The paper reports defect detection techniques using X-rays in order to find three dimensional defects such as funnels in printed circuit board (PCB) patterns. X-ray images of fine PCB patterns, however, show patterns on more than one layer, having degrees of distortion that vary over an image, are subject to intensity variation due to fluctuations in the intensity of the X-ray source, and so on. To overcome these problems, we examined distortion caused by perspective transforms on the sphere surface of an X-ray detector and developed a defect detection algorithm based on feature extraction. Fluctuation of X-ray strength and heavy shading of the acquired image are compensated by signal processing. Evaluation of the system confirmed that the system could detect defects deeper than 20 /spl mu/m and with surface areas greater than 70/spl times/35 /spl mu/m/sup 2/ in fine PCB patterns and that the inspection time for a 600/spl times/800 mm/sup 2/ PCB was only 20 minutes.>