Innocolor: a Pioneering Image Recoloring Framework Driven by Salient Attention Networks

Image recoloring technology plays a crucial role in enhancing color contrast perception for individuals with Color Vision Deficiency (CVD). Current recoloring algorithms primarily focus on improving contrast, preserving color naturalness, and minimizing computational costs. However, enabling CVD individuals to perceive visual saliency similarly to those with normal vision has become a key research challenge. In this paper, a real-time recoloring framework, InnoColor driven by saliency detection is proposed. InnoColor performs preliminary recoloring using learnable 3D-LUTs, followed by precise regression of the optimal hue rotation angle through a novel framework, RA-NET. Finally, hue shifting is applied to the salient regions in the image. Experimental results demonstrate that the proposed method has achieved state-of-the-art improvements in both salient consistency, with a Dice score of 0.89 compared to DALT-NET's score of 0.86, and in computational efficiency. InnoColor has a significant impact in initiating newly optimized, accessible, and salient pathways for CVD users in the growing digital network.