Application of Different Median Filter Algorithms for Welding Defects Clarification in Radiographic Images

Abstract

The main aim of the inspections is to ensure that the quality of the welds meets the design requirements and operating conditions, as well as meeting safety and reliability requirements in many industrial sectors. There are many different ways of non-destructive testing NDT are employed for checking welds. The radiographic testing RT is one of the most prevalent and widely used processes in the industry to detect both surface and subsurface defects. In recent years, industrial radiography has been well developed to be used alongside with the combination of techniques of image processing to enhance both the process of inspection and the time of operation. The RT images are often affected by some external surroundings and degraded by noises during the process of acquisition of them. The filtering process is intended to recover the original image from the corrupted one. Median filtering can be considered as an appropriate procedure since it is used by many researchers. In this paper, a comparison among four types of median filter is used to get the best one, to adopting it in welding inspection images. The peak signal to noise ratio (PSNR) and root means square error (RMSE) indexes are used to perform the quality of image filtering techniques. The results shown that the median filter is a good tool for clarification of welding defects and both adaptive median filter (AMF) and decision based median filter (DBMF) gives the best results respectively. The results obtained were consistent with the results of the engineering examination for qualified and certified welding defects inspectors.