Low-light image enhancement aims to improve the visibility of severely degraded images captured under insufficient illumination, alleviating the adverse effects of illumination degradation on image quality. Traditional Retinex-based approaches, inspired by human visual perception of brightness and color, decompose an image into illumination and reflectance components to restore fine details. However, their limited capacity for handling noise and complex lighting conditions often leads to distortions and artifacts in the enhanced results, particularly under extreme low-light scenarios. Although deep learning methods built upon Retinex theory have recently advanced the field, most still suffer from insufficient interpretability and sub-optimal enhancement performance. This paper presents RetinexWT, a novel framework that tightly integrates classical Retinex theory with modern deep learning. Following Retinex principles, RetinexWT employs wavelet transforms to estimate illumination maps for brightness adjustment. A detail-recovery module that synergistically combines Vision Transformer (ViT) and wavelet transforms is then introduced to guide the restoration of lost details, thereby improving overall image quality. Within the framework, wavelet decomposition splits input features into high-frequency and low-frequency components, enabling scale-specific processing of global illumination/color cues and fine textures. Furthermore, a gating mechanism selectively fuses down-sampled and up-sampled features, while an attention-based fusion strategy enhances model interpretability. Extensive experiments on the LOL dataset demonstrate that RetinexWT surpasses existing Retinex-oriented deep-learning methods, achieving an average Peak Signal-to-Noise Ratio (PSNR) improvement of 0.22 dB over the current State Of The Art (SOTA), thereby confirming its superiority in low-light image enhancement. Code is available at https://github.com/CHEN-hJ516/RetinexWT (accessed on 14 October 2025).

Underwater images often affect the effectiveness of underwater visual tasks due to problems such as light scattering, color distortion, and detail blurring, limiting their application performance. Existing underwater image enhancement methods, although they can improve the image quality to some extent, often lead to problems such as detail loss and edge blurring. To address these problems, we propose FENet, an efficient underwater image enhancement method. FENet first obtains three different scales of images by image downsampling and then transforms them into the frequency domain to extract the low-frequency and high-frequency spectra, respectively. Then, a distance mask and a mean mask are constructed based on the distance and magnitude mean for enhancing the high-frequency part, thus improving the image details and enhancing the effect by suppressing the noise in the low-frequency part. Affected by the light scattering of underwater images and the fact that some details are lost if they are directly reduced to the spatial domain after the frequency domain operation. For this reason, we propose a multi-stage residual feature aggregation module, which focuses on detail extraction and effectively avoids information loss caused by global enhancement. Finally, we combine the edge guidance strategy to further enhance the edge details of the image. Experimental results indicate that FENet outperforms current state-of-the-art underwater image enhancement methods in quantitative and qualitative evaluations on multiple publicly available datasets.

Camouflaged Object Detection (COD) aims to identify objects that share highly similar patterns—such as texture, intensity, and color—with their surrounding environment. Due to their intrinsic resemblance to the background, camouflaged objects often exhibit vague boundaries and varying scales, making it challenging to accurately locate targets and delineate their indistinct edges. To address this, we propose a novel camouflaged object detection network called Edge-Guided and Multi-scale Fusion Network (EGMFNet), which leverages edge-guided multi-scale integration for enhanced performance. The model incorporates two innovative components: a Multi-scale Fusion Module (MSFM) and an Edge-Guided Attention Module (EGA). These designs exploit multi-scale features to uncover subtle cues between candidate objects and the background while emphasizing camouflaged object boundaries. Moreover, recognizing the rich contextual information in fused features, we introduce a Dual-Branch Global Context Module (DGCM) to refine features using extensive global context, thereby generating more informative representations. Experimental results on four benchmark datasets demonstrate that EGMFNet outperforms state-of-the-art methods across five evaluation metrics. Specifically, on COD10K, our EGMFNet-P improves $F_{\beta }$ by 4.8 points and reduces mean absolute error (MAE) by 0.006 compared with ZoomNeXt; on NC4K, it achieves a 3.6-point increase in $F_{\beta }$. On CAMO and CHAMELEON, it obtains 4.5-point increases in $F_{\beta }$, respectively. These consistent gains substantiate the superiority and robustness of EGMFNet.

Semantic segmentation in street scenes is a crucial technology for autonomous driving to analyze the surrounding environment. In street scenes, issues such as high image resolution caused by a large viewpoints and differences in object scales lead to a decline in real-time performance and difficulties in multi-scale feature extraction. To address this, we propose a bilateral-branch real-time semantic segmentation method based on semantic information distillation (BSDNet) for street scene images. The BSDNet consists of a Feature Conversion Convolutional Block (FCB), a Semantic Information Distillation Module (SIDM), and a Deep Aggregation Atrous Convolution Pyramid Pooling (DASP). FCB reduces the semantic gap between the backbone and the semantic branch. SIDM extracts high-quality semantic information from the Transformer branch to reduce computational costs. DASP aggregates information lost in atrous convolutions, effectively capturing multi-scale objects. Extensive experiments conducted on Cityscapes, CamVid, and ADE20K, achieving an accuracy of 81.7 $\mathrm{\%}$ Mean Intersection over Union (mIoU) at 70.6 Frames Per Second (FPS) on Cityscapes, demonstrate that our method achieves a better balance between accuracy and inference speed.

Rail fasteners are a crucial component of the railway transportation safety system. These fasteners, distinguished by their high length-to-width ratio, frequently encounter elevated failure rates, necessitating manual inspection and maintenance. Manual inspection not only consumes time but also poses the risk of potential oversights. With the advancement of deep learning technology in rail fasteners, challenges such as the complex background of rail fasteners and the similarity in their states are addressed. We have proposed an efficient and high-precision rail fastener detection algorithm, named YOLO-O2E (you only look once-O2E). Firstly, we propose the EFOV (Enhanced Field of View) structure, aiming to adjust the effective receptive field size of convolutional kernels to enhance insensitivity to small spatial variations. Additionally, The OD_MP (ODConv and MP_2) and EMA (Efficient Multi-Scale Attention) modules mentioned in the algorithm can acquire a wider spectrum of contextual information, enhancing the model’s ability to recognize and locate objectives. Additionally, we collected and prepared the GKA dataset, sourced from real train tracks. Through testing on the GKA dataset and the publicly available NUE-DET dataset, our method outperforms general-purpose object detection algorithms. On the GKA dataset, our model achieved a mAP@0.5 value of 97.6% and a mAP@0.5:0.95 value of 83.9%, demonstrating excellent inference speed. YOLO-O2E is an algorithm for detecting anomalies in railway fasteners that is applicable in practical industrial settings, addressing the industry gap in rail fastener detection.

While single-modal visible light images or infrared images provide limited information, infrared light captures significant thermal radiation data, whereas visible light excels in presenting detailed texture information. Combining images obtained from both modalities allows for leveraging their respective strengths and mitigating individual limitations, resulting in high-quality images with enhanced contrast and rich texture details. Such capabilities hold promising applications in advanced visual tasks including target detection, instance segmentation, military surveillance, pedestrian detection, among others. This paper introduces a novel approach, a dual-branch decomposition fusion network based on AutoEncoder (AE), which decomposes multi-modal features into intensity and texture information for enhanced fusion. Local contrast enhancement module (CEM) and texture detail enhancement module (DEM) are devised to process the decomposed images, followed by image fusion through the decoder. The proposed loss function ensures effective retention of key information from the source images of both modalities. Extensive comparisons and generalization experiments demonstrate the superior performance of our network in preserving pixel intensity distribution and retaining texture details. From the qualitative results, we can see the advantages of fusion details and local contrast. In the quantitative experiments, entropy (EN), mutual information (MI), structural similarity (SSIM) and other results have improved and exceeded the SOTA (State of the Art) model as a whole.

Semantic segmentation of driving scene images is crucial for autonomous driving. While deep learning technology has significantly improved daytime image semantic segmentation, nighttime images pose challenges due to factors like poor lighting and overexposure, making it difficult to recognize small objects. To address this, we propose an Image Adaptive Enhancement (IAEN) module comprising a parameter predictor (Edip), multiple image processing filters (Mdif), and a Detail Processing Module (DPM). Edip combines image processing filters to predict parameters like exposure and hue, optimizing image quality. We adopt a novel image encoder to enhance parameter prediction accuracy by enabling Edip to handle features at different scales. DPM strengthens overlooked image details, extending the IAEN module’s functionality. After the segmentation network, we integrate a Depth Guided Filter (DGF) to refine segmentation outputs. The entire network is trained end-to-end, with segmentation results guiding parameter prediction optimization, promoting self-learning and network improvement. This lightweight and efficient network architecture is particularly suitable for addressing challenges in nighttime image segmentation. Extensive experiments validate significant performance improvements of our approach on the ACDC-night and Nightcity datasets.