This paper addresses scene understanding in the context of a moving camera, integrating semantic reasoning ideas from monocular vision with 3D information available through struct...
—Image-based computation of a 3D map for an indoor environment is a very challenging task, but also a useful step for vision-based navigation and path planning for autonomous sys...
In this paper, we present a two-layer generative model that incorporates generic middle-level visual knowledge for dense stereo reconstruction. The visual knowledge is represented...
In this work we recover the 3D shape of mirroring objects such as mirrors, sunglasses, and stainless steel objects. A computer monitor displays several images of parallel stripes,...
Stas Rozenfeld, Ilan Shimshoni, Michael Lindenbaum
We propose a novel formulation of stereo matching that considers each pixel as a feature vector. Under this view, matching two or more images can be cast as matching point clouds i...