oiii drizzle integration, 30 frames each of 360s
    3.  The best calibrated Oiii frame

    Could it be that even at 360s there is just not enough Oiii signal ?


  • BlockHeadBlockHead
    This looks correct to me.
    Yes, not all object emit strongly. You likely need more... much much more of the 360sec exposures.

    -teh BLockhead
Sign In or Register to comment.