Radiographic image enhancement based on a triple constraint U-Net network

Abstract

Radiographic testing (RT) images of complex components are affected by several factors, including low greyscale levels, low contrast and blur. These factors can significantly restrict the accuracy and effectiveness of defect recognition. To address this issue, this paper proposes a radiographic image enhancement method based on a triple constraint U-Net network. Firstly, a radiographic image preprocessing target dataset is constructed based on conventional image preprocessing technology and previous experience. The U-Net model is then used to design a model loss function, including the parameters of image consistency, texture consistency and structural similarity, in order to achieve structure preservation and noise removal in the images. Finally, radiographic images of actual complex components are used to illustrate and verify the effectiveness of the proposed method. The results show that the proposed method can effectively convert an original image to a target image, enhance the details of the defect area and improve the accuracy of defect recognition by 5.2%.