A method for compressing an image, includes: calculating a level of hues of the image over at least all of one layer of the image; depending on the type of hues of the representative layer, classifying the image in one of the following three classes: a first class if the image is of a graphics type; a second class if the image is of a highly contrasted type; a third class if the image is of a low-contrasted type; and, choosing a compression processing type depending on the class of the image: difference processing, if the image is of the first class; frequency processing, if the image is of the third class; and, if the image is of the second class: for lossless or low-loss compression, preferably using difference processing, and, in the other cases preferably using frequency processing.