fix matlab bridge results

[jabooth]

We have some kind of scaling issue that is happening, at least with yzt_iccv_2013.

[jabooth]

@patricksnape I've started branch matlab to try and fix these issues.

So far, I've got this:
7ac74af

Which:

• Moves landmark saving to behave like PTS files (flip XY-YX, +1 to matlab, -1 back to Python)
• Converts the RGB unit image to a float greyscale in the yzt_iccv_2013 fit method (glancing through the fit method it seems like this is the required format?)

[grigorisg9gr]

Why don't you use some interactive bridge with matlab instead of this workaround?
E.g. with something like https://pypi.python.org/pypi/pymatlab you don't need to export everything to a .mat file, import with scipy etc. You can call the m-script and return the outcome right in python.

I have used that for debugging in conversions matlab/python, so even if calling it should be fairly easy, let me know if you have any questions.