An image fusion algorithm based on improved PCNN is proposed in compressive sensing domain.Original images are compressed by compressed sensing to obtain measurements.Improved PCNN model is built based on physical significance of measurements.Namely link coefficient,weighting matrix and threshold amplification coefficient are set adaptively.Measurements are selected as input neuron to obtain ignition-map as fusion operator.Fusion image is obtained according to reconstructing algorithm.The results show that this algorithm overcomes the detail missing of traditional compressed sensing image fusion,and the subjective and objective evaluation indexes show that this method is superior to the existing compressed sensing image fusion methods.