Article
Computer Science, Artificial Intelligence
Jun Wang, Peilin Liu, Fei Wen
Summary: This work aims to develop a self-supervised learning method for RGB-guided depth image enhancement, which does not require any noisy-clean pairs but can significantly boost the enhancement performance on real-world noisy depth images.
IEEE TRANSACTIONS ON IMAGE PROCESSING
(2023)
Article
Chemistry, Multidisciplinary
Enqi Zhang, Lihong Guo, Junda Guo, Shufeng Yan, Xiangyang Li, Lingsheng Kong
Summary: This paper proposes a low-brightness image enhancement algorithm based on multi-scale fusion. By using brightness transformation and illumination estimation techniques, advantageous features are extracted and images are fused to improve image quality. Experimental results demonstrate that the proposed method has better enhancement effect.
APPLIED SCIENCES-BASEL
(2023)
Article
Computer Science, Information Systems
Yongqiang Chen, Chenglin Wen, Weifeng Liu, Wei He
Summary: In this paper, an end-to-end low-light image enhancement network based on the YCbCr color space is proposed to address brightness distortion and noise issues in the RGB color space. The network utilizes the characteristics of YCbCr color space and includes a CNN branch, a U-net branch, and a fusion module for enhancing contrast and eliminating noise. Experimental results demonstrate that the proposed method generates enhanced images with richer details, more realistic colors, and less noise.
Article
Engineering, Electrical & Electronic
Chufan Liu, Xin Shu, Lei Pan, Jinlong Shi, Bin Han
Summary: In this paper, we propose a multiscale dual-color space underwater image enhancement network (MSDC-Net) to address the color deviations and hazy effects in underwater images. The network consists of a color correction block and a deep learning-based encoder-decoder architecture, which extracts rich features and produces competitive outputs. Experimental results on real-world and synthetic underwater images demonstrate the outstanding performance of the proposed MSDC-Net.
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT
(2023)
Article
Optics
Sangjae Ahn, Joongchol Shin, Heunseung Lim, Jaehee Lee, Joonki Paik
Summary: In this paper, a novel low-light image enhancement method is proposed, which enhances brightness and contrast simultaneously by combining optimization-based decomposition and enhancement network. The method works in two steps, including Retinex decomposition and illumination enhancement, and can be trained in an end-to-end manner. Extensive experiments demonstrate that our method outperforms state-of-the-art low-light enhancement methods in terms of both objective and subjective measures.
Article
Engineering, Electrical & Electronic
Hao-Tian Wu, Xin Cao, Ruoyan Jia, Yiu-Ming Cheung
Summary: A new 2D histogram based contrast enhancement with reversible data hiding (CE-RDH) scheme is proposed in this paper, taking brightness preservation into account. The scheme can preserve image color and brightness while achieving better image quality than existing schemes.
IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY
(2022)
Article
Computer Science, Information Systems
Lalit Maurya, Viney Lohchab, Prasant Kumar Mahapatra, Janos Abonyi
Summary: Many vision-based systems suffer from poor levels of contrast and brightness due to inadequate and improper illumination during image acquisition. By using nature-inspired optimization, a balance between contrast and brightness can be achieved in image enhancement, improving image quality.
JOURNAL OF KING SAUD UNIVERSITY-COMPUTER AND INFORMATION SCIENCES
(2022)
Article
Optics
Yuehan Chen, Yafeng Li, Yulin Wang, Zetian Mi, Yujia Wang, Xianping Fu
Summary: This paper proposes a robust polarization-based underwater image enhancement method using anchor brightness adaptation (ABA) to solve the problem of inaccurate parameter estimation caused by amplified camera noise in low illumination conditions. By relying on the relationship between the Stokes Vector and the angle of polarization (AoP), the proposed neighborhood high fidelity constraint (NHFC) can select the most suitable region for parameter estimation, reducing camera random noise interference under low illumination conditions. The estimation of background light intensity and scene transmission map based on polarization characteristics effectively enhances the image, and the introduction of ABA ensures the best exposure of the enhanced image.
OPTICS AND LASERS IN ENGINEERING
(2023)
Article
Computer Science, Information Systems
Shih-Chia Huang, Da-Wei Jaw, Bo-Hao Chen, Sy-Yen Kuo
Summary: This paper introduces a single image enhancement algorithm based on the human visual system, which can assess and enhance image illumination based on Luminance Perception procedure, with experimental results showing superior performance compared to other algorithms.
IEEE TRANSACTIONS ON EMERGING TOPICS IN COMPUTING
(2021)
Article
Computer Science, Information Systems
Yinhua Su, Mian Wu, Ying Yan
Summary: This study proposes two algorithms to improve the visual effect of images captured in low illumination environments and preserve image information integrity. The image enhancement algorithm combines the WASOBI denoising method and adaptive gamma correction for noise reduction and brightness adjustment. The image brightness equalization algorithm divides the image based on illumination area and uses guided filtering and adaptive gamma correction to equalize brightness. Experimental results demonstrate the effectiveness and stability of the proposed image enhancement algorithm. The subjective evaluation score of the brightness equalization algorithm is 4.7, indicating good visual effects. Objective evaluation results show the capability of the equalization algorithm, with the shortest operation time of 0.9452 seconds compared to other algorithms. Hence, the proposed algorithms have certain application value.
Article
Multidisciplinary Sciences
Hosang Lee
Summary: This paper proposes a method for enhancing sandstorm images by balancing color components using image-adaptive eigenvalues, followed by a dehazing procedure using a multiscale convolution neural network, resulting in images with balanced colors and brightness-adaptive features.
Article
Computer Science, Artificial Intelligence
Qian Chen, Keren Fu, Ze Liu, Geng Chen, Hongwei Du, Bensheng Qiu, Ling Shao
Summary: Salient object detection (SOD) using RGB-D images has drawn significant attention in computer vision. The proposed multi-modal enhancement and fusion network (EF-Net) utilizes color hint map module, depth enhancement module, and layer-wise aggregation module to effectively detect salient objects by combining information from RGB images and depth maps. Extensive experiments show that EF-Net outperforms 12 state-of-the-art RGB-D saliency detection approaches in terms of key evaluation metrics.
PATTERN RECOGNITION
(2021)
Article
Computer Science, Artificial Intelligence
Yue Wang, Xu Jia, Lu Zhang, Yuke Li, James H. Elder, Huchuan Lu
Summary: RGB-D saliency detection combines RGB images and depth maps to improve the prediction of salient regions. A transformer-based structure is proposed to fully integrate information at multiple scales and modalities. Experimental results show that the proposed network outperforms state-of-the-art methods in terms of performance and efficiency.
PATTERN RECOGNITION
(2023)
Article
Engineering, Electrical & Electronic
Lili Dong, Weidong Zhang, Wenhai Xu
Summary: This paper proposes a method for underwater image enhancement using integrated RGB and LAB color models. By correcting color channels with dedicated fractions and applying local contrast enhancement and gain equalization strategies, the proposed method improves the color shift and visibility issues in underwater images. Experimental results demonstrate that the method outputs high-quality underwater images with natural color and high visibility.
SIGNAL PROCESSING-IMAGE COMMUNICATION
(2022)
Article
Computer Science, Artificial Intelligence
Ya'nan Wang, Zhuqing Jiang, Chang Liu, Kai Li, Aidong Men, Haiying Wang, Xiaobo Chen
Summary: This study proposes a user-friendly neural network for multi-level low-light image enhancement. By decomposing the image into content and luminance components, the image brightness is enhanced to different levels, and different brightness references are selected based on user requirements. Furthermore, information except for brightness is preserved to alleviate color distortion.
PATTERN RECOGNITION
(2022)
