ImageIntegration problem?

[cfranks]

Hi,

I'm having a problem with the current version (01.07.05.0779 Starbuck (eng x86_64)) installed today.  I took 4 images each of Ha, O-III amd S-II from last night, succesfully calibrated with ImageCalibration and registered with ImageRegistration and output 32 bit results.  ImageIntegration fails when I try and integrate each group of 4.  Since there are only 4 images, I used Percentile Clipping rejection with No Normalization.  On a Windows XP x64 system with 8Gb RAM, PixInsight crashes back to Windows when I 'Global Apply'.  On a Windows 7 x64 Laptop with 4Gb of RAM, PixInsight reads in the 4 images then 'grays out' and freezes until I eventually force it closed.  Previously integrated large groups of images still correctly integrate when I use my usual Rejection algorithm of Linear Fit Clipping.  This is such a straight forward process that I cannot think it is something I am doing/not doing but I'm willing to be told!      

Charles

[Juan Conejero]

Hi Charles,

Bug confirmed. This is a Windows-specific bug. It only happens under the following conditions:

- Percentile clipping rejection selected.
- No rejection normalization selected.

It happens irrespective of the number of integrated images. It does not happen on the FreeBSD, Mac OS X and Linux versions of PixInsight (it is probably a bug in the MS Visual C++ compiler).

However, why are you selecting no rejection normalization? Scale + zero offset normalization should _always_ be used unless you are integrating images with heavy and strongly varying irregular illumination problems; in such case flux equalization normalization can be preferable (but this must be checked carefully in each case). Disabling rejection normalization only makes sense to integrate bias and dark frames.

Thanks for reporting. We'll fix this bug in the next version of the ImageIntegration module.