Color Compensation Based Enhancement of Dark Underwater Images

The intensity of light is very low in the deep sea and hence the captured images appear dark. This article introduces a new approach for enhancing the weakly illuminated underwater deep sea images in two stages. The first stage decomposes the image using Retinex Theory into reflectance and luminance. The reflectance is enhanced by utilizing the significance of contrast limited adaptive histogram equalization (CLAHE) and the arctangent function by employing the multiscale image fusion technique. The enhanced reflectance is then multiplied by the luminance to remove the darkness. In the second stage, the color mixing method is utilized for color balancing to remove the color casts. The experimental results illustrate that the approach proposed in this work outperforms the existing methods in terms of underwater color image quality evaluation metric (UCIQE) with an average score of 0.6207.