This paper, the second part of our three-part contribution, is concerned with state estimation in a discrete-time hybrid system model, described in Part I. The original contributions of the paper are as follow. The observable dynamics, seen through a measurement restriction imposed on the system, is characterised in terms of measurability and distinguishability of transitions between activity states. Based on the characterisation, an algorithm, to construct an Observer for estimation of the current state of the overall system composed from its component models, is developed. The whole theory is illustrated with an example of a heating system explained in Part I. This methodology for state estimation is used in a companion paper, Part III, for solving the problem of fault diagnosis.