We present a new algorithm, called LB-Triang, which computes minimal triangulations. We give both a straightforward O(nm ) time implementation and a more involved O(nm) time implementation, thus matching the best known algorithms for this problem.Our algorithm is based on a process by Lekkerkerker and Boland for recognizing chordal graphs which checks in an arbitrary order whether the minimal separators contained in each vertex neighborhood are cliques. LB-Triang checks each vertex for this property and adds edges whenever necessary to make each vertex obey this property. As the vertices can be processed in any order, LB-Triang is able to compute any minimal triangulation of a given graph, which makes it significantly different from other existing triangulation techniques.We examine several interesting and useful properties of this algorithm, and give some experimental results.
Background and motivationComputing a triangulation consists in embedding a given graph into a triangulated, or chordal, graph by adding a set of edges called a fill. If no proper subset of the fill can generate a chordal graph when added to the given graph, then this fill is said to be minimal, and the resulting chordal graph is called a minimal triangulation. The fill is said to be minimum if its cardinality is the smallest over all possible minimal fills, and the corresponding triangulation is called a minimum *