Abstract: The automatic detection of weld defects based on image processing of X-ray digitized film is important in the engineering field.Ideas of respectively processing the weld edge area and the other area of weld and respectively processing the slim line defects and the other defects were proposed for automatic detection of the double sides weld defects.The outer edges and the edges of the overlapped area of double sides weld were segmented by grey waveform analysis with column by column.And then the large templates of median filter and mean filter were combined to simulate the weld background,and the weld defects were detected by segment threshold after the background being subtracted.The algorithms of adaptive image binarization with column by column and modified Hough transform were proposed to detect the slim line defects.The result shows that the proposed algorithms avoid false alarms on the edges of the overlapped area for the double sides weld and weak line defects are detected effectively.