Real-Time Underwater Image Communication Guided by Depth Estimation

To address the challenges of limited bandwidth and complex channel conditions in underwater image transmission, this letter proposes a novel task-guided transmission framework that incorporates depth information and semantic constraints into the optimization process. The proposed method employs a lightweight encoder to extract task-relevant features, and introduces a staged reconstruction mechanism at the receiver to progressively restore semantic information. This design enhances the extraction efficiency of critical semantic features essential for depth estimation, thereby improving task performance under constrained transmission resources. Experimental results demonstrate that the proposed method achieves higher communication efficiency and robustness, and better task performance in measured underwater acoustic channels compared to traditional approaches.