Radiography testing (RT) is a well-established non-destructive method for the detection and measurement of defects in the welded objects. Unfortunately, however, the image quality of radiographs is often low mainly due to the unavoidable detection of scattered X-rays. This superimposed noise on the images could result in low detectability and inaccurate dimensional measurement of such defects. Application of image processing solutions has found wide application since hardware-based removal of the noise is often impractical. The Chambolle algorithm is a method for minimizing the total statistical variance within an image which is fast and efficient and is based on a projection-based algorithm. The enhanced images processed through the Chambolle method have significantly improved contrasts and yield higher accuracy for defect dimensional measurements when compared to original images. In this study, the Chambolle algorithm was used to increase the contrast in the defect regions of the radiographs of ‘Sonaspection kit’ objects where contrast enhancements by up to a factor of about 3.5 were obtained. Also, dimensions of standard defects from the ‘Sonaspection kit” were measured to an accuracy of better than 5% using images from the processed images compared to inaccuracies of about 15% when original radiographs of the same defect were analyzed.

射线照相检测 (RT) 是一种成熟的无损检测方法，用于检测和测量焊接物体中的缺陷。然而不幸的是，由于不可避免地检测到散射的 X 射线，射线照片的图像质量通常较低。这种叠加在图像上的噪声会导致此类缺陷的可检测性低和尺寸测量不准确。由于基于硬件的噪声去除通常是不切实际的，因此图像处理解决方案的应用得到了广泛的应用。Chambolle 算法是一种用于最小化图像内总统计方差的方法，它基于投影算法，快速有效。与原始图像相比，通过 Chambolle 方法处理的增强图像显着提高了对比度，并为缺陷尺寸测量提供了更高的精度。在这项研究中，Chambolle 算法用于增加“Sonaspection kit”物体射线照片缺陷区域的对比度，其中对比度增强高达约 3.5 倍。此外，“Sonaspection 套件”中标准缺陷的尺寸使用来自处理图像的图像进行测量，准确度优于 5%，而分析相同缺陷的原始射线照片时的不准确度约为 15%。Chambolle 算法用于增加“Sonaspection kit”物体射线照片缺陷区域的对比度，其中对比度增强高达约 3.5 倍。此外，“Sonaspection 套件”中标准缺陷的尺寸使用来自处理图像的图像进行测量，准确度优于 5%，而分析相同缺陷的原始射线照片时的不准确度约为 15%。Chambolle 算法用于增加“Sonaspection kit”物体射线照片缺陷区域的对比度，其中对比度增强高达约 3.5 倍。此外，“Sonaspection 套件”中标准缺陷的尺寸使用来自处理图像的图像进行测量，准确度优于 5%，而分析相同缺陷的原始射线照片时的不准确度约为 15%。