Parked vehicle edge computing (PVEC) is proposed to enhance the resource capacity of vehicular networks by utilizing the resources from parked vehicles (PVs). Toward the flexible and fine-grained resource usage, we integrate the container-based virtualization with PVEC for ensuring the task execution in PVs with fast response, increased scalability, and high efficiency. Then we study a social welfare maximization problem to maximize the overall utilities of users and PVs in computation offloading. The problem is decomposed into two subproblems for determining container deployment and computing resource allocation, respectively. Finally, numerical results are provided to demonstrate the effectiveness and efficiency of our scheme.