This paper presents an approach for recognizing urban scene in distant views. Previously, we have developed a method using dynamic programming technique to recognize key buildings appearing as silhouette in a distant view. Based on those locations of buildings in silhouette, this paper develops a method to find locations of other buildings which do not appear as silhouette in the image by using 2D template matching. We then describe some applications of the approach such as incorporating building texture information to 3D building models and estimating the height of abuilding. We also show its effectiveness in experiments on real images.