Abstract: The purpose of this paper is to study and improve the performance of real-time media streaming in wireless multimedia sensor networks (WMSNs). This work begins with thorough investigation and understanding of the performance bottleneck of real-time MPEG video transmission over WMSNs. By analyzing the relationship between end-to-end throughput and video frame rate, we choose IFQ queue length as an important indicator of network congestion. On this basis, an adaptive flow control strategy based on cross-layer design is proposed for performance improvement of wireless media streaming. Our strategy works as follows when the wireless link condition is getting poor, a sensor node discards partial low-priority video frames actively so as to increase the probability that high-priority video frames are delivered successfully. This strategy could contribute to improve video playing quality, while reducing unnecessary energy consumption. To verify the effectiveness of this strategy, this paper set up two groups of experiments, a simple one without background traffic flow and a complex scenario with interference from other nodes. A series of key parameters, such as average end-to-end delay, packet loss rate, decodable frame ratio are considered with different video frame rates. Our simulation results demonstrate that the proposed strategy alleviates the performance bottleneck and achieves significant performance improvement for real-time media streaming over WMSNs.