The application of satellite-sensed thermal infrared imagery to the detection of changes in temperature before outbreaks of earthquake is faced with two problems: the interference with image by clouds and the scientific record of temperature-increase phenomena. This study has solved the problems through using spatial-temporal interpolation to make inference for zones with information on abnormal temperature increases negatively affected by clouds as well as mosaic of temperature-increase images sensed at different time points so as to recover images free of clouds. In addition to the methods for eliminating the interference by clouds, this study specifies normalized standard values and threshold numbers so as to use computerized semi-automatic detection to identify zones with relatively high temperatures, a scientific approach to facilitating automatic coloring of such zones and mosaic of images. The methods for processing images, deleting negative influence of clouds and scientifically detecting pre-earthquake abnormal increases in temperature as described in this paper are intended to serve as reference for other researchers to improve precision in subsequent studies.