As an update, I can corroborate the answers previously given.
I have used both OpenCV with C++ and the EMGU Wrapper for both research and work and I agree partially with most people.
EMGU CV can be easy to learn and your code will be much easier to understand, it will require less lines of code and there is a decent range of functions you can call. But it is not complete.
Many functions will not provide all the originally intended functionality that you will find with OpenCV 3.0 e.g. FindNonZero() does not work exactly the same as the function that is wrapping and SetValue() gave me a hard time throwing runtime errors difficult to debug.
On the other hand, OpenCV provides great support compared to any other Wrapper you might find online and you will be working with the most up-to-date libraries. OpenCV 3.1 is a great release and I found no problems running it on Xcode or Visual Studio.
Therefore, I would recommend anyone wanting to immerse themselves into Computer Vision and Image Processing to go straight for OpenCV C++. If you just want to work on less complex problems you can opt for EMGU (I cannot recommend others as I am not familiar).