In this tutorial, I have presented what is probably the most commonly used techniques for parameter estimation in computer vision. Particular attention has been devoted to discussions about the choice of appropriate minimization criteria and the robustness of the different techniques. Hopefully, the reader will find this tutorial useful. Comments are extremely welcome.
Another technique, which I consider to be very important and becomes popular now in Computer Vision, is the Minimum Description Length (MDL) principle. However, since I have not yet myself applied it to solve any problem, I am not in a position to present it. The reader is referred to [19, 10, 11].