[论文解读] Enhancing Underwater Images via Adaptive Semantic-aware Codebook Learning

Bosen Lin, Feng Gao+3 more|arXiv (Cornell University)|Feb 11, 2026

Image Enhancement Techniques被引用 0

一句话总结

SUCode 引入语义感知的像素级码本和三阶段训练范式，以适应区域级退化，在全参考指标上达到最先进结果，在无参考性能上具竞争力。

ABSTRACT

Underwater Image Enhancement (UIE) is an ill-posed problem where natural clean references are not available, and the degradation levels vary significantly across semantic regions. Existing UIE methods treat images with a single global model and ignore the inconsistent degradation of different scene components. This oversight leads to significant color distortions and loss of fine details in heterogeneous underwater scenes, especially where degradation varies significantly across different image regions. Therefore, we propose SUCode (Semantic-aware Underwater Codebook Network), which achieves adaptive UIE from semantic-aware discrete codebook representation. Compared with one-shot codebook-based methods, SUCode exploits semantic-aware, pixel-level codebook representation tailored to heterogeneous underwater degradation. A three-stage training paradigm is employed to represent raw underwater image features to avoid pseudo ground-truth contamination. Gated Channel Attention Module (GCAM) and Frequency-Aware Feature Fusion (FAFF) jointly integrate channel and frequency cues for faithful color restoration and texture recovery. Extensive experiments on multiple benchmarks demonstrate that SUCode achieves state-of-the-art performance, outperforming recent UIE methods on both reference and no-reference metrics. The code will be made public available at https://github.com/oucailab/SUCode.

研究动机与目标

• 通过引入语义感知的离散表示来解决水下图像的病态性和区域级退化问题。
• 学习由语义掩码引导的像素级、类别特定的码本以进行联合修复与增强。
• 通过三阶段训练范式分离码本学习、表征与增强，减轻伪真值污染。
• 提出 GCAM 和 FAFF，以在保持语义一致性的同时改善色彩保真度与纹理细节。

提出的方法

• 使用原始水下图像和语义掩码学习 C 个语义类别的语义类别特定码本 Z_c。
• Stage II 自我恢复：通过权重预测器对各类别量化特征进行加权聚合，合成统一的离散表示。
• Stage III 增强：通过 FAFF 的域自适应特征调制融合原始与增强特征，并使用带 GCAM 的双解码器进行颜色感知的修复。
• 采用双解码器架构，其中 G_q 恢复原始图像，G_r 通过一个基于 Swin Transformer 的权重预测器支持域转换。
• GCAM 重新加权颜色通道以解决水下色偏并保持颜色真实感。
• FAFF 通过实部快速傅里叶变换、相位保持、幅度调制和仿射特征调制进行频域特征融合，以在保持结构的同时转移增强纹理。

Figure 1: The comparison of the training and testing pipeline and enhance results between different codebook generation methods. The proposed SUCode’s result is sharper and clearer, with more natural color.

实验结果

研究问题

• RQ1 如何将语义信息融入离散码本学习以应对UIE中的区域特定退化？
• RQ2 三阶段训练范式在UIE中是否能在学习鲁棒表征的同时缓解对伪真值参考的依赖？
• RQ3 面向像素级且语义感知的码本是否比单 shot 或类别无关的码本能提升修复质量？
• RQ4 频域特征融合（FAFF）和 GCAM 是否能在水下图像中提供更好的色彩保真度与纹理恢复？

主要发现

Method	SUIM-E SSIM	SUIM-E PSNR	SUIM-E LPIPS	SUIM-E UCIQE	SUIM-E UIQM	UIEB SSIM	UIEB PSNR	UIEB LPIPS	UIEB UCIQE	UIEB UIQM
Fusion	0.876	16.824	0.226	58.413	2.811	0.907	18.483	0.211	52.823	3.251
IBLA	0.788	16.019	0.221	62.498	1.870	0.771	15.009	0.341	53.816	2.346
ULAP	0.860	16.574	0.232	59.746	2.174	0.902	17.871	0.233	52.620	3.309
UDCP	0.581	11.694	0.308	62.172	1.815	0.603	11.001	0.399	59.492	2.147
WaterNet	0.907	22.295	0.144	60.999	2.807	0.898	21.566	0.237	61.805	3.314
UColor	0.898	22.860	0.145	62.436	2.860	0.906	22.266	0.187	59.176	3.316
UShape	0.851	21.369	0.147	53.451	2.969	0.819	20.266	0.219	48.406	3.296
CCMSR	0.896	22.028	0.161	60.129	2.875	0.914	22.761	0.180	57.084	3.274
WfDiff	0.853	16.176	0.184	57.052	2.701	0.888	18.994	0.214	53.269	3.255
SMDR-IS	0.896	22.082	0.146	62.600	2.749	0.924	22.232	0.166	61.559	2.952
AMSIN	0.902	21.923	0.125	61.399	2.762	0.921	22.635	0.146	62.332	3.309
RUE-Net	0.923	22.902	0.121	62.500	2.776	0.923	22.743	0.164	62.357	3.260
HCLR-Net	0.902	22.317	0.124	58.765	3.360	0.902	22.317	0.124	58.599	3.279
FDCE-Net	0.923	23.039	0.141	58.765	3.360	0.923	23.039	0.141	58.765	3.360
SS-UIE	0.871	21.713	0.182	59.538	2.815	0.850	21.006	0.255	58.919	3.066
CDF-UIE	0.892	22.089	0.116	54.826	2.838	0.886	21.592	0.159	54.219	3.333
FeMaSR	0.908	22.749	0.100	62.605	2.841	0.883	22.733	0.137	62.675	3.301
AdaCode	0.886	22.329	0.105	62.409	2.812	0.818	21.792	0.156	60.835	3.216
RIDCP	0.509	13.407	0.572	42.184	2.533	0.573	14.915	0.487	48.679	2.246
IPC-Dehaze	0.823	13.869	0.381	50.837	2.252	0.852	16.923	0.226	54.777	2.352
CodeUNet	0.590	17.349	0.447	54.769	2.705	0.836	21.468	0.196	59.650	3.383
SUCode(Ours)	0.939	23.908	0.087	62.618	2.878	0.925	23.857	0.124	63.136	3.174

• SUCode 在 SUIM-E 与 UIEB 数据集上实现了全参考指标（PSNR、SSIM、LPIPS）的最先进性能。
• SUCode 在无参考指标（UCIQE、UIQM）上表现竞争力，尤其在 SUIM-E 与 UIEB 上达到最高的 UCIQE。
• 跨数据集评估显示出较强的泛化能力，在以 UIEB 训练、在 LSUI 与 UFO-120 上测试时超越了若干基线。
• 语义感知的码本相比非语义码本方法能带来更清晰、自然的色彩恢复与更好的纹理保持。
• 三阶段训练策略有效处理病态的伪真值问题，并使增强与语义内容对齐。

Figure 2: The overall structure of the proposed SUCode. In stage I, the semantic-aware category‑specific codebooks are updated with the mask $m$ . Stage II is a partition and synthesis process of the codebook, achieved through the self-reconstruction of raw underwater images. In stage III, domain co