Abstract: A printer model and human visual model based method on tone-dependent error diffusion(TDED) is proposed.First the printer model and HVS(human visual system) used in halftoning algorithm are analyzed.And based on the Floyd-Steinberg error diffusion algorithm,the tone-dependent weights matrix is trained on the different tones using an error within a human visual system.In the experiment,the printer is used to test the result and compare the quality of the halftone image created by the new algorithm,and both them show very good halftoning quality.