An image-processing apparatus for quantizing multi-level (M-level) image data into N-level values, where M>N>1, using a multi-level error-diffusion process or a minimum-average multi-level error method is disclosed.The image-processing apparatus includes: means for outputting correction data; means for setting a quantization-threshold value; means for comparing the correction data and the quantization-threshold value so as to output N-level image data; and means for calculating an error generated with generating the N-level image data.