In many important industrial machines, such as nuclear power equipment and steam turbines, the weld quality affects the overall safe operation of the equipment. Although many artificial intelligence (AI)-based technologies are applied to defect recognition using radiographic images,
at present the final decision is still made based on human visual inspection. Therefore, it is necessary to enhance the radiographic images to aid the inspection process. In this paper, a method for radiographic testing (RT) weld image enhancement based on phase symmetry, which is based on
the principles of human vision, is proposed. Phase symmetry does not depend on the greyscale and contrast information of an image, so it is suitable for RT images with low greyscale values and low contrast. To evaluate the proposed method, 260 RT images acquired by a professional radiographic
film digitiser (JD-RDT) and general international RT weld images are used. The results demonstrate that the proposed method can enhance both weld seam and marking information (numbers, letters and symbols), this being suitable for human visual inspection. The proposed method is compared with
commonly used methods and the comparison results show that the proposed method achieves an improvement over state-of-the-art methods.