Abstract: Infrared thermal wave testing technology was a kind of infrared thermal technology which heats the sample and detects the surface temperature. The infrared thermal image which had low contrast degree and high background noise was difficult to be analyzed quantitatively by inspector. In order to solve thermal image problem, the Watershed method based on mathematical morphology was presented. The method was an efficient gray level segmentation way which had strongly eliminated noise. Therefore, the Watershed method based on mathematical morphology which segmented defects and extracted the image feature was adopted. The result showed that the Watershed method had eliminated noise effectively, and had computed the defect locations and areas with quantitative analysis and had an engineering applied value.