Image recoloring for multiple types of Color Vision Deficiency

There are many techniques for recoloring images with different effects and improving color discrimination in patients with color vision defects. However, certain issues still persist, such as the unnatural and discordant colors of objects in the converted image. To address these problems, we have explored a comprehensive set of methods to achieve image recoloration. Our approach enables the resulting images to possess three essential characteristics: naturalness, harmonization, and distinguishability, thereby fulfilling the requirements of Color Vision Deficiency individuals. The method comprises two components: recommended palette generation and image recoloring. The former can learn the color distribution of different objects in nature, while the latter can recolor the image in conjunction with the recommended palette. Our experimental findings demonstrate that our approach is feasible and provides a direction for future research.