A New Line Segment Grouping Method for Finding Globally Optimal Line Segments

Abstract. In this paper we propose a new method for extracting line segments from edge images. Our method basically follows a line seg-ment grouping approach. This approach has many advantages over a Hough transform based approach in a practical situation. However, since its process is purely local, it does not provide a mechanism for finding more favorable line segments from a global point of view. Our method overcomes the local nature of the conventional line segment grouping approach, while retaining most of its advantages, by incorporating some useful concepts of the Hough transform based approach into the line segment grouping approach. We performed a series of tests to compare the performance of our method with those of other six existing meth-ods. Throughout the tests our method ranked almost highest both in detection rate and computation time.