3D planar patch extraction from stereo using probabilistic region growing

This article presents a novel 3D planar patch extraction method using a probabilistic region growing algorithm. Our method works by simultaneously initiating multiple planar patches from seed points, the latter determined by an intensity-based 2D segmentation algorithm in the stereo-pair images. The patches are grown incrementally and in parallel as 3D scene points are considered for membership, using a probabilistic distance likelihood measure. In addition, we have incorporated prior information based on the noise model in the 2D images and the scene configuration but also include the intensity information resulting from the initial segmentation. This method works well across many different data-sets, involving real and synthetic examples of both regularly and non-regularly sampled data, and is fast enough that may be used for robot navigation tasks of path detection and obstacle avoidance.

Sparse motion segmentation using multiple six-point consistencies

We present a method for segmenting an arbitrary number of moving objects in image sequences using the geometry of 6 points in 2D to infer motion consistency. The method has been evaluated on the Hopkins 155 database and surpasses current state-of-the-art methods such as SSC, both in terms of overall performance on two and three motions but also in terms of maximum errors. The method works by finding initial clusters in the spatial domain, and then classifying each remaining point as belonging to the cluster that minimizes a motion consistency score. In contrast to most other motion segmentation methods that are based on an affine camera model, the proposed method is fully projective.

Affine Invariant, Model-Based Object Recognition Using Robust Metrics and Bayesian Statistics

We revisit the problem of model-based object recognition for intensity images and attempt to address some of the shortcomings of existing Bayesian methods, such as unsuitable priors and the treatment of residuals with a non-robust error norm. We do so by using a refor- mulation of the Huber metric and carefully chosen prior distributions. Our proposed method is invariant to 2-dimensional affine transforma- tions and, because it is relatively easy to train and use, it is suited for general object matching problems.