Abstract: The slim line defect whose average width is not more than three pixels in X-ray weld images is usually weak and interrupted. Normal X-ray defects detection algorithm can not detect it effectively. A new method is proposed to detect this kind of defects. After filter the original X-ray image to reduce the noise, the self-adaptive threshold value is adopted line by line to segment the slim line defects in the image preliminarily, and the improved partial Hough Transform is used to eliminate the noise and segment the line defect. The experimental results show that the proposed method can adapt to X-ray digitized images of different grey range and avoid the influence of much noise and interruption, and is effective to detect the low-contrast slim line defects.