Abstract: In order to reduce network load, energy consumption and delay in video sensor network, a layered multi-quality image fusion method based on attention is proposed. A dynamic-attention based wakeup mechanism is established by structured deployment and region-mapping between views of nodes. The low-quality image acquired by high-layer node is fused with high-quality images from low-layer node, and the attention target is enhanced. Experiment results further showed that the fusion method is effective.