This paper describes a technique to represent relevant information of tree-like structures in a compact way. The technique is general. In the application described here, the images are obtained with contrast-enhanced magnetic resonance angiography (MRA). After segmentation, the vessels are reduced to fully reversible surface skeletons. Thereafter, a novel approach to curve skeletonization based on the detection of junctions and curves in the surface skeleton is used. This procedure results in a good description of the tree structure of the vessels, where they are represented with a much smaller number of voxels. This representation is suitable for further quantitative analysis, e.g., measurements of vessel width and length.