To solve the problem that multi-focus image fusion results in virtual shadow at the target object edge, a multi-focus image fusion algorithm is proposed based on the guided filtering and improved pulse coupled neural network (PCNN). The source image is decomposed by a guided filter with the multi-scale edge-preserving decomposition, and the preliminary fusion, and the obtained basic and detail images are fused preliminarily by different guided filtering weighted fusion strategies. Preliminary fusion image is used as external input excitation to stimulate the improved PCNN model. The source images are according to the fusion weight map to obtain the final fusion image. Experimental results show that, compared with traditional fusion algorithms, the detail information of edge, region boundary and texture of source images are preserved by the proposed algorithm, which avoids virtual shadow at target object edge, and improves fusion image quality.