The increasing growth of maritime activities leads to the challenges for processing the maritime data. However, the resources-limited maritime devices cannot meet the requirements of transmission delay and energy consumption. In this paper, we investigate the resource allocation for computation offloading in maritime communication networks via game theory to improve the offloading efficiency and reduce the energy consumption of maritime devices. Specifically, we propose an optimization problem that jointly optimizes the offloading data, the computation resource allocation of unmanned surface vehicle (USV) and the allocation of acoustic channels, with the objective of minimizing the total energy consumption of underwater wireless sensor (UWS). Despite the non-convexity of the joint optimization problem, we propose a layered structure and decompose it into a top-problem for optimizing the data offloading, a middle-problem for optimizing the computation resource allocation of USV, a bottom problem for optimizing the channel allocation. We conduct simulations to validate the effectiveness and efficiency of the proposed algorithms.