The Python Imaging Library, PIL, is a good compromise between speed-to-market and good performance (and you can always use scipy and its core part, numpy, to enrich it for more advanced image-processing needs, if you pick Python as your pivot language!-). Similarly, visualization (including 3D) are excellently covered in third-party Python extensions -- check out EPD, the Enthought Python Distribution, for a good idea of what libraries might best help you in such tasks (you can always build your own versions if you don't want to partner with Enthought for commercial distribution... but it might be worth checking them out, as they have excellent commercial contacts as well as tech skills;-).
When and if you want to dip down into C++ for some specific component, Boost.Python, SIP, or Cython will make it child's play to integrate the component into your Python mainstream. For UI &c, PyQt is great...
In other words, while I'm obviously biased, in your shoes I'd unhesitatingly go for Python as the "core" and investigate the various options I've mentioned for visualization, UI, etc, etc. One caveat: for quick time-to-market, stick with Python 2.6: the newest 3.1, while great in many respects, is likely to still miss compatible versions of many third party extensions that will make your life way easier and sweeter with Python 2.6!