Underwater image re-enhancement with Simplest Color Balance, CLAHE Algorithm, De-noising Convolutional Neural Network and Unsharp Masking

Restoring Underwater images is quiet challenging task. In an attempt to resolve this, a proposal of fusion algorithm that consists of a Simple Color Balancing algorithm, Contrast Limited Adaptive Histogram Equalization, De-noising Convolutional Neural Network, and Unsharp Masking. To minimize the impact of color cast in underwater images, use of a simple color balance algorithm to stretch the r, g, and b values to their maximum range [0,255] to get a similar distribution of rgb channels in the histogram. To improve the perceptibility of underwater images, increase of the contrast of underwater images using the CLAHE algorithm. De-noising Convolutional Neural Network is used to remove the noise present in underwater images. An unsharp masking approach is implemented to eliminate the blurriness in output denoising images and sharpen the output images. PSNR values show that the resulting images of the proposed system have the best image quality. Some gave the best results based on their SSIM values compared to the MSRCR model. The UCIQE values of the resulting images dominate the results of other models. The efficiency of the fusion algorithm is proved by the results obtained and then compared with the results of current models.