An image processing method that puts input image data represented by multi-value pixels into gradation representation of dots having binary data of plural minute pixels while performing error diffusion processing. The method includes a first step that when a dot is formed which corresponds to correction image data produced by adding correction data to multi-value image data of a multi-value target pixel, only in the case where the dot is to be formed so as to be adjacent to a dot formed for a proper adjacent multi-value pixel, forms the dot even if it has less than k (2 or greater integer) number of minute pixels. In other cases, the method forms only a dot having k number of minute pixels or more.