Abstract: To realize X-ray digital image computerized analysis and processing, a defect-extraction automation technique is studied. An image processing method based on subarea and self-adaptive median filtering is introduced with the evaluation of turbine blade. In the light of gray level variety in the different subareas of X-ray digital image, a kind of scan-direction-based one-dimensional median filtering method, in which the length of filter can be adjusted automatically with the size of defects being extracted, is used. Defects in the inspected workpiece can be extracted rapidly and accurately with this method. The experimental results indicate that, with this image processing method, defect distortion can be avoided, which will lay a good foundation for defect-recognition automation in industrial digital radiography.

Key words: nondestructive testing, defect extraction, self-adaptive filtering, digital radiography, image processing