A toy model for a shape recognition problem in computer vision is studied within the framework of the Monge-Kantorovich optimal transportation problem with a view to understanding the underlying geometry of boundary matching. The questions addressed are the existence, uniqueness and geometric characterization of the optimal solutions.