We show that, in images of man-made environments, the horizon line can usually be hypothesized based on a-contrario detections of second-order grouping events. This allows constraining the extraction of the horizontal vanishing points on that line, thus reducing false detections. Experiments made on three datasets show that our method, not only achieves state-of-the-art performance w.r.t. horizon line detection on two datasets, but also yields much less spurious vanishing points than the previous top-ranked methods.

1 (a) Image from the database (b) Facade detection and recognition + warping of a 2D virtual object (c) Pose computation and projection of a 3D virtual object Figure 1: Facade proposals for Urban AR. A facade of the Nantes Event Center building (a) is automatically detected and recognized in two views of the building (b, red polygons). From these results, any planar virtual object added to the facade (here the ISMAR logo) can be warped according to the transformation of the facade. (c) When some geometric information about the facade is available, any 3D virtual object expressed in the same reference frame as the facade (here the ISMAR building) can be added to the view.

This article presents an efficient approach for accurate registration of a building facade model "dressed" with dense semantic information. Localization sensors such as the GPS as well as vision-based methods are able to provide a camera pose in an efficient and stable way, but at the expense of low accuracy. We propose here to rely on semantic maps to improve the accuracy of a rough camera pose. Simultaneously we aim to iteratively improve the quality of the semantic map through the registration. Registration and semantic segmentation are jointly refined in an Expectation-Maximization framework. We especially introduce a Bayesian model that uses prior semantic segmentation as well as geometric structure of the facade reference modeled by Generalized Gaussian Mixtures. We show the advantages of our method in terms of robustness to clutter and change of illumination on urban images from various databases.

We present a method for automatic facade rectification and detection in the Manhattan world scenario. A Bayesian inference approach is proposed to recover the Manhattan directions in camera coordinate system, based on a prior we derived from the analysis of urban datasets. In addition, a SVM-based procedure is used to identify right-angle corners in the rectified images. These corners are clustered in facade regions using a greedy rectangular min-cut technique. Experiments on a standard dataset show that our algorithm performs better or as well as state-of-the-art techniques while being much faster.