Abstract: A plurality of uniform density pixel matrices Di are prepared so that each matrix Di will have all pixels of a uniform density i. First, in S110, the pixels of one uniform density pixel matrix Di are converted into binary values (0 or 1) while performing an error diffusion operation. Thus obtained binary value pixel matrix Fi is stored in the working memory 14. Then, while the uniform density value i is repeatedly incremented, the above-described converting-and-storing processes are repeated. Afterward, binary values throughout all the binary value pixel matrices are summed at each pixel position. Thus, in S150, an accumulated value matrix M1 is produced to have a corresponding element constructed from the total value. Next, threshold values of a dither matrix DM are determined based on the elements of the matrix M1. That is, the threshold values are determined one by one using the elements of the matrix M1 from its element having the lowest value.