Recently, Ushio and Takai have proposed a Mealy-automata-based framework to study the non-blocking supervisory control problem under partial observation. This framework can handle observation uncertainties by taking both state-dependent observations and nondeterministic outputs into account. In this technical note, we propose a model-transformation-based approach to solve the supervisor synthesis problem in this framework. First, we propose a transformation algorithm that transforms the non-blocking supervisor synthesis problem for Mealy automata to a conventional supervisory synthesis problem under partial observation, which can be solved effectively by an existing algorithm. Then we show that the supervisor synthesized for the transformed problem indeed solves the original problem. Our results bridge the gap between the conventional supervisory control framework under partial observation and the recently proposed Mealy automata framework where nondeterministic output function is used.