Method of estimation of text labels recognition systems efficiency parameters with use of likelihood characteristics tree

In article the method of estimation of text label recognition systems efficiency is offered. Processing of input images is presented as sequence of events of a kind «algorithm has returned the correct (wrong) answer». On this basis the diagram of all possible sequences of these events is generated. It is tree of likelihood characteristics. With use of the tree functional dependence of recognition system characteristics from characteristics of separate algorithms is determinated.