Abstract: The defect size of sample was obtained by different image processing of infrared image method, and the different test reaults were evaluated. Results showed that the best time of detection defect size was at around the half greatest contrast image immediately following the greatest contrast image. With the increase of defect depth, the defect size in same diameter sample was increased. On the basis of this method, the optimization precision of defect size by changing selection method of threshold value can be obtained.