This paper investigates the offloading of image segmentation tasks and the allocation of corresponding computing resources for the Internet of Vehicles (IoV) supported by edge intelligence. With the convergence of 5G technology and artificial intelligence, the demand for high-precision sensors and navigation in smart connected vehicles is growing. Image segmentation technology, a crucial component of autonomous driving systems, requires substantial computing power and communication bandwidth. Faced with the shortage of onboard computing power and rising costs, edge computing offers a solution by offloading computing tasks to roadside servers close to the data source, i.e., connected cars, thereby alleviating network bandwidth and power consumption pressures and reducing system latency. This paper constructs an efficient edge computing resource allocation system based on an improved model-free Soft Actor-Critic (iSAC) algorithm with maximum entropy, and enhances the offloading efficiency by employing an integrated computing and networking scheduling framework to minimize task completion time. By incorporating Prioritized Experience Replay (PER), the iSAC algorithm accelerates the learning process while maintaining stability and improving the efficiency and accuracy of computation offloading. Simulation experiments compare the performance of iSAC with baseline algorithms, demonstrating its advantages in reducing error rates and optimizing task completion time. Future research will investigate task diversity and priority requirements in IoV.