To solve the problem of multi-focus image fusion, a new multi-focus image fusion algorithm based on the visual perception feature was proposed. Because the threshold of human visual contrast sensitivity was proportional to the image background brightness, the visual uniform parameter was adopted to separate clear objects from fuzzy objects obtained by different image sensors. Firstly, the image was decomposed at RGB level separately. Secondly, the R,G,B single gray image was divided into sub-blocks. Thirdly, the sub-blocks with higher uniform value were selected as the corresponding sub-blocks of fusion image. Then, the retained sub-blocks were reconstructed to compose the fusion image. The immune genetic algorithm was applied to calculate the optimal number of sub-blocks, and the image quality criterion data, root-mean-square error and image entropy, were chosen as the affinity function of the optimal algorithm. The results have shown that the image fusion algorithm proposed was suitable to multi-focus image fusion and easy to realize.