Python is getting popular for 3d visualization - several labs use it for brain imagery, for example. So it is definitely usable for that kind of problems. Library-wise, I guess you are aware than things like vtk and co can be used from python.
http://neuroimaging.scipy.org/site/doc/manual/html/faq/why.html
Of course, doing it in C++, in theory, will result in faster programs, everything else being equal (same algorithms, programmers proficient in the language, etc...) and given 'infinite time'. But python is certainly more productive than C++: in a given amount of time, you will get more things done in python than in C++, unless you are a much better programmer in C++ than in python. Also, if there is a need to interact with scientific people who are not programmers, python will be more readable than C++. C++ produced by researchers/scientific people tend to be really horrible in my experience (even worse than Fortran :) ).
The biggest problem I can foresee is if python is not fast/memory efficient enough, so that it involves a lot of constraints from an architectural point of view. I don't think it will be the case for imagery (python, as most high level languages, has scalability problems when you need ten of thousand of objects with a lot of interactions - that's one of the few areas where using C++ for a project started from scratch still makes sense IMHO nowadays).
Disclaimer: I should mention that I am involved quite a lot in the scientific python community, and that I am not an image processing specialist.