Drizzling introducing artifacts?

[cmarcus]

I'm having problems with drizzling that is introducing artifacts see attached screen short - presumably these are dithered hot pixels?
The image on the left is the normal intregrated (winsorized sigma clipping) integrated image. On the right is the drizzled image taking the drizzle files from the integration as per normal. As you can see these artefacts have appeared...
any idea what is going on?
TIA for any input

[pfile]

do you have "enable pixel rejection" turned on in DrizzleIntegration?

are you running the latest PI with all updates?

i seem to recall a bug that caused a problem like this with DrizzleIntegration, but it was a long, long time ago.

rob

[cmarcus]

Many thanks for your reply.

I have enable pixel rejection turned in DrizzleIntegration.
But I have not installed the latest updates... doing that now and see what happens.

Regards

[cmarcus]

It still happens even with the latest updates...
 

[pfile]

when you last ran ImageIntegraton on the non-drizzled files, did you have pixel rejection turned on? if you did a test run of II without rejection then i think the rejection tickbox in DI does nothing.

here is the thread i was thinking about. are you sure you are on the latest version of the PI core application? it can not be update with the update mechanism; you have to download a new installer.

https://pixinsight.com/forum/index.php?topic=10398.15

rob

[cmarcus]

There is also this thread that I didn't get round to pursuing even tho' I started it :

https://pixinsight.com/forum/index.php?topic=10228.msg64688#msg64688

I did have pixel rejection turned on during II.

I'll make some further investigations today. My position is slightly complicated because I am currently on my laptop as the desktop machine is at the "doctor's". The last time I drizzled on that machine it did actually work without this problem, and it may be a PI update issue on this laptop. (For which I have indeed downloaded the new installer).

Many thanks once again for your input.
Clifford

[cmarcus]

I have tried a couple of things and am still getting the effect.
My version of PI is 0.1.08.04.1195 Ripley (x64) and all updates have been installed.
The only thing I can think of is that on doing these tests I was using files that had been calibrated under a previous version of PI. I'm presuming that the bug is not at or before that stage?
Next step could be start everything ab novo with the fully updated version?
Many thanks
Clifford

[pfile]

i happened to process some images that i had laying around that had been drizzled 3x... peeping in really close i see a similar problem with what look like hot pixels. i wonder if there is something about the drizzle process that emphasizes such hot pixels somehow? it might make sense to try doing CosmeticCorrection on your calibrated lights to simply remove any hot pixel artifacts before going thru the drizzle flow.

rob

[RickS]

I use Drizzle a lot and don't see problems like this... but I have to use CosmeticCorrection to repair bad columns so I normally do a default autodetect removal of hot & cold pixels at the same time.

Cheers,
Rick.

[Juan Conejero]

Hi Clifford,

I cannot reproduce this problem with DrizzleIntegration. To help you, I need your calibrated (not aligned) data set.

peeping in really close i see a similar problem with what look like hot pixels. i wonder if there is something about the drizzle process that emphasizes such hot pixels somehow?

Not at all. However, there was a bug in DI where the tool was making some mistakes interpreting pixel rejection data stored in .drz files. This bug was fixed in October 2016 with an update to the DI tool. So if you have images that had been processed with older versions, you should re-process them. The current version of the DrizzleIntegration tool should give the same results as ImageIntegration rejection wise. The new release 1.8.5 of PixInsight includes a highly improved version of this tool, along with a new XML-based drizzle data format (.xdrz files) that is much more efficient and flexible.

[pfile]

it is certainly possible that the files were processed with the old DI, i just found them laying around on my disk yesterday and they are pretty old.

rob

[cmarcus]

Hi Juan,
Just noticed your answer. I am still getting the effect with other data which I have processed ab novo with 01.08.04.1195. (see attached screen shot of drizzle integration on the left and regular integration on the right) How can I get a data set to you?
Interestingly enough, I have not always had this problem, it seems to have come in the last six months or so, have been a big fan of drizzling for some time...
Regards
Clifford

[Juan Conejero]

Hi Clifford,

As noted before, I cannot reproduce this issue with any data set. To help you, and in case we have a bug, to fix it, I need a set of calibrated but not aligned frames, where you are having this problem. Please upload the data to dropbox, Google drive, or a similar service. I don't need a lot of data; just a moderate sample where you can get this problem consistently. Thank you!

[cmarcus]

Many thanks Juan,
I have PMed you a d/box link with a data set
Clifford

[Juan Conejero]

Hi Clifford,

Thank you for the data. There is no problem at all with DrizzleIntegration working with these files:


(Click for a full-size version)

This is with the current version. As you can see, there are no problems with pixel rejection. I have made the same test with the next version 1.8.5 and the new .xdrz format with similar (better, faster) results. Are you sure you have the Enable pixel rejection option checked in DrizzleIntegration?