Tuesday, April 8, 2014

pyAstroStack 0.3.0 - Lots of speedups and OpenCL no longer required

I released 0.2.1 just about a week ago and now I'm about to push 0.2.5 out. First the updates were indeed minor, but they grew and became quite important considering the whole program. After finishing 0.2.5 and hunting down its bugs I decided this version should be 0.3.0 instead.

Installation still requires Python-pip on Ubuntu 13.10, but I'm already targeting this on soon to be released 14.04. There might even be a deb-package or repository on Launchpad if I figure them out.

Here's a summary of changes:

  • Bilinear and Variable Number of Gradients (VNG ) debayering algorithms. This makes PyOpenCL not required. With OpenCL the methods are a bit faster but the difference is about 10-20%.
  • PyAstroStack will work on Ubuntu 14.04 without too much hassle. There are a lot of packages to install but everything seems to be on package manager.
  • Fixed a huge memory leak on registering operations.
  • Registering speedups on 
    • correctly Cythonizing _step2.pyx
    • changing ImageMagick to Scikit-image
    • not requiring SExtractor and star matching if registering (aligning) same batch again
Full list of changes in distribution package on in Git repository.


Prints from AstroStack