X-ray image denoising for Cargo Dual Energy Inspection System
Bilel Yagoub, Hatem Ibrahem, Ahmed Salem, Jae‐Won Suh, Hyun-Su Kang · 2021 International Conference on Electronics, Information, and Communication (ICEIC) · 2021
The current instruments for cargo inspection systems use dual energy system e.g. 9/6 MeV or 6/3 MeV for material discrimination, the low energy transmission system introduces a special kind of noise, a random vertical lines noise, this noise may cause severe problems in material discrimination. We propose a novel convolutional neural network-based technique for 3 MeV x-ray image denoising and we also introduce a mathematical model to artificially generate this kind of noise, due to the lack of a large x-ray cargo images. We trained our model on artificially generated images with the simulated random vertical noise, while we tested the model on both of artificially generated images and a real image generated by an x-ray dual energy system. The system showed a competitive peak-to-noise ratio PSNR of 41.2 and a structure similarity index of 0.956.