Abstract: The binarizing method uses a diffusion coefficient matrix for error diffusion that has a two dimensional frequency and fixedly corresponds to each pixel of the input image, and error that is produced by comparing the referred input pixel value and threshold value is divided in the proportion of the diffusion coefficients in the sub matrix of the diffusion coefficient matrix and those divided errors are added to pixels neighboring the referred pixel, and the diffusion coefficient matrixes for each color are rotated with mutually different angles in the case of color printing.