Abstract: Based on the analysis of mathematical concept and geometric transformation .the author consider that the decorrelation stretch can selectively exaggerate the least correlated part of thermal multispectral image information,and enhance the spectral contrast of different ground materials. It is one of the effective methods to enhance the highly correlated images. The information of a thermal infrared multispectral image mainly include two parts:surface temperature and emissivity. In principle .the surface temperature is the dominant information. For the temperature of terrestrial surface is commonly less than 30℃.the thermal infrared multispectral image data are always highly correlated from one channel to the next. It is unfavourable for the extraction and analysis of the thematically geological information. The decorrelation stretch,which tend to increase the saturation and affect the hues little,is usful in exaggerating the poorly correlated emissivity information of thermal infrared rnultispectral image data,rather than the highly correlated surface temperature information. The paper presents the effectiveness of decorrelation stretch image with an example in extracting lithostratigraphical information by use of the information about the composition ( i. e.emissivity),rather than the surface temperature.