Mobile edge computing (MEC)-assisted Virtual Reality (VR) rendering has emerged as a promising solution to alleviate the computational burden on mobile devices and significantly enhance the Quality of Experience (QoE) for VR users. However, with the rapid surge in VR rendering requests, MEC-assisted rendering systems are confronted with the challenge of ensuring high rendering quality and ultra-low latency delivery while operating under resource constraints. To address this challenge, this paper proposes an efficient MEC-assisted VR rendering scheme aimed at maximizing the overall QoE for all users. We introduce a novel VR rendering framework that integrates frame caching and reuse, supporting rendering frames through rendering pipeline or reprojection, thereby enhancing rendering efficiency. Then we formulate the joint optimization of computation offloading and rendering quality for QoE as a long-term optimization problem. To solve this complex long-term joint optimization problem, we use Lyapunov optimization theory to transform it into a series of single-step problems and further decouple each single-step problem into two interdependent subproblems. We propose a genetic-based algorithm that alternately optimizes the two subproblems using a genetic algorithm and a heuristic method. Simulation results demonstrate that our proposed scheme outperforms other existing schemes in terms of overall user QoE, average display delay, and other key metrics.