Problem
Research questions and friction points this paper is trying to address.
X-ray image enhancement
noise reduction
COVID-19 diagnosis
R2C-GAN: Restore-to-Classify Generative Adversarial Networks for blind X-ray restoration and COVID-19 classification
🤖 AI Summary
To address the challenge of jointly modeling blind-domain X-ray image degradation (e.g., blur, noise, exposure artifacts) and COVID-19 diagnosis, this paper proposes a “Restore-to-Classify” end-to-end framework that simultaneously optimizes image restoration and disease classification—without requiring paired degraded images. Built upon a dual-path GAN architecture, the method integrates perceptual loss, classification consistency regularization, and self-supervised degradation modeling to avoid error propagation inherent in cascaded pipelines and enhance discriminative feature learning. Evaluated on multiple public X-ray datasets, it achieves state-of-the-art performance in both restoration quality (improved PSNR/SSIM) and classification accuracy (+3.2%), significantly boosting clinical diagnostic robustness under low-quality imaging conditions. The core contribution lies in the first unified modeling of blind image restoration and medical classification, enabling joint structural–semantic optimization.
Innovation
Methods, ideas, or system contributions that make the work stand out.
R2C-GANs
simultaneous restoration and disease identification
X-ray blind restoration
🔎 Similar Papers
No similar papers found.
