A robust method for detecting planar regions based on random sampling using distributions of feature points

We propose a robust method for detecting local planar regions in a scene with an uncalibrated stereo. Here, we assume that the correspondences between the two images have been established. Our method is based on RANSAC for estimating homographies to the local planar regions in the scene. For doing this, we adopt double random sampling scheme by a uniform distribution and the local probability distribution of each pair, which is defined by the distances from the point to the others in one image. We first choose a pair as a seed by the uniform distribution, and then choose four pairs by the local probability distribution with respect to the seed. By introducing the local probability distribution, we can efficiently choose four potentially coplanar pairs in the scene. The same scheme can be applicable to detect line segments in an image. We demonstrate that our method is robust to the outliers in a scene by simulations and real image examples.