We propose a personalized, real-time framework for generating color vision deficiency (CVD)-friendly images that enhance color discriminability while preserving natural appearance and structural fidelity. The method integrates individualized CVD simulation with a conditional GAN that learns a neutral gray-layer representation, enabling perceptually adaptive color enhancement without introducing artifacts. A soft-light blending strategy further refines high-resolution results while maintaining real-time efficiency. Experiments on Flickr30K, DIV2K, and Ishihara test images demonstrate that our approach achieves superior color contrast preservation and comparable visual quality to state-of-the-art methods, while running over 10 FPS at 768×768 resolution. The results highlight the effectiveness of the proposed graylayer-guided recoloring pipeline for accessible and perceptually consistent visualization in real-world applications.

color vision deficiency, image enhancement, generative adversarial networks