Retinex Theory Based Automated Contrast Enhancement of Gamma Radiographic Images of Pipe Welds

Gamma radiography has found widespread use for the inspection of welded pipelines enabling the detection of a various type of defects, including surface, subsurface, and deeper flaws. However, Radiography Testing, RT, images are susceptible to poor image quality with several factors contributing to the blurring of radiographs and necessitating image-enhancing methods. In this study, a Retinex-based (RB) enhancement technique was implemented and applied to enhance the contrast of twenty-three radiographic images acquired using gamma and film. The outcomes were also compared with those from Total Variation (TV) and Non-Local Operators (NO) based algorithms. It was found that the application of the RB algorithm to RT radiographs obtained by Gamma-ray significantly improves the visualization of defect areas, the weld root, and the Image Quality Indicator (IQI) region. Comparing the RB approach to the methods based on TV and NO algorithms, it was found that the RB and NO methods yielded similar results in identifying different regions within the radiographs and the TV method did not enhance the IQI, nor the visualization in the slug regions. The outcome of the evaluation by RT radiographers indicated that the RB method was more effective in increasing the image contrast than both the TV and NO methods, allowing for better visibility of various weld regions. Given the single parameter and hence the shorter optimization requirement for the RB algorithm, the overall time for RT image enhancement using the RB approach was shorter. Furthermore, the RB approach is expected to be more suitable for automation of the image enhancement process, lending itself to the potential of AI (Artificial Intelligence)-based analysis.