The work aims to solve the problem of image boundary blurring and visual worms generated by the existing halftone algorithm and get a better digital halftone image of visual perception. The α-truncated average filter was introduced based on the original classic filter and a new error diffusion filter considering the distance and the grey value similarity between pixels. Compared to the traditional error diffusion algorithm, the proposed algorithm reproduced the halftone image details in a better way and its boundary was also well maintained. Moreover, the PSNR and SSIM values of objective evaluation parameter were increased in some way, and the NSME value also got an ideal result, which not only solved the fuzzy boundary of traditional error diffusion algorithm, but also better expressed the detailed information on texture of the halftone images. The halftone image generated by the proposed algorithm has a better visual effect and the algorithm is simple and effective.