Generalised syntactic pattern recognition as a unifying approach in image analysis

Abstract

In this paper we analyse theoretical foundations of syntactic pattern recognition and its relationships with mathematical linguistics, structural pattern recognition, and statistical pattern recognition. We point out that all the above mentioned methods are based on two fundamental operations: the combination of parts into wholes and abstraction. We provide examples of those operations in different contexts. Finally, we note that at this level of generalisation, all three main approaches to pattern recognition (structural, statistical and syntactic) can be merged into one powerful method. We outline a system based on these principles and intended to be used for the recognition of unconstrained handwriting.