An image compression method for encoding gradation image data for use in, for example, a printer or a copying machine handling the gradation image. When a difference between two representative values in one block E)f the image data is less than a predetermined amount, a block code made up of a single piece of gradation data and one pixel array pattern is generated. When the difference is not less than the predetermined amount, the representative values are replaced by any of sixteen substitute values, and a block code is formed to include two gray scale data each of four bits and a single pixel array pattern, thereby realizing coding compression of the image information.