Interlaced bilateral filtering and wavelet thresholding for flaw detection in the radiography of weldments

Abstract

The recognition of damages and defects is very important in welded joints radiographic images. The experts’ evaluations of the weld images are dependent on both image quality and interpreter’s experience. This evaluation will become difficult if the radiography images have low quality and have been corrupted with noise and other quality decreasing factors. Therefore, some image processing methods are required to improve the radiography images’ quality. Here, an interlaced multistage bilateral filtering and wavelet thresholding have been implemented to radiography images of welded objects for better detection of weld flaws, which can effectively obviate the defect regions in real noisy radiography images. The line profile method has been implemented for the evaluation of the results, i.e., the quality of the radiographs before and after applying the image processing algorithm. The line profiles were compared in the image quality indicator region. The comparison of the line profiles shows that the quality of the reconstructed radiographs, which is measured here by the contrast to background level, enhances almost by a factor of two for the subtracted images in comparison with the original radiographs.