Abstract: There are some weak spots in medium-resolution remote sensing image, such as weak information in seismic disaster of buildings and change detection method prone to be affected by non-seismic disaster information. According to those problems, this paper established a method for quickly extracting residential seismic damage information based on the change detection method, which uses principal component transform to enhance damaged information, and adopts the supervised classification method to extract the similar residential area, which were further optimized by using the nighttime lights data, eliminating the influence of non-seismic disaster factors in the change detection method. And then the meizoseismal area of India M_W7.6 earthquake in 2001 is chosen as the studied area by using Landsat thematic mapper (TM) imagery pre- and post-earthquake and nighttime lights data of earthquake-stricken area so as to verify and analyze the method proposed in this paper. The results show that the reflectivity becomes larger is the the most significant damaged characteristic for the building residential area in the resolution of 30--50 m medium-resolution remote sensing image. The method for extracting residential earthquake damage information established in this paper has good effects in solving weakly damaged information and background noise based on medium-resolution images.