As increasingly more applications are deployed in vehicles, how to provide the demanded computational capability becomes crucial. This paper proposes the hybrid vehicular cloud (HVC) framework for cloud computing on the road to increase the computational capability of vehicles by using resources from the centralized cloud, RSU, and neighboring vehicles. The proposed scheme can be used as a general framework for all types of applications. It is able to satisfy various types of job requirements, such as real time or host requirements, and it is adaptive to the dynamic vehicular environment. Specifically, this paper defines the procedures to support the autonomous organization of a vehicular cloud. An online scheduling algorithm is proposed to solve the job assignment problem with the objective of processing more jobs and reducing cellular network usage. Extensive simulations are conducted, and the results show the superiority of the proposed scheme over competing schemes in typical urban scenarios.