Abstract: In this paper, a multilevel error diffusion algorithm with adaptive noise modulation is proposed to alleviate the banding effects and other artifacts exited in the multilevel halftoning algorithms in literature.At first, the coefficients of the noise modulator are computed based on the weighted mean square error balancing, and the noise modulator is inserted into the quantization process of the error diffusion algorithm.Then, a blue noise spectrum based genetic algorithm is introduced to optimize the adaptive error diffusion filter with noise modulation.Finally, the output image is obtained through the quantization error filtering with the adaptive error diffusion filter.Experimental results show that our algorithm could not only effectively reduce the banding effects and other artifacts, but also achieve a nice signal-to-noise ratio.