sdh4

The speed problem here seems to come from the self._synchronize() call completely rebuilding the inner list and then the node list from the inner list. When you do that iteratively...

Implemented the first option mentioned above (Image2D class handles both cases) and submitted a pull request #2 . Feel free to reject if you don't like this approach.

.... Happy to help out if you want me to test out changes or propose an abstract class interface. Anything terribly substantive might have to wait until the new year...