PatchNet: a tiny low-light image enhancement net

Liu, Zhenbing; Wang, Kaijie; Wang, Zimin; Lu, Haoxiang; Yuan, Lu

doi:10.1117/1.jei.30.3.033023

Cited by 4 publications

(3 citation statements)

References 25 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Incorporating Retinex theory, Retinex-Net 35 divided the enhancement process into separate reflectance and illumination components. PatchNet 36 introduced an approach to preserving the incremental information transition from low-light to normal images. LAE-Net 37 uses adaptive kernel selection and feature adaptation to balance light intensity, detail representation, and color fidelity.…”

Section: Related Workmentioning

confidence: 99%

Low-light image enhancement via multistage Laplacian feature fusion

Liu,

Huang,

Zhang

et al. 2024

J. Electron. Imag.

Self Cite

View full text Add to dashboard Cite

Low-light images inevitably suffer from degradation problems during the enhancement process, such as loss of detail and local overexposure or underexposure. Many existing methods target only one of these issues, leading to suboptimal results. We propose a multistage Laplacian feature fusion network (MLFFNet) capable of simultaneously mitigating both degradation difficulties. MLFFNet employs a pyramid framework that incrementally learns the degradation functions across various frequency bands, leveraging Laplacian feature maps at each stage.A key innovation of our approach is the supervised refinement module, which refines features through a dual strategy: an attention mechanism that enriches the detail capture, including edges, textures, and colors, and a residual mechanism that adjusts the luminance for a balanced exposure. The resultant enhanced image benefits from channel-wise attention, ensuring superior enhancement. Finally, the enhanced image is acquired by several channel attention blocks. Extensive experiments on various datasets indicate that our proposed MLFFNet outperforms the state-of-the-art methods both qualitatively and quantitatively.

show abstract

Section: Related Workmentioning

confidence: 99%

Low-light image enhancement via multistage Laplacian feature fusion

Liu,

Huang,

Zhang

et al. 2024

J. Electron. Imag.

Self Cite

View full text Add to dashboard Cite

show abstract

“…One possible solution to this problem is to enhance images featuring poor illumination. Although low-light enhancement techniques have been successful in improving visual quality for classification tasks [4][5][6][7][8], extreme variations in illumination still pose significant challenges. As part of this study, experiments were conducted using images enhanced by a state-of-the-art (SOTA) method [6].…”

Section: Introductionmentioning

confidence: 99%

Two-Stage Progressive Learning for Vehicle Re-Identification in Variable Illumination Conditions

Wu,

Jin,

2023

Electronics

View full text Add to dashboard Cite

Vehicle matching in variable illumination environments can be challenging due to the heavy dependence of vehicle appearance on lighting conditions. To address this issue, we propose a two-stage progressive learning (TSPL) framework. In the first stage, illumination-aware metric learning is enforced using a two-branch network via two illumination-specific feature spaces, used to explicitly model differences in lighting. In the second stage, discriminative feature learning is introduced to extract distinguishing features from a given vehicle. This process consists of a local feature extraction attention module, a local constraint, and a balanced sampling strategy. During the metric learning phase, the model expresses the union of local features, extracted from the attention module, with illumination-specific global features to form joint vehicle features. As part of the study, we construct a large-scale dataset, termed VERI-DAN (vehicle re-identification across day and night), to address the current lack of vehicle datasets exhibiting variable lighting conditions. This set is composed of 200,004 images from 16,654 vehicles, collected in various natural illumination environments. Validation experiments conducted with the VERI-DAN and Vehicle-1M datasets demonstrated that our proposed methodology effectively improved vehicle re-identification Rank-1 accuracy.

show abstract

“…Javadi et al proposed a piecewise linear histogram-equalization algorithm to enhance the contrast in the frequency domain by stretching the intensity of the entire spectrum [30]. In recent years, with the rapid development of deep learning, a series of powerful deep-learning-based low-illuminance image-enhancement methods have been proposed [31][32][33][34][35]. However, owing to the lack of both color information and training data, they are difficult to apply to X-ray image processing.…”

Section: Introductionmentioning

confidence: 99%

Image contrast enhancement algorithm for X-ray observation of space materials in situ

Liu

Pan

et al. 2022

J. Inst.

View full text Add to dashboard Cite

Chinese Space Station has planned a high-temperature material science experiment rack, equipped with an X-ray projection imaging module, to support the development of material experiments and research in space. BiFeO3 has been selected as the first batch of experimental materials for Chinese Space Station. The melting and solidification process of BiFeO3, an opaque, high-temperature material, is observed by X-ray observation module in situ. X-ray is the dominant way to observe opaque materials due to its penetrability. In-situ observation of materials is the top priority of this study, so we have strict requirements on image quality, and high-quality images can better analyze the properties and properties of materials. Limited by narrow size and high temperature conditions, the X-ray images collected have low contrast, serious noise pollution, and poor imaging quality. To enhance the contrast and improve the edge details of such images, a grayscale weighted histogram equalization combined with high-frequency enhancement (GWHE-HFE) algorithm is proposed. First, we add a mask to the input image to obtain the region of interest (ROI), and then filter out the low-frequency components of the image by Gaussian high-pass filter to preserve high-frequency details. Second, the image obtained in the previous and the X-ray image of ROI are respectively multiplied by a coefficient and added to obtain the edge-emphasized X-ray image. And then, we use grayscale weighted histogram equalization (GWHE) to process the image obtained in the second step to obtain the contrast enhanced X-ray image. The enhanced image shows the crystal grains and the thin bands where the solid and the melt intersect, and it is helpful to accurately locate the solid solution interface. Experiments on X-ray images of BiFeO3 growth demonstrate that this combined method outperforms existing ones both qualitatively and quantitatively, providing an in-depth and effective analysis method for in high-temperature material-science experiments.

show abstract

PatchNet: a tiny low-light image enhancement net

Cited by 4 publications

References 25 publications

Low-light image enhancement via multistage Laplacian feature fusion

Low-light image enhancement via multistage Laplacian feature fusion

Two-Stage Progressive Learning for Vehicle Re-Identification in Variable Illumination Conditions

Image contrast enhancement algorithm for X-ray observation of space materials in situ

Contact Info

Product

Resources

About