Defect detection in X-ray images using fuzzy reasoning

Abstract Since X-ray images contain sensory noise, and objects in X-ray images are often distorted by various effects such as uneven lighting, classical image processing techniques and methods based on ordinary crisp set theory are poor at detecting small low contrast objects. In this paper, we propose a more effective method based on fuzzy theory. With the proposed algorithm, images are filtered by applying fuzzy reasoning using local image characteristics. The proposed algorithm was applied to detect internal weld defects from radiographic films, which are taken from steel butt weld parts. Results show success in detecting defects at a similar level to human vision. A comparison between a visual and an automatic evaluation demonstrates the efficiency of this method.