Underwater images are degraded by scattering and absorption in the water, often resulting in color distortion and loss of detail. To ameliorate these issues, this paper proposes a multi scale underwater image enhancement method based on adaptively fusing gradient and color information. First, during texture detail enhancement, multi scale gradients and high frequency components are fully integrated to restore texture details. Second, underwater light‐field information is incorporated as a global color reference to enhance color restoration. A hierarchical feature extraction structure is employed to comprehensively capture latent features from the original underwater images, while residual blocks and attention mechanisms embedded at critical points of the network to strengthen its feature modeling capability. Finally, an adaptive fusion mechanism integrates these multi dimensional cues to comprehensive global enhancement of underwater images. Experimental results demonstrate that, compared with other state of the art methods, the proposed method delivers superior improvements in underwater image quality.