Pixel Rejection

[Corries]

Can anyone please help?

I have a question about the Image Integration process.  When the Image Integration process has completed and I examine the rejection_high.fit file there seems, to me, to be a rather large number of pixels that are being rejected (see link below).  Am I correct in my thinking or is this about normal for pixel rejection?

Many thanks, Terry.

https://drive.google.com/open?id=0B7H6w19vQ2UPbFlWb0hodVk1dnc

[Geoff]

Have you looked at the Pixel rejection (2) panel?  You can tweak the parameters there to get more or less rejection.
Have a look at the documentation for ImageIntegration for more detail on how to tweak rejection parameters. (Imageintegration is one of the documented processes and as is always the case where there is documentaion, it is excellent)
Geoff

[Corries]

Hi Geoff,

Thanks for the reply.  I’m afraid I didn’t explain my problem very well.  I was aware that you can adjust the rejection parameters.  But, being new to PixInsight I don’t know what is normal.  Is the amount of rejected pixels, as shown in the rejection_high.fit file, normal or excessive?  Shall I adjust the parameters to reduce them or not?  Can you please give me any guidance here?

Many thanks, Terry.

[Geoff]

Hi Terry
Have a look at this webpage by Jordi Gallego
http://www.astrosurf.com/jordigallego/articles.html
About halfway down the page there is an one titled 'Image Integration Techniques....."
It is a bit dated in terms of where PI is now, but the underlying ideas are what you need.
Geoff

[aworonow]

One thing in Jordi Gallego's page that is suspect is the suggestion that including an image with noticeable defects (tracking failure, in his case) can improve results of the integrated image.  True, it may increase the estimated-noise-reduction (ENR) metric, but that metric is base on the areas free of signal (at least as free as can be). Those areas are just noise, so ENR improves by providing more samples of the background. But signal may be distorted and contrast within signal-containing regions may be reduced, as may the sharpness of the signal-background interface, when flawed images are included. ENR does not speak to those issues. At least that 's my read of the ImageIntegration documentation.

I did  take a look at a set of data that had about 20 'quality' images and another 10 with poor tracking, poor focus (light clouds?), etc. The ImageIntegration of the ensemble of all had a higher ENR than did quality images alone (after optimizing both), but the ensemble had obviously less definition of fine structure than did the 20 best.

Alex

[Corries]

Thanks, everyone.  This is a great help.

Terry.