Underwater Image Enhancement via Triple-Branch Dense Block and Generative Adversarial Network

The complex underwater environment and light scattering effect lead to severe degradation problems in images, such as color distortion, noise interference, loss of details. However, the images bring a significant challenge applications. To address detail we propose triple-branch dense block-based generative adversarial network (TDGAN) for quality enhancement images. A residual block is designed generator, which improves performance feature extraction efficiency retains more image dual-branch discriminator also developed, helps capture high-frequency information guides generator use global content detailed features. Experimental results show that TDGAN competitive than many advanced methods from perspective visual perception quantitative metrics. Many application tests illustrate can significantly improve accuracy target detection, it applicable segmentation saliency detection.