Defect detections in industrial radiography images by a multi-scale LMMSE estimation scheme

摘要

The defect detection of welded objects is at the core of nondestructive testing and has found many applications in many industries. Industrial radiography is widely used for inspection of welded objects. In the low contrast radiography, image processing can assist to improve image quality. In this paper, an interlaced over complete wavelet expansion (OWE) and linear minimum mean square error estimation (LMMSE) are implemented to detect defects and improve the contrast of industrial radiography images. In the first stage, OWE is implemented to the radiograph and the wavelet coefficients can be calculated. Then, LMMSE can be applied to calculate the threshold level of each scale and the new wavelet coefficient. Finally, to reconstruct the output image, an inverse OWE transform will be applied. The experimental results show that the reconstructed images have higher contrast than the original radiograph in the defect regions.