Unsupervised Underwater Image Enhancement Method Based on Data-Driven and Atmospheric Light Optimization

In response to the difficulty of obtaining paired datasets in existing underwater image enhancement methods., an unsupervised underwater image enhancement method based on data-driven optimization and atmospheric light adjustment is proposed. The degradation module., utilizing a parallel network., extracted latent components of the original image and utilized an improved Koschmieder model to recombine and generate the degraded image. The data-driven optimization module compared the original and re-degraded images to extract perceptual transmissions of different regions., incorporating dual-branch gating units and channel attention mechanisms. Finally., the atmospheric light optimization module processed global information in the frequency domain to offset the influence of atmospheric light on underwater images., enhancing image quality. Experimental results demonstrate improvements of 3.3%., on evaluation metrics (UCIQE., UIQM., and NIQE) on the UIEB datasets., respectively. Presently., our method exhibits significant advantages in visual quality.