1 August 2026 in IEEE Transactions on Emerging Topics in Computational Intelligence
Underwater image enhancement (UIE) presents a significant challenge within computer vision research. Despite the development of numerous UIE algorithms, a thorough and systematic review is still absent. To foster future advancements, we provide a detailed overview of the UIE task from several perspectives. Firstly, we introduce the physical models, data construction processes, evaluation metrics, and loss functions. Secondly, we categorize and discuss recent algorithms based on their contributions, considering six aspects: network architecture, learning strategy, learning stage, auxiliary tasks, domain perspective, and disentanglement fusion. Thirdly, due to the varying experimental setups in the existing literature, a comprehensive and unbiased comparison is currently unavailable. To address this, we perform both quantitative and qualitative evaluations of state-of-the-art algorithms across multiple benchmark datasets. Lastly, we identify key areas for future research in UIE.
2 June 2025 in IEEE Transactions on Antennas and Propagation
Data-driven deep learning techniques have made notable advancements in modeling electromagnetic scattering problems. However, its accuracy on the testing dataset can be heavily reduced when data availability is constrained. To address the overfitting problem caused by limited data, we propose a novel hybrid physics-data-driven neural network (HPDNN), which incorporates both data loss and physical loss. Specifically, Maxwell’s equations are used as the physical constraint to evaluate the prediction accuracy of neural network, and the physical loss is combined with the data loss in the backpropagation for extrapolation tasks. The results show that the proposed method can achieve accuracy enhancement up to 15.7% over the typical data-driven method, and the field profile of prediction can be significantly improved with the aid of physical constrain. Besides, this work reveals that the convergence speed and accuracy of HPDNN are better than those of purely physics-informed neural network (PINN), exhibiting the importance of loss hybrid on accurate modeling of electromagnetics.
13 January 2025 in Information Fusion
7 December 2023 in IEEE Transactions on Geoscience and Remote Sensing
Pan-sharpening aims to super-resolve the low-resolution (LR) multi-spectral (MS) image under the guidance of high-resolution (HR) panchromatic (PAN) image. Existing deep learning-based pan-sharpening methods usually adhere to a common philosophy of learning complementary information between MS and PAN images. Despite remarkable advances, few studies consider the band-private characteristics which differ greatly from band to band. An ideal MS image, however, is jointly determined by its diverse spectral bands, thus the accurate restoration of every band will benefit the pan-sharpening performance. In this work, we propose a novel yet effective solution to reconstruct HRMS image by explicitly modulating every spectral band, conditioning on the PAN image. As a result, we design a spatially-adaptive spectral modulation network, dubbed SSMNet, which consists of three core designs: source-adaptive spectral modulator (SSM), cross-band information aggregation (CBIA) module, and cross-stage feature integration (CSFI) module. The first predicts a series of spatially-adaptive kernels to capture the local information of every spectral band. Followed by, the second is responsible for facilitating the information communication among various bands to guarantee the continuous spectral representations. Furthermore, the third attends to integrate the cross-stage output features to produce the pan-sharpened result. In addition, we also introduce the histogram loss to constrain the band-wise distribution of the final fused products. Extensive experiments demonstrate that our SSMNet achieves the favorable performance against other state-of-the-art methods on multiple satellite datasets. The code is available at https://github.com/ez4lionky/SSMNet/.