This article is devoted to improving the efficiency of segmentation of radiographic images of welded joints. The article presents an algorithm for defect image segmentation, which is performed in two stages: determination of an array of thresholds for detecting defect image pixels (various detection thresholds for defect image pixels located in areas of digital radiographic images of welded joints with characteristic brightness distribution and background brightness estimation errors) on the background sample based on the criterion of avoiding the detection of “false” defect images, and the actual search for defect images. The experimental results confirm the applicability of the developed algorithm for effective detection of defect images in radiographic images of welded joints without the use of reference sensitivity standards.
Grigorchenko et al. (Mon,) studied this question.