Identification of planes from a pair of uncalibrated stereo images is a challenging problem as it can lead to extracting virtual planes instead of physical ones, especially for complex scenes. We propose a new homography-based approach to extract physical planes and to distinguish them from virtual ones for general scenarios. The proposed approach uses noncoplanar points inside a plane to decide whether the plane is physical or virtual. Depending on the distribution of the points inside the convex hull of the plane, the plane is classified as likely virtual, likely physical, very likely virtual or very likely physical. To estimate homographies, we use our method which computes the homography for three points with no necessity to assume the fourth point being coplanar with the three. Experiments on real images demonstrate the validity of the proposed approach for general scenarios. © 2009 Springer Berlin Heidelberg.
CITATION STYLE
Amintabar, A., & Boufama, B. (2009). The Distinction between Virtual and Physical Planes Using Homography. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 5627 LNCS, pp. 727–736). https://doi.org/10.1007/978-3-642-02611-9_72
Mendeley helps you to discover research relevant for your work.