Underwater image enhancement model based on GAN with enhanced generative capability

Underwater images are often plagued by issues like color distortion, exposure, insufficient contrast, and severe haze, which greatly limits their application value in ocean exploration tasks. In recent times, a great many underwater image enhancement approaches have drawbacks when it comes to detail improvement and color adjustment.To overcome these challenges, this study proposes propose underwater image enhancement model based on GAN with enhanced generative capability, named GEGC, for enhancing the details and correcting color of underwater image. Concretely, GEGC adopts multi-stage feature enhancement in its generator：After U-Net encoding, the image is calibrated for channel and spatial features by TripletAttention，then, through the designed Color Focused Detail Enhancement Module named CFDE, feature channel weight optimization is implemented to achieve color correction,combining two convolutions and residual connections to finely mine and enhance high-frequency details, gradually generating high fidelity images.This process forms an adversarial loop with the discriminator, continuously improving the color accuracy and detail richness of the generated image. Numerous experiments conducted on openly accessible underwater image datasets have demonstrated that the proposed method exhibits significant effectiveness, superiority, and competitiveness compared to representative image enhancement methods, which has been fully demonstrated through qualitative and quantitative evaluations.