As the multi-focus image fusion method can not deal with the noisy image effectively,this paper pr oposes a multi-focus image fusion method in the non-subsampled shearlet transform (NSST) domain with pulse-coupled neural network (PCNN) and g uide filtering. Firstly,high frequency sub-band coefficients and low frequency sub-band coeff icients are obtained by performing NSST respectively.The fusion rules are improved by comb ining guide filtering with improved PCNN model for high-frequency sub-band coefficients.Th e low level visual characteristics of phase consistency,sharpness and brightness are extract ed to guide the low-frequency sub-band coefficient fusion.Finally,the fusion result is obtaine d from the inverse NSST.The experimental results show that the proposed method c an effectively accomplish multi-focus fusion under the condition of noisy interference,and im prove mutual information by nearly 0.15under the condition of standard deviation of noise of 20,and the edge and texture information is kept good and effective.