Upon scaling input image data and outputting the scaled image data, an address arithmetic processor determines phase information, which indicates the distances between the position of the pixel of interest after scaling and neighboring pixels of an original image, in accordance with the scale of that scaling, a coefficient arithmetic processor calculates filter coefficients in a filter process in accordance with the determined phase information, and a filter processor executes the filter process of the input image data by convolution operations on the basis of the calculated filter coefficients and outputs multi-valued image data. An LUT converts the density of the multi-valued image data in correspondence with the density characteristics of a printer, and a binarization processor binarizes the multi-valued image data and outputs binary image data to an image storage unit.