Real-time vanishing point detection using the Local Dominant Orientation Signature

The vanishing point can be defined as a point generated by converged perspective lines, which are parallel in the real world. We propose a novel algorithm to detect the vanishing point in various images in real-time. The proposed algorithm unfolds into three steps. In the first step, we introduce the Local Dominant Orientation Signature (LDOS) descriptor to extract structural feature of an image. Then, we detect vanishing point candidates using dynamic programming. Finally, we estimate the location of the vanishing point from detected vanishing point candidates. Unlike the previous methods, the proposed method, which uses the dominant orientation of the local image structure, is fast and not limited to specific image contents. Experiments are performed on diverse images to confirm the efficiency of the proposed method and to show that it can be employed in various real-time applications.