More recently, some people began to realize that real data do not completely satisfy the classical assumptions.
Rousseeuw and Leroy [239 ]
Problems requiring a best guess [12] for the value of some parameters based on noisy input data are estimation problems. Alternatively, you might view estimation problems as problems where you try to characterise a dataset using a mathematical construct such as a point or a line. This chapter is intended as a self-contained introduction to the basic concepts in the field of linear estimation. As such, it will give you an understanding of the analysis in the following chapters where we focus specifically on motion estimation. The discussion here is quite informal: our aim is to introduce the broad ideas of estimation more detail is available from a vast range of sources [64, 122, 148, 201, 239, 267, 303]. This chapter begins with a brief introduction to probability theory and some key terms before moving onto the basics of estimation, which then allows us to spring into the field of robust estimation. It is assumed you are familiar with basic calculus.
[12]We will come to the meaning of best guess later.
This section defines a range of basic concepts from probability theory that appear in the estimation literature. Concepts are presented in order with earlier concepts building on later concepts, thus it may help if this section...
TABLE OF CONTENTS 
