The primary work to recognize butt weld defects automatically with X-ray radiographs is to achieve effective assessment area accurately. A specific image processing design was proposed to overcome extracting difficulties by combining X-ray imaging mechanism and weld gray features. Firstly, it conducts gray-scale transformation of the butt weld image based on X-ray imaging fundaments. Secondly, it corrects the offset of weld bead based on the regional gray variance. Thirdly, it adopts Least-squares method to fit the weld seam and determine its center. Lastly, it converts weld seam from annular to rectangular using the distance map. The results of 3 000 samples showed that the design can detect the center of weld seam and fit it with 100% success rate. The fitting ellipses showed that the weld seam deflection angle error is -0.216°±0.826°. Therefore, it is feasible to extract the effective weld seam area from low contrast digital X-ray image for future assessment.
