XTC-DDPM: Diffusion Model for Tone Mapping and Pseudo-Color Imaging of Dual-Energy X-ray Security Inspection Images

Da Cai, Tong Jia, Bowen Ma, Hao Wang, Mingyuan Li, Dongyue Chen

Abstract: Dual-energy X-ray security inspection imaging technology plays a crucial role in ensuring public safety, focusing on two tasks: tone mapping and pseudo-color imaging. This paper introduces XTC-DDPM, a diffusion model tailored for dual-energy X-ray security inspection tone mapping of grayscale images and pseudo-color imaging. The objective is to enhance image quality and facilitate accurate item identification by security personnel. The paper begins by constructing a dual-energy X-ray security inspection imaging dataset, named TCX-ray. Subsequently, our proposed method incorporates a multi-source blur module based on the distribution of high-frequency information in the image. This module is applied to 16-bit X-ray security inspection raw images. During the multi-channel forward diffusion process of the diffusion model, noise is gradually introduced to the multi-channel target image until the image reaches a state of pure noise. The dual-energy 16-bit X-ray security inspection image, its blurred counterpart and the noise image at each time period are combined. A dual-branch feature extraction denoising network is employed to predict the noise, followed by a reverse denoising process. This results in obtaining high and low energy 8-bit X-ray security inspection grayscale images, as well as pseudo-color images. Finally, the tone-mapped high and low energy X-ray images are fused to produce the grayscale output, while the pseudo-color imagery serves as an additional output. The proposed method is validated through both qualitative and quantitative experiments. The extensive experimental results demonstrate the superiority of our approach compared to existing tone mapping and pseudo-color imaging methods for X-ray grayscale images.