In order to preserve more features of source image information and overcome the defect, such as the fusion algorithms based on gradient which is prone to ghosting, this paper proposes a multi-focal image fusion algorithm based on weighted degree of focus. The algorithm calculates the focus metric for each source image with the spatial frequency, the improved operator sum of modified-Laplacian, and the gradient weighting sum, get a focus state images MSF, Mi-SMLand MSoG, then find the corresponding membership value ? SF(z), ? i-SML(z) and ?SoG(z), according to the membership function of these measures, get the weight of each source image, finally generate the final fusion image according to the linear normal method. The algorithm and the current several typical image fusion algorithms have the subjective and objective evaluation. The results show that the standard deviation (σ) is higher than 4.699, the average gradient ( G ) is higher than 0.382, the information entropy (H) is the highest, the other evaluation parameters are also within the acceptable range, the fusion image is clearer, and no artifacts are produced.