Welded Pipe Defect Detection Enhancement Using Regularized Kernel Estimation-Based Image Processing in Radiographic Testing

Abstract

Confirmation of the integrity of welded pipes is of critical importance in the oil and gas industry and radiography testing (RT) is used as the gold standard for non-destructive testing (NDT) inspection. Radiography is based on the penetration of  X-rays or gamma-rays through the object and the measurement of the intensity of the traversing radiation. Unfortunately, intrinsic to the measurement is the undesired detection of scattered radiation which constitutes the imaging noise and manifests in image blurriness. Seamless detection of the presence and evaluation of weld defects necessities high testing sensitivity which requires the selective removal of the blurring and enhancement of the imaging contrast. Software-based removal of the blurriness is achievable by estimation of a “blur kernel” and data processing using various methods. Two such methods that are widely used include the “fast ${{\ell }_{0}}$” regularized kernel estimation (${{\ell }_{0}}$-RKE) and a modification of the Goldstein–Fattal method (MGFM). In the study reported here, the ${{\ell }_{0}}$-RKE and the MGFM methods were implemented and applied to the radiographs of welded pipes. Improvement of defect region and weld root evaluation was achieved using these methods through the sharpening of the radiograph feature edges/detail. Comparison with the unaided evaluation of the original blurred radiographs by expert operators confirmed the improved visualization for both algorithms. Although the ${{\ell }_{0}}$-RKE algorithm enhanced the image to a greater extent than the MGFM method, the MGFM algorithm was much faster with run-times of almost half of those for the ${{\ell }_{0}}$-RKE algorithm.