Color Enhancement for the Colorblind Using Color Correction Intensity Map and Pix2pix Image Conversion

In this paper, we propose a method for creating easy-to-see images by estimating regions where color discrimination is difficult for people with color blindness and performing color correction. There are various prior studies on color correction of images for the colorblind. However, in many cases, the goal is only to identify text or objects, and thus the hue of the image is often unnatural compared to the original image. To solve this problem, we estimated the color discrimination gap between colorblind and normal subjects based on confusion loci theory, and performed color correction using the pix2pix model. The aim is to prevent unnecessary hue changes by estimating the correction intensity, and to achieve highly accurate color correction by deep learning. From the validation using a colorblindness simulator, we obtained the result that color discrimination becomes easier for colorblind people while maintaining the hue close to that of the original image. This method allows easy color correction of existing images for the colorblind without any loss of hue.