›

Cosmetic correction vs pixel rejection with drizzle

Woodbine
in PixInsight
I thought I understood cc until I recently imaged my first star cluster (Hercules). I have learned the value of dither and drizzle to achieve a remarkable improvement in image quality with my otherwise undersampled OSC data on nebulae and galaxies. I had also been using cc by default to remove ~7sigma hot pixels.

However, my first integration attempt on the Hercules cluster resulted in horrible donut like bright stars all through the cluster. After some Googling, I simply dropped the CC step in WBPP and the drizzled stars became fine. The original “donut” stars were bright, but not saturated and I didn’t see the problem on brighter stars outside the cluster.

As this came up in as Google search, it seems to be a known effect, but I don’t understand what was happening and I’d really appreciate it if one of the PI experts on this forum could help explain.

Many thanks - Paul




Comments

  • BlockHeadBlockHead
    Try turning off the high range rejection in ImageIntegration. 
    This will likely solve the issue. 

    You can test by reintegrating your data (you do not need to run the whole WBPP thing again). 
    You need to use ImageINtegration without high range rejection (everything the same). This will update your drizzle files. Then you will use drizzle integration again. 

    The process container in the logs folder will be of great assistance. Please watch my WBPP video to learn how to take advantage of it.

    -the Blockhead
  • Woodbine
    Thanks Adam,

    I am rewatching the entire WBPP lessons again as I begin to appreciate the importance of really understanding this step in the process.

    I’m now trying to get my head around the weighting process and whether/when adding a lower quality subframe improves or degrades the final result (ie: how do I decide quantitatively when to reject a sub/ frame).

    -Paul
  • BlockHeadBlockHead
    In an ideal world you should be able to leave any weighted image in the stack and it will contribute for what it is worth and improve the SNR. However, there is a correlation between bad frames (poor quality/weights) and other kinds of issues (gradients, seeing...etc etc) in such a way that establishing a base weighting threshold usually does the job. A conservative value is not to include anything that has less weight than 50% (0.5). This advice applies if the images are all the same exposure time. 

    However, when I originally made this suggestion... people very quickly discovered how crappy their data is! They found they may have a small percentage of good frames with a bulk of poor quality ones. This lead to a wholesale rejection of many frames. 

    So... you need to find a weight threshold that is basically the minimum quality you would accept. In this way you do not need to rigorously examine every frame (if you have of good idea of the data quality)- the minimum weight threshold will do the work. This takes the decision out of your hands. 

    I have found that the weighting schemes do exactly as you would expect and I have been very please with using a conservative weighting threshold.

    -the Blockhead
Sign In or Register to comment.