Abstract: In this paper,an image fusion algorithm based on visual correlation is proposed for video sensor networks. Given the severe resource constraints on individual video sensors,our algorithm takes fully account the redundant visual information among multiple video sensors and partitions a sensing task into the two highly correlated ones.Each video sensor only needs to cover a fraction of targeted area.In particular, we use the epipolar constraint to fuse the received multiple partial images,thus reconstruct a complete visual scene.The experimental results show that our algorithm is simple,feasible,and it can not only reduce the transmission workload and save network energy,but also effectively perform visual monitoring task.