Mobile edge computing (MEC) has emerged as a promising paradigm to realize user requirements with low-latency applications. The deep integration of multi-access technologies and MEC can significantly enhance the access capacity between heterogeneous devices and MEC platforms. However, the traditional MEC network architecture cannot be directly applied to the Internet of Vehicles (IoV) due to high speed mobility and inherent characteristics. Furthermore, given a large number of resource-rich vehicles on the road, it is a new opportunity to execute task offloading and data processing onto smart vehicles. To facilitate good merging of the MEC technology in IoV, this article first introduces a vehicular edge multi-access network that treats vehicles as edge computation resources to construct the cooperative and distributed computing architecture. For immersive applications, co-located vehicles have the inherent properties of collecting considerable identical and similar computation tasks. We propose a collaborative task offloading and output transmission mechanism to guarantee low latency as well as the application- level performance. Finally, we take 3D reconstruction as an exemplary scenario to provide insights on the design of the network framework. Numerical results demonstrate that the proposed scheme is able to reduce the perception reaction time while ensuring the application-level driving experiences.