We propose an approach for detecting and matching facades from an aerial view to recover a set of corresponding facades from street-view images. Our approach exploits the regularity of urban scene building facades as captured by their lattice structures. Color, texture, and shape context similarity, as well as estimated surface orientations are used in screening and matching candidate views. Our experimental results demonstrate effective matching of oblique facades between ground and aerial views of urban scenes. Quantitative comparisons convincingly show superior performance of our proposed method over baseline SIFT, Root-SIFT and the more sophisticated Scale-Selective Self-Similarity and Binary Coherent Edge descriptors, for automated matching of aerial and street views. We also demonstrate a regularity-based approach for removing occlusions from street views and image enhancement through higher-resolution texture-replacement in aerial views.
