Abstract : We present a new approach to the simplification of line drawings, that maintains the morphological structure of the original drawing while decreasing the number of lines. The technique works by analyzing the structure of the drawing at a certain scale and identifying clusters of lines. These clusters are then processed to create new lines in a separate stage where different scenarios can be favored based on the target application. This two-stages decomposition naturally adapts to various applications: density reduction of a drawing, where too many lines project in a given region of the image; Level-of-detail (LOD) representations for line-based rendering (contours and hatching), where the number of lines must vary with scale; and progressive editing, where the user refines a curve by successive sketches, viewed as an iterative simplification of the set of line sketches drawn by the user. Our goal is to provide a low-level tool common to these applications.