Convolutional Transformer-Based Deblurring Model for X-Ray Images

Image deblurring is an important pre-processing for improving relevant computer vision tasks. In this paper, we are interested in conducting deblurring X-ray images. Using a convolutional transformer as the main building block, we build an AutoEncoder-style deblurring model for X-ray images. From the experiments using the public X-ray image dataset, we show that our model conducts the deblurring operation well. For example, in terms of structural similarity (SSIM) as a performance metric, our model improves SSIM by up to 27% compared to the blurry images.