To meet the requirements of various computing intensive and delay-sensitive applications in vehicular networks, it is considered an effective approach to transmit the tasks of mobile users to nearby mobile edge computing (MEC) servers through computation offloading. In such a vehicular edge computing network, in addition to MEC resources, vehicles are also equipped with edge computing capabilities by carrying a cloudlet platform. Therefore, this paper designs a collaborative vehicular edge computing framework to realize the collaborative processing of computing-intensive and delay-sensitive tasks on resource-rich vehicular cloudlets. Based on this framework, multiple vehicular cloudlets can form a vehicular coalition and simultaneously process tasks of mobile users, thereby further improving the computing efficiency of vehicular edge computing networks. We formulate the collaborative task processing problem as a combinatorial auction with the objective of minimizing the total task processing cost of all mobile users. Moreover, a low-complexity greedy-based collaborative computing algorithm is proposed as our solution. Numerical results show that compared with the benchmark schemes, the proposed scheme performs better in minimizing the total task processing cost.