Line-drawing interpretation using probabilistic relaxation

Abstract

This paper shows that probabilistic relaxation is an effective method in the automatic interpretation of line drawings consisting of lines, symbols, and characters, such as electricity distribution diagrams superimposed on maps. The line interpretation problem has been newly formulated as a labeling problem in which probabilistic relaxation is used to obtain globally consistent results. The proposed automatic interpretation method consists of two stages. The first is segmentation and recognition of primitive components, such as symbols, characters, and long lines. The second is long-line interpretation, where probabilistic relaxation is introduced.