This paper studies decentralized output-feedback control of large-scale nonlinear systems in the presence of sensor noise. The measurement errors are dealt with by set-valued maps in the recursive control design of the subsystems. With appropriately designed decentralized controllers, the subsystems are rendered input-to-state stable (ISS) with respect to the outputs of the other subsystems and the disturbances. The robust stability of the closed-loop decentralized system is guaranteed by a recently developed cyclic-small-gain theorem.