Drizzle integration leaves what looks like hot pixels

[mhultstrom]

Hi,

I have a problem with artifacts when using drizzle integration. Here is a standard processing of M20, 50 x 300s dithered exposures with normal integration using median stacking with sigma rejection. It looks fine.



However in the drizzle integrated version I get these spots that look like hot pixels, for example they seem to follow the dither pattern. The originals are calibrated using both flats and darks in a temperature controlled camera. In addition, what ever it is, it is easily rejected in the standard integration. I have tried different rejection algorithms and integration methods, but they are very persistent when using drizzle integration.



Thanks for any help,

Michael

[Juan Conejero]

Hi Michael,

Looks like your .drz files don't include valid pixel rejection data, or pixel rejection parameters have not been properly adapted to the data set with ImageIntegration.

median stacking with sigma rejection

This can be the origin of the problem. Maybe it is necessary in other applications, but in PixInsight, with very few and very special exceptions, you should always avoid median combination. You should use average combination and fine tune pixel rejection. See the documentation for ImageIntegration (now outdated, but still useful) for detailed information.

On the other hand, if you use median combination your .drz files won't include, in general, the necessary pixel rejection data. Bear in mind that median combination performs an implicit rejection of outlier pixels that cannot be emulated by the drizzle integration process.

[mhultstrom]

Thanks for your reply. I did try different rejection methods as well as different integration methods, but I will try some more..

I do get a warning for "Illegal pedestal keywords" can that be related?

[Juan Conejero]

Indeed it can. Where have you calibrated the data? Importing calibrated data from other applications does not work. You must calibrate the entire data set in PixInsight, including individual bias, dark, flat and light frames.

[mhultstrom]

Calibration was automatially done in Maxim DL by the iTelescope system.  And, why is that? All the data is in the data, and outliers are still outliers.
Is this something specific for Drizzle, or are there other things that also don't work unless I do everything in PixInsight?

Cheers,

Michael

[Juan Conejero]

If the calibrated frames are in 32-bit floating point format (and if they are not, then they are incorrect anyway), then they have been stored with numeric ranges different from the ranges used by PixInsight and our preprocessing tools. In addition, there may be incompatible metadata (such as PEDESTAL keywords with wrong values) and other problems. You should request the original raw data, which IMO is what you pay for when you rent an observatory.

[mhultstrom]

Fair point.

[jkmorse]

Michael,

What rejection algorithms did you try?  I used to see similar things with certain rejection methods but it would go away when I switch methods.  With that many subs, you really should only be looking at Linear Fit and Winsorized rejection options.  Try both of those and see if either helps.

Jim