Detection of thin structure defects in radiographic images using an improved Gabor filter

Defect detection has become an important part of quality control in the process of manufacturing mechanical parts and components. In this paper, a new approach is proposed for the detection of cracks in radiographic images. With this method, an improved Gabor filter is proposed and a prototype pattern is designed according to the geometric profiles of the defects to be detected. Subsequently, the prototype pattern is utilised to configure the filter to achieve a high filter response. A filter bank with eight different orientations is configured and the maximum response of the filter bank is computed as the weighted mean of the simple responses of the filters. In the newly designed filter, a parameter to control the range of responses of the filter to detect various kinds of cracks is introduced. The proposed method is tested using several radiographic images. The results show that the proposed approach is very efficient in detecting various kinds of defects and performs exceedingly well at detecting cracks.