Accurate extraction of urban built-up areas is an essential prerequisite of urbanization research. This paper proposes a local optimal threshold method based on NPP-VIIRS nighttime light images to extract urban built-up areas in mainland China. First, according to the regional economic differences, the image is divided into eight regions. Second, the maximum variance ratio threshold method and particle swarm optimization algorithm are combined to extract the optimal threshold for segmenting the night light images in each region. The optimal threshold for each region is used to extract the urban built-up areas from the NPP-VIIRS images of each region. Finally, eight urban built-up areas are merged. In order to verify the extraction accuracy, one city-level city and one county-level city are extracted in each region, and the boundaries of urban built-up areas are compared with a Google Earth historical image from June 15, 2015. The experimental results show that the kappa for the eight selected city-level cities (Shanghai, Chengdu, Lanzhou, Changsha, Changzhi, Tianjin, Shenzhen, Harbin) is approximately 0.80. The kappa for the eight selected county-level cities (Baoying, Luoping, Fuhai, Dongxiang, Dongsheng, Miyun, Longchuan, Shuangyang) is approximately 0.75. The method exhibits high precision for large-scale urban built-up area extraction without the aid of auxiliary data. The experimental results indicate that light intensity can reflect the development level of a city.