Earthquakes often induce collapse or cause extreme damage to large areas of buildings. One of the most important requirements for earthquake emergency operations is staying up-to-date on the extent of structural damage in earthquake-stricken areas. Terrestrial laser scanning (TLS) technology can directly obtain the coordinates of mass points while maintaining a high measurement accuracy, thereby providing the means to directly extract quantitative information from surface cracks on damaged buildings. In this paper, we present a framework for extracting wall cracks from high-density TLS point clouds. We first differentiate wall points from nonwall points using the TLS data. Then, a planar triangulation modeling method is used to construct a triangular irregular network (TIN) dataset, after which a raster surface is generated using an inverse distance weighting point cloud rasterization method based on the crack width. Then, cracks are extracted based on their shape features. We extract six sets of wall cracks from a damaged building wall in Beichuan County as an example of employing the above-mentioned method; the damage was caused by the Wenchuan earthquake. Quantitative calculations reveal that the extraction accuracy of the proposed method is greater than 91% and that the rate of leakage detection is less than 10%. In addition, the main limiting factor of the extraction accuracy is the crack width, that is, a wider crack will result in a higher extraction accuracy. In addition, the crack connectivity and leakage rate are negatively correlated, that is, a higher connectivity corresponds to a lower rate of missed extractions.