ABSTRACT
In signal and image processing the operations performed on the signals are of a deterministic nature and best expressed as algorithms that uniquely describe the relation between the input and the output signals. The signals and images, on the other hand, are best described in probabilistic terms in the framework of probability and statistics. In many signal processing tasks, e.g., in speech and audio signal processing, it is most natural to analyze and process the signals according to their frequency content, i.e., using the Fourier transform representations. In image processing where the frequency content of the signal has a less natural meaning than for electrical or audio signals, the strength of statistical methods is even more pronounced. A complicating factor is that even the basic statistical concepts are based on quite deep mathematical theories.
