Correspondences of straight lines using graph-theoretic approach

A method of matching straight line segments from image pairs is proposed. Each line has a rich set of features that are used in finding candidate matches to a line. Along with this local property, there is a structural relationship that each line possesses which is exploited for the matching process. This allows each image to be represented as a relational graph with nodes representing lines and with the edges representing the line´s structural relationship with the other lines. An association graph is obtained from the two relational structures. The graph is then refined by an iterative process. The largest maximal clique in the refined graph represents a consistent set of assignments that yields the best match