A boosted spiking neural network with retina mechanism, termed ReBo-SNN, for the reconstruction of hazy images that draws inspiration from the antagonistic mechanisms of retinal neurons and innovatively devises a retinal module to circumvent the issue of excessive smoothing.
Abstract
The haze effect induces light attenuation and color shifting in distant objects, thereby diminishing the contrast and saturation of images. In pursuit of restoring image clarity, a plethora of dehazing methods have been successively introduced. Nevertheless, the development of effective and practical methodologies remains a formidable challenge due to the requirement for powerful feature extraction capability alongside consideration of computational efficiency and model scalability. To address these issues, we put forward a boosted spiking neural network with retina mechanism, termed ReBo-SNN, for the reconstruction of hazy images. Specifically, the proposed network draws inspiration from the antagonistic mechanisms of retinal neurons and innovatively devises a retinal module. By extracting highly expressive features through ON and OFF roads, the retinal module can effectively augment visual resolution and contrast. Furthermore, to circumvent the issue of excessive smoothing, the network introduces a novel boosting strategy that realizes progressive feature enhancement and assists in reconstructing high-quality dehazed images. In addition, the network also incorporates dense residual connections with strided convolution to compensate for potential information loss during the translation, which significantly enhances the efficiency of feature transmission and integration. Extensive experiments on both synthetic and real-world hazy images demonstrate that our proposed network achieves superior quantitative and qualitative results. Based on the spiking mechanism and compact architecture, our network also exhibits a substantial energy efficiency advantage, which promotes its deployment in vision hardware to enhance overall performance.
This work pioneers the synergistic integration of SNNs into Transformer architectures for LLIE, establishing a compelling pathway toward powerful, energy-efficient low-level vision on resource-constrained platforms.
Hongzhi Wang, Xiubo Liang, Jinxing Han et al.· Advances in Neural Informati...· 0 citations
Experiments on HDRTV1K show that Bio-SFT achieves competitive perceptual quality and consistently improves HDR-VDP-3 and $\Delta E_{ITP}$ while reducing artifact propagation in symmetric guidance pipelines.
Tingyu Cheng, Ting Zhang, Chongyi Li et al.· 0 citations
Image quality degrades under complex imaging environments due to factors such as mixed noise and illumination degradation. This paper proposes a denoising and enhancement method based on a deep convolutional neural network. The network employs an end-to-end fully convolutional architecture, designing dual-branch multi-scale dilated convolutional modules to collaboratively capture microscopic texture and global topological information; a dual attention mechanism of concatenated spatial and channel attention is used to achieve adaptive feature calibration; dense residual skip connections are introduced to enhance the reuse of shallow geometric information; and a composite supervision function is constructed by combining smoothing L1 loss, perceptual loss, and structural similarity loss. This method achieves a peak signal-to-noise ratio of 31.72 dB, a structural similarity of 0.905, and a perceptual distance as low as 0.101. Processing time for a single $256 \times 256$ image is 21.3 ms, and the generalization performance across datasets is stable. The proposed method effectively suppresses complex noise while preserving fine structural and texture information.
Xiaojie Teng· 2026 5th International Confe...· 0 citations
Abstract - Underwater images frequently suffer from strong color casting and uneven lighting, both of which arise from light absorption, scattering, and suspended particles in the water. These distortions significantly hinder visual clarity and reduce the reliability of downstream analysis. Existing state of the art methods as Glad-Net and MSRCC-Net also suffer from color bias and uneven lighting and showed square effects in images recovery. This study proposes a deep learning based framework using MSRCC-Net by adding Adaptive Feature Decoupling Module (AFDM) and Hierarchical Feature Encoder(HFE) as novelty. This approach is designed to correct color imbalance while simultaneously normalizing illumination variations This proposed framework employs advanced neural architectures to extract global brightness inconsistencies and localized color deviations by proposing an approach named as CILA-SegNet (Color Casting and Illumination architectural segregation Network). It’s a combination of Multi-Scale Retinex with color correction module combined with CNN and transformer for axial-self attention module. Noise is removing by passing the preprocessed and augmented images to Adaptive Feature Decoupling Module(AFDM) for segregation in LAB channels and then to Hierarchical Feature Encoder(HFE) as Adaptive CLAHE and A,B soft color correction with MSRCC-Net model to improve the results. The initial results are promising, showing noticeable improvements in brightness, contrast, and image details with mean intensity growing from 116.35 to 119.54, contrast improving from 39.61 to 41.87, and entropy rising from 6.98 to 7.27 .This study obtained the noticeable reduction in computational cost with 11m parameters, 43.60g FLOPs and 0.150s latency. After results image grew slightly brighter and more detailed after CLAHE and produces visually coherent, high-quality underwater images suitable for real-world operational use.
Ishmal Nayab, M. Umair, Maham Mehr et al.· Journal of Independent Studi...· 0 citations
Images captured by consumer electronic devices, such as mobile phones and digital cameras, often suffer from low-light degradation due to sensor limitations and imaging pipelines, which degrades visual quality and affects downstream vision tasks. Existing methods based on Convolutional Neural Networks (CNNs) and Transformers have dominated current low-light image enhancement (LIE) due to their excellent ability to model hierarchical features. However, CNNs operate in local receptive fields that cannot model long-range dependencies, while Transformers overcome this problem but incur substantial computational costs. To address these challenges, we propose MambaLIE, a Scene Light Intensity-Boosted Low-Light Image Enhancement method based on a State Space Model (SSM). We first introduce scene light intensity to improve the structural distribution of illumination, which is then gated with the low-light input to guide enhancement. To better model the illumination while maintaining computational efficiency, we propose the Locally Enhanced State Space Model (LESSM) for efficient light enhancement. Our LESSM contains two branches: an SSM branch and a Local Enhanced branch, where the former is used to model the long-range dependencies with linear time complexity, while the latter is used to enhance local feature representations. Extensive experiments demonstrate that MambaLIE outperforms state-of-the-art CNN-based and Transformer-based LIE methods on four widely used synthetic benchmarks and five publicly available real-world benchmarks in terms of accuracy, speed, and model size, making it suitable for practical deployment on resource-constrained devices.
Wanshu Fan, Xiangyu Li, Cong Wang et al.· IEEE transactions on consume...· 0 citations
Images captured at night are often degraded by both low-light conditions and motion blur, primarily caused by insufficient illumination and relative motion during exposure. Existing methods typically address only a single type of degradation or rely heavily on data-driven deep neural networks, which limits interpretability and generalization capability. To overcome these limitations, we propose a model-driven deep neural network to effectively handle the joint degradation of low light and blur. By integrating image restoration and Retinex-based modeling, we formulate a unified optimization framework and employ a proximal gradient descent (PGD)-based deep unfolding strategy to recover reflectance and illumination components from low-light blurry images. Furthermore, we design an illumination enhancement module (IEM) and a reflectance refinement module (RRM) to improve brightness, restore fine details, and suppress noise. Extensive experiments on both synthetic and real-world datasets demonstrate that the proposed method outperforms state-of-the-art approaches. Code is available at https://github.com/cassiopeia-yxx/MDNet.
Yao Xiao, Youshen Xia, Zhenyu Lu et al.· IEEE Transactions on Neural...· 0 citations