ABSTRACT A broad range of defects can be formed in metallic equipment . The present study adopted active infrared thermography with optical excitation sources to detect the artificial flat-bottom holes in a steel sample. Raw thermal data always have some problems, such as noise and blurred edge. A proposed method, namely TPPT, was adopted to eliminate these problems. In addition to the developed technique, several well-known processes such as PCT were employed. The Contrast Enhancement parameter (CE) was defined to compare preprocessed images. CE data indicate that the TPPT has the best performance in improving the raw data. A post-processing procedure was also proposed to improve the segmentation results. This strategy consisted of two steps, histogram equalisation of preprocessed images and applying 2D wavelet noise-reduction. Ultimately, several segmentation methods were applied to the post-processed images. The outstanding performance of TPPT was proved one more time as the TPPT segmented images revealed the highest number of defects, almost 14 out of 20, on average. Moreover, the performance of segmentation algorithms was compared together regarding the number of detectable defects and estimation of defective regions’ area. FCM was recognised as the most appropriate segmentation algorithm for this study based on the obtained results.