PI hangs on [A] button press in STF

[Yuriy Toropin]

Win 7 64bit ultimate SP-1
PixInsight 1.7.0 Starbuck

Last 2 days work with PI becomes almost impossible - consistently, attempt to apply automatic STF to just generated result of ImageIntegration leads to hang on of PI.
First, application of A STF to rejection maps works fine, attempt to apply to sum itself leads to PI freezeing and hanging.

Application should be killed and restarted....

How could I help in investigation of the issue?

Thanks,
   Yuriy

[Nocturnal]

I'd post the image that makes this happen. Can you save the II result, open it and then apply STF?

[Juan Conejero]

Hi Yuriy,

I have fully identified and fixed this bug, even before you reported it, in fact, as I have been working on the ImageIntegration tool a couple weeks ago

The bug happens upon the following conditions:

- Min/max pixel rejection algorithm selected.
- Most of the integrated frames have black borders at the same side of the image.

The explanation of this bug is as follows. The min/max rejection routine generates some pixels with non-finite values (NaN). The ScreenTransferFunction tool computes image statistics to calculate automatic stretch parameters, and the NaN or infinite values are not being properly handled by these routines, leading to a crash.

This bug is now fixed in the next version of the ImageIntegration module, which I'll publish during the next week. By the way, the new version provides excellent scalability and performance on multiprocessor/multicore machines.

[Yuriy Toropin]

Wow, Juan, thanks a lot for confirmation!!!
I was really.. disoriented when, after that terrible day (yes, I was using Mix/Max and 'cause it was "comet" session there were some frames with black regions on the same side) I've not observed that strange behavior working with different material...

Looking forward for that great "next version"!

***
Could you also please convert default "buffer size" in ImageIntegration to editable configurable parameter, in Global Preferences or somewhere else? It's boring to change those 16 Mb in each next session....

[Yuriy Toropin]

This bug is now fixed in the next version of the ImageIntegration module, which I'll publish during the next week. By the way, the new version provides excellent scalability and performance on multiprocessor/multicore machines.

Juan, could I kindly ask you to publish fix... soon... Garradd is flying, workaround like $T*1+0.0000001 doesn't work at all on the sum

[Juan Conejero]

Hi Yuriy. I've just uploaded a compiled ImageIntegration module for Windows x64:

http://pixinsight.org/tmp/II/ImageIntegration-pxm.dll.zip

Just uncompress it on your bin installation folder. This version fixes the min/max bug and all performance problems. Please read the tool tip information, especially for the new "Stack size" parameter.

PLEASE NOTE: Although this module works perfectly as far as I've tested it, this is not a final release version. The official version will be released as an update for all supported platforms during this week.

[Yuriy Toropin]

Thanks a lot, Juan!
It works, bug has gone and... it's REALLY fast!