With the increase in bandwidth and computing power of wireless devices, video applications over wireless ad-hoc networks are expected to become widespread in intelligent vehicles, emergency communication, and battlefield command. However, a crucial problem is how to select the best paths for video streaming of qualify-on-service (QoS) in multi-hop wireless networks. Most of the existing research done on this topic tend to ignore the impacts of time- varying channel and wireless interference on the quality of multi-path video streaming. This paper proposes an optimized multi-path selection algorithm which takes not only network congestion into account, but also interference. Packet collision and delay in a MAC layer is predicted using the interference model. Each node is modeled as an M/M/1/K queuing system. Packet delay and loss, due to congestion, are predicted using the queuing theory. The distortion of path is defined as a function of packet losses and delays along the path. The paths with the minimum estimated distortion are selected as the optimal routings. Extensive experiments in NS-2 simulation environment have been carried out. The experimental results show that this algorithm achieves a certain level of satisfaction in the QoS of video streaming.