An image compression method which successively diminishes in size the image to form a compressed image. The compression is done by dividing an image into partial areas with equal size, calculating a mean pixel value in each area, and substituting the mean pixel value into the diminished image. The image is then re-enlarged in the same way, and the enlarged image is compared against the original image to determine error values. The error values and the diminished images are used to represent a compressed version of the original image.