Advances in communication and computation technology facilitate closed loop control via communication networks. Such networked control systems may include distributed sensors and actuators as well as distributed computational resources. In the field of vision-based control, the possibility of distributed computation is of particular interest due to the large amounts of visual data to be processed. Time-delay caused by data transfer through the network and by processing algorithms impair stability and performance of the controlled closed-loop system. In this article, a novel switching control scheme for networked vision-based control systems is proposed accounting for varying time-delay in order to improve control performance. The closed-loop system is modeled as a continuous system with varying time-delay considering processing and transfer time-delays as well as sampling intervals. Mean exponential stability is proven based on occurrence probabilities of delays. The article concludes with comparative experiments showing a significant improvement of control performance of the proposed approach with respect to a non-switching reference controller.