PDF version of this entire document
This chapter has explained, but has not yet demonstrated, some of the advantages gained by using an MDL approach for choosing landmark points in a set of shapes. The notion of an objective function was explained, as well the idea of using an information-theoretic objective function. Once this function is in place, there are various issues that are concerned with the optimisation. The way by which good solutions are sought is rather crucial. Subsequent chapters, which cover detailed experiments, elaborate further, using more practical examples.
