Page 1 of 1

Xisf Format

Posted: Wed Jul 20, 2016 3:37 pm
by Starry Eyes
Any plans to include the xisf format to ST?. For what ever reason I've had to give up DSS for my stacking and am using that other really complicated PI for linear processing. Of course I save to 32 bit fits for ST. Seems to come out alright. Is there a practical difference between the integer and floating point versions of fits?
PJM

Image

Re: Xisf Format

Posted: Tue Jul 26, 2016 11:07 pm
by admin
Apologies for the late reply - I've been traveling a whole lot the past couple of weeks...

It appears XISF is a PixInsight-specific format (though open and royalty/patent free), however PI will keep supporting FITS (as it should).

As for the practical difference between Integer and Floating Point FITS, there are some small, subtle differences, which have to do with the way floating point numbers can encode data that aren't really numbers (such as Not-a-Number or +/- infinity).StarTools includes some extra routines to interpret such non-numerical values when importing floating point data, so that such missing samples get converted into 'best guesses' in order not to 'upset' the many algorithms in StarTools that require an artifact-free image as input.

Given that StarTools' internal engine uses 64-bit integer arithmetic, it is also theoretically possible (if you're feeling particularly evil) to specify a dynamic range that cannot be sufficiently encoded using 64-bit integers, causing banding.

Long story short, you will probably notice very little between FP or integer encoding, but Integer-saved FITS will make StarTools' job a lot easier (and loading times possibly shorter).

Re: Xisf Format

Posted: Sun Aug 07, 2016 2:32 am
by Starry Eyes
Very good, Sir!