PIC-UIE: Predicting Image-Adaptive Corrections for Lightweight Underwater Image Enhancement

Cunhao Zhu; Dongliang Xu; Xiangtao Kong; Xiaoyan Lu; Tianyu Wang; Yue Yao

cs.CVSep 27, 2026

PIC-UIE: Predicting Image-Adaptive Corrections for Lightweight Underwater Image Enhancement

Authors: Cunhao Zhu, Dongliang Xu, Xiangtao Kong, Xiaoyan Lu, Tianyu Wang, Yue Yao

Organizations: Shandong University · School of Airspace Science and Engineering, Shandong University · The Hong Kong Polytechnic University · Shandong Tongyu Network Security Technology Co., Ltd. · Mohamed bin Zayed University of Artificial Intelligence

Abstract

Underwater image enhancement (UIE) aims to restore visibility, color fidelity, and structural detail from images degraded by wavelength-dependent attenuation and backscatter. State-of-the-art UIE methods often rely on large backbones and dense image-to-image prediction, limiting their practicality for edge deployment. Moreover, operating entirely in a single color space couples degradation estimation with luminance and chroma correction. To address these challenges, we propose PIC-UIE, a lightweight predictor--executor framework that predicts image-adaptive corrections from a fixed $256\times256$ RGB thumbnail and applies them to the native-resolution input in the YCbCr color space. The predictor produces seven outputs, organized into spatial correction, nonlinear luminance and coupled chroma mapping, and image-level color calibration. A depth map regularizes the transmission proxy during training, whereas inference uses only the RGB input. With 9,486 parameters and 0.094 GFLOPs at $256\times256$ , PIC-UIE achieves 24.137 dB PSNR and 0.9216 SSIM on UIEB-90 and 21.320 dB PSNR on zero-shot LSUI. It further processes native 4K images at 55.0 FPS under the comparison protocol. These results show that structured correction prediction provides an effective and practical alternative to dense RGB reconstruction for underwater image enhancement.

Figures & tables

Figure 1: Quality–efficiency comparison on UIEB-90. Bubble area indicates parameter count, color indicates FLOPs, and the red marker denotes PIC-UIE.

Figure 2: Overview of PIC-UIE. The predictor estimates seven corrections from a fixed $256\times 256$ thumbnail, and the YCbCr executor applies them at native resolution. Depth supervises $t$ only during training.

Figure 3: Qualitative comparison on UIEB-90 and zero-shot LSUI. Columns follow the method order in Table 1 ; red boxes identify enlarged regions used to compare color correction, local contrast, and structural-detail preservation.

Method	Venue	Params	FLOPs (G)	Time (ms)	UIEB-90					LSUI zero-shot
Method	Venue	Params	FLOPs (G)	Time (ms)	PSNR $\uparrow$	SSIM $\uparrow$	MSE $\downarrow$	LPIPS $\downarrow$	PCQI $\uparrow$	PSNR $\uparrow$	SSIM $\uparrow$	MSE $\downarrow$	LPIPS $\downarrow$	PCQI $\uparrow$
FUnIE-GAN	RA-L’20	7.02M	7.16	3.11	18.969	0.7834	0.01506	0.2302	0.7472	18.058	0.7583	0.01763	0.2917	0.7684
Shallow-UWnet	AAAI’21	219K	43.3	2.01	17.279	0.7237	0.02383	0.2885	0.5212	20.311	0.8005	0.01344	0.2794	0.5961
PUIE-Net	ECCV’22	1.9M	935.2	194.3	21.695	0.8941	0.00885	0.1259	0.6493	20.995	0.8249	0.01005	0.2306	0.7076
NU 2 Net	AAAI’23	3.15M	20.8	10.68	23.202	0.9019	0.00644	0.1244	0.8164	21.497	0.8261	0.01007	0.2397	0.7216
DM-underwater	MM’23	18.3M	1338	132.4	21.728	0.9047	0.00928	0.1254	0.8774	18.157	0.7757	0.02096	0.2765	0.7310

Table 1: Full-reference results on UIEB-90 and zero-shot LSUI; dark and light green cells denote the best and second-best results. Baseline sources are ( Islam et al. 2020 ; Naik et al. 2021 ; Fu et al. 2022 ; Guo et al. 2023 ; Tang et al. 2023 ; Zhang et al. 2024b ; Zhang et al. 2024a ; Zhou et al. 2024 ; Guan et al. 2024 ; Zhao et al. 2024 ; Yang et al. 2024 ; Peng and Bian 2025 ; Lin et al. 2026 ) .

Method	U45 zero-shot				C60 zero-shot
Method	UIQM $\uparrow$	UCIQE $\uparrow$	NIQE $\downarrow$	Entropy $\uparrow$	UIQM $\uparrow$	UCIQE $\uparrow$	NIQE $\downarrow$	Entropy $\uparrow$
FUnIE-GAN	3.247	0.3057	3.859	7.541	3.143	0.2847	5.313	7.167
Shallow-UWnet	2.979	0.2334	4.722	6.880	2.433	0.2237	6.199	6.525
PUIE-Net	3.172	0.2744	4.504	7.323	2.630	0.2651	5.789	7.086
NU 2 Net	3.238	0.3040	4.311	7.526	2.808	0.2946	5.727	7.272
DM-underwater	3.094	0.3229	4.274	7.513	2.675	0.3051	6.936	7.218

Table 2: No-reference quality assessment on U45 and C60 under zero-shot transfer at

256\times 256

. Higher UIQM, UCIQE, and entropy and lower NIQE indicate better scores; dark and light green cells denote the best and second-best values in each column, respectively.

Method	Native 1080P				Native 4K				FLSea		URPC2019
Method	GFLOPs	FPS $\uparrow$	UIQM $\uparrow$	UCIQE $\uparrow$	GFLOPs	FPS $\uparrow$	UIQM $\uparrow$	UCIQE $\uparrow$	Matches $\uparrow$	Inliers $\uparrow$	mAP@50 $\uparrow$
FUnIE-GAN	228.4	77.6	2.973	0.2503	913.4	17.5	3.043	0.2630	1438	1359	0.723
Shallow-UWnet	1369	24.4	2.076	0.2209	5475	5.8	2.345	0.2335	665	630	0.725
PUIE-Net	29589	0.6	2.419	0.2466	118356	0.15	2.613	0.2779	1791	1706	0.745
NU 2 Net	662.0	29.7	2.524	0.2729	2648	7.4	2.657	0.2756	2097	2007	0.766
DM-underwater	–	–	–	–	–	–	–	–	574	544	0.742

Table 3: Native-resolution efficiency, no-reference quality, and downstream task performance. GFLOPs and FPS use batch-one FP32 on an RTX 4090; FLSea reports SIFT matches and RANSAC inliers, and URPC2019 reports YOLO11n mAP@50. Dark and light green cells denote the best and second-best available results.

Figure 4: Native 4K qualitative comparison on UIQAD. Marked regions are shown as enlarged crops to compare color recovery, local contrast, artifact suppression, and fine-detail preservation across methods.

Hardware	Type / Memory	1080P FPS $\uparrow$	4K FPS $\uparrow$
Ryzen 7 4800H	CPU	2.6	0.5
GTX 750 Ti	GPU / 2 GB	9.1	2.2
Tesla P4	GPU / 8 GB	23.7	6.0
Jetson Orin NX	Edge GPU / 8 GB	29.5	5.9
RTX 2060	GPU / 6 GB	41.0	10.7
Tesla T4	GPU / 16 GB	51.7	13.3

Table 4: Cross-hardware FP32 throughput.

Output parameterization	Params	UIEB PSNR
Direct RGB residual	9,427	20.940
Direct YCbCr residual	9,427	20.880
Structured corrections	9,486	24.137

Table 5: Matched-capacity output parameterizations.

Figure 5: Predictor-capacity scan on UIEB and LSUI.

Variant	Params	UIEB	LSUI
Predictor: YCbCr input	9,486	23.837	21.178
Predictor: RGB+YCbCr	9,918	23.766	21.241
Chroma: two 1D curves	7,276	23.869	21.239
Chroma: global affine	6,834	23.721	21.260
w/o RGB statistics	9,374	23.586	21.064
Statistics: concatenation	10,670	23.781	21.196

Table 6: Representation choices (PSNR).

Variant	UIEB PSNR	LSUI PSNR
w/o physical luminance	23.700	21.259
w/o tone curve	23.745	21.219
LUT $\rightarrow$ affine	23.721	21.260
w/o CCM	23.407	21.020
w/o covariance calibration	23.227	20.864
w/o transmission gate	23.697	21.242

Table 7: Leave-one-out ablation (PSNR).

Figure 6: Depth-consistency diagnostic on UIEB-90.

Explore similar work

RPL-UIE: Reliable Prior Learning for Underwater Image Enhancement
Jul 31, 2026Yifan Chen, Jiaming Liu, Ye Zheng +2Underwater Image Enhancement Methods Visual Priors
College of Future Information Technology, Fudan University, Shanghai 200433, China · Institute of Artificial Intelligence (TeleAI), China Telecom, China · Faculty of Robot Science and Engineering, Northeastern University, Shenyang 110819, China +1
DY-LUT: Depth-Aware YCbCr Lookup Tables for Real-Time Underwater Image Enhancement
Jul 24, 2026Cunhao Zhu, Xiangtao Kong, Dongliang Xu +3Underwater Image Enhancement Methods Visual Degradation
1Shandong University · 2The Hong Kong Polytechnic University · 3Mohamed bin Zayed University of Artificial Intelligence
CoRe-UIE: Rethinking Coexisting and Region-wise Degradation for Underwater Image Enhancement
Aug 10, 2026Weifeng Kong, Chenghao Xu, Lin Chen +2Underwater Image Enhancement Methods Query-Specific Degradation-Aware Representation
Hohai University Jiangsu, China

Jul 31, 2026, eess.IVJ/K move · Enter open · S save

RPL-UIE: Reliable Prior Learning for Underwater Image Enhancement

Yifan Chen, Jiaming Liu, Ye Zheng, Zhe Sun+1

College of Future Information Technology, Fudan University, Shanghai 200433, China · Institute of Artificial Intelligence (TeleAI), China Telecom, China · Faculty of Robot Science and Engineering, Northeastern University, Shenyang 110819, China+1

Underwater Image Enhancement Methods Visual Priors U-Net

Open paper AskPDF arXiv

Underwater image enhancement (UIE) aims to recover clear images from observations affected by wavelength-dependent absorption, scattering, and spatially nonuniform degradation. Although existing generative methods can handle complex degradations, severe information loss may lead to semantic drift in the restored results. To address this issue, we propose RPL-UIE, a two-stage teacher--student framework for reliable prior learning. In the teacher stage, the network learns reliable and complementary spatial priors characterizing appearance and photometric properties from paired degraded and reference images. In the student stage, the network takes only degraded images as input and learns to emulate the teacher's prior extraction capability, thereby providing more reliable restoration guidance for the enhancement process without requiring reference images at inference. To reduce the prior-learning discrepancy between the teacher and student models, we further develop Residual Prior Refinement Diffusion (RPRD) and Frequency-Aware Prior Residual Calibration (FPRC). RPRD uses the coarse priors as anchors and progressively predicts the necessary corrections in the residual space. FPRC retains stable low-frequency residual components and selectively modulates high-frequency detail residuals, producing calibrated priors to support high-quality reconstruction. Experiments on multiple UIE benchmarks demonstrate competitive restoration performance. Downstream underwater object detection and instance segmentation experiments further demonstrate the improved utility of enhanced images for visual perception, while tests on real-world data captured by a remotely operated vehicle (ROV) support the practical applicability of RPL-UIE.

Abstract

Explore similar work

RPL-UIE: Reliable Prior Learning for Underwater Image Enhancement

DY-LUT: Depth-Aware YCbCr Lookup Tables for Real-Time Underwater Image Enhancement

CoRe-UIE: Rethinking Coexisting and Region-wise Degradation for Underwater Image Enhancement