You should provide more information about your objects, conditions and performance requirements. Here is very extensive survey on Object Tracking: http://vision.eecs.ucf.edu/papers/Object%20Tracking.pdf. You will probably find something for your case.
Here are some possible approaches:
- If you're going to track 12 color balls which can't intersect and be occluded (view from the top), you can use some foreground/background segmentation algorithm, i.e. BackgroundSubtractorMOG. Having the motion mask you can filter noise with morphological operations, detect blobs and track them between frames.
- If your objects are planar textures objects, you can use some feature-based approach. Here is a short video showing how you can track a book or logo. Sample is in Python, but you can easily convert it to any other language.
- If your objects are not planar (have different view from different angles, i.e. people), you can think about something like TLD (aka Predator).
So, the best approach really depends on the type of your objects and the view from camera.
