Disclaimer: this information is essentially the same an one I provided in my answer about particle filters on SO
This online course deals with Monte Carlo localization, Kalman fitlers and particle filters.
It is called "Programing a Robotic Car", so not really image processing, but: It is explained in very general way, so the principles explained are easily applicable to any kind of input (e.g. sonic, gps, visual...).
It has examples in python, including "in class tasks" and "homeworks" (various non-programming and programming tasks meant to help you understand and learn how to implement the various filters).
It is taught by a Standford professor if I'm not mistaken, and it's very easy to follow. By the end of the course, you should have all three methods implemented in python and ready to apply to any kind of data, including visual.
Disclamer 2: I assume you mean applications in Computer Vision, since Image processing deals mostly with static images. To estimate something, you would need multiple data, that is a video sequence, which makes it lean more towards CV than IP.