One important component of Hugin is the control point detector, that identifies automatically the overlapping features in two images to align them in space. Hugin users currently have the choice of either setting these control points manually, or to use one of the flavors of autopano:
- The original autopano, by Alexandre Jenny, for which the source code is not available. Alexandre has moved on to develop his own excellent stitching software, Autopano Pro.
- Autopano-SIFT, a C# / Mono implementation by Sebastian Nowozin.
Autopano-sift-C, a faster C implementation of Autopano-SIFT that is hosted in Hugin’s SVN repository and available to Linux and OSX users was not available to Windows users until recently.
Last week Pablo ported the Autopano-sift-C build process to cmake, however a glue shell script is required to control the binaries. The script autopano-c-complete.sh, also written by Pablo, works well on Linux and OSX. A translation for Windows was needed. I took the translation of the script upon myself.
There is not much choice of scripting on a default Windows install. Perl, PHP, Phython and the likes have all been ported to the Windows architecture and can be downloaded, installed and run. But we don’t want to impose an additional download to our users. So I was left with two alternatives: prehistoric batch programming from the DOS era, or the more recent Windows Host Scripting. For all practical purposes, my skills at both were ZERO. My last experience with DOS batch files predates the Internet. So I looked at what Pablo’s autopano-c-complete.sh does and spent a few hours writing autopano-c-complete.vbs. It works form the command prompt on a simple test case, and I hope that testers of the recently released Hugin Windows Installer will help find eventual bugs.
Unfortunately it does not work from the Hugin GUI. A wxWindows error message appears instead, and I have reached my limits with Windows Script Host. I committed the script to the subversion repository.Maybe somebody with more WSH experience me can have a look at the committed script and improve it? That’s the beauty of Open Source.