Abstract: In order to improve the multi-focus image sharpness and the practicability of PCNN in the aspect of image fusion, a new multi-focus image fusion method was proposed based on nonsubsampled Contourlet transform (NSCT) and dual-channel pulse coupled neural network (DCPCNN). Firstly, multi-focus images which from the same scene were decomposed using NSCT respectively; and then visual properties contrast based on SML was used to fuse low frequency subband coefficients; High frequency sub-band component fused through DCPCNN which motivated by modified spatial frequency. DCPCNN had two channels corresponding to different linking strength, high frequency sub-band fusion coefficient was determined according to the number of ignition. Finally, the fusion image achieved by inverse transformation of NSCT. The experimental results demonstrate that the proposed algorithm improves fusion image sharpness significantly, subjective and objective indicators are better than other fusion algorithms, having better visual effects.