Computationally efficient vanishing point detection for camera applications

This paper describes a computationally efficient approach to obtain vanishing points in images for consumer electronics applications. It involves modifications of the conventional J-linkage algorithm that is widely used for vanishing point detection. The developed approach involves reducing the number of hypothesis sets and utilizing an initial set of clusters using a skeletal set of edges. Both the computational efficiency and performance aspect of our approach are compared with the conventional J-linkage algorithm by examining the York Urban image database. It is shown that our approach significantly improves the computational efficiency of vanishing point detection while at the same time providing lower errors.