Abstract: A practical detection method,which takes advantage of the information contained in the image itself,is presented based on local treatment gradually for attacking the difficulties in accurately detecting and extracting weld seam and defect in ray testing image.First of all,the approximate location of the weld zone in the original image is determined by detecting the regional characteristics of bands;then,the local area where the weld boundaries and defect are in is determined by analyzing gray column waveforms,and the area types are divided into the categories of non-crack and crack class roughly;finally,the detection and extraction of welding defects are realized by using the methods based on Watershed Transform and Beamlet Transform respectively.The results show that the method can accurately detect the weld boundaries and weld defect in various ray images and has good adaptability and practicality.