In this paper, a new improved approach is built to obtain the information of building land, which is based on NDBI, a popular method in extracting building land information. Based on extracting the indexes of NDBI/NDVI/MDWI, the information of urban building land can be accurately extracted and the data of these indexes is contrasted. We analyzed the distribution discipline of building land, vegetable and water land in order to find these three indexes association pattern in order to find the best building land pattern on these indexes (The pixels are obtained which have most maximum data than NDVI and MDWI). Moreover, this paper gives an accuracy assessment using ALOS data of 2010. The results of accuracy assessment show this new method is satisfied for obtaining NDBI, the accuracy is up to 89% for high-density urban pixels. One the other hand, the NDBI information (19902010) of Guangzhou is extracted using this new method, and the trend of urban spatial expanding is analyzed and discussed. The results show as followed:(1) the improved method is effective to extracting the building land; (2) among the 20 years, the mean value of NDBI increases 0.15 and the area is increasing 220% from 192 km2 to 617 km.