Accurate reconstruction of urban façades remain a popular and important research topic. Usually, structure from motion (SFM) and multi-view stereo are used to obtain a sparse or dense 3D point cloud from a set of images. The sparse point set is then improved with the help of user interaction, vanishing points [Sinha et al. 2008], repetitive elements, symmetries, etc. Our aim is to use the point cloud created by SFM and optimize the resulting model by incorporating additional information. Particularly, we are investigating the potential of shadows and selfshadowing as seen in 1(a) for reconstruction purposes. Shadows have been used in some work as a tool to obtain 3D models, as for carving and as a weak structured light. In contrast, we use shadow information to improve the achievable reconstruction results. Shadows cause problems in SFM as they change over time, leading to confused correspondence results. Instead, we propose to use shadows to our advantage and non-locally couple the façade planes. Copyright is held by the author / owner(s).