The problem addressed in this paper is the high-level problem of distinguishing among photographs, graphics, texts and compound documents. To cope with the great variety of compound documents we have designed a hierarchical classification strategy which first classifies images as compound or not-compound by verifying the homogeneity of the sub-images in terms of low-level features. Not-compound images are then classified as photographs, graphics or texts. Results of our experiments on a database of over 35000 images collected from various sources will be reported and discussed in the final paper.