The application of synthetic aperture radar (SAR) in urban areas is far from resolved with an increased spatial resolution, so building extraction from SAR images remains a difficult task. According to SAR imaging principles, the outline of a building is usually incomplete in a single-aspect SAR image, and microwave interactions between adjacent targets further complicate this phenomenon in urban areas. Thus, in this study, dual-aspect high-resolution SAR images obtained, respectively from ascending and descending orbits are introduced to extract the building footprints in urban areas, and a method for building footprint extraction based on the fusion of dual-aspect SAR images is proposed. First these dual-aspect SAR images are co-registered, and then the preliminary positions of each potential building are determined using Markov random field models and Hough transform. Next the test images are partitioned into several subimages that contain only one building target. Then the edge of a building is extracted within the subimage of each aspect using a region-growing method and gradient algorithm, and then detection results obtained from each aspect are fused to produce the ultimate outline of the buildings based on Dempster-Shafer evidence theory. Experiments using TerraSAR-X images demonstrate that this method can extract the complete footprint of buildings in urban areas and can also improve the accuracy when estimating the dimensions of buildings.