Some question about the paper & code.

[Deleted user]

Sorry for disturbed, But some question I had think deeply two days, Still don't quite understand.

1. For Gaussian mixture :
   Why use pixels of the transformed templates as the centres of isotropic Gaussian components?
   However, the standard deviation is calculated in the original Image.
   What's means of that?

2.Why they add up all the transformed templates and then calculate the likelihood at this two-dimension data set?

3.Where is the alpha channel of the template described at paper?

I was really confused, Could you offer some perspective to understanding this!