TL;DR: This paper develops the Scale-Invariant algorithms, which incorporate prior shape information by defining prior probabilities on support vectors, where a support vector is a vector formed from the lateral displacements of a particular set of support lines of an object.