Journals > > Topics > Imaging Systems and Image Processing
Imaging Systems and Image Processing|85 Article(s)
Image-free cross-species pose estimation via an ultra-low sampling rate single-pixel camera
Xin Wu, Cheng Zhou, Binyu Li, Jipeng Huang, Yanli Meng, Lijun Song, and Shensheng Han
Cross-species pose estimation plays a vital role in studying neural mechanisms and behavioral patterns while serving as a fundamental tool for behavior monitoring and prediction. However, conventional image-based approaches face substantial limitations, including excessive storage requirements, high transmission bandwidth demands, and massive computational costs. To address these challenges, we introduce an image-free pose estimation framework based on single-pixel cameras operating at ultra-low sampling rates (6.260 × 10-4). Our method eliminates the need for explicit or implicit image reconstruction, instead directly extracting pose information from highly compressed single-pixel measurements. It dramatically reduces data storage and transmission requirements while maintaining accuracy comparable to traditional image-based methods. Our solution provides a practical approach for real-world applications where bandwidths and computational resources are constrained. Cross-species pose estimation plays a vital role in studying neural mechanisms and behavioral patterns while serving as a fundamental tool for behavior monitoring and prediction. However, conventional image-based approaches face substantial limitations, including excessive storage requirements, high transmission bandwidth demands, and massive computational costs. To address these challenges, we introduce an image-free pose estimation framework based on single-pixel cameras operating at ultra-low sampling rates (6.260 × 10-4). Our method eliminates the need for explicit or implicit image reconstruction, instead directly extracting pose information from highly compressed single-pixel measurements. It dramatically reduces data storage and transmission requirements while maintaining accuracy comparable to traditional image-based methods. Our solution provides a practical approach for real-world applications where bandwidths and computational resources are constrained.
Chinese Optics Letters
- Publication Date: Aug. 22, 2025
- Vol. 23, Issue 9, 091101 (2025)
MWIR-LWIR dual-band imaging system with hybrid refractive-diffractive-metasurface optics for spatially separated focal planes
Zhiang Qian, Bingxia Wang, Bowei Xia, Zhihao He, Xinyi Lei, Jianbao Xu, Jierong Gu, Xiang Shen, and Shuming Wang
The integration of mid-wave infrared (MWIR) and long-wave infrared (LWIR) imaging into a compact high-performance system remains a significant challenge in infrared optics. In this work, we present a dual-band infrared imaging system based on hybrid refractive-diffractive-metasurface optics. The system integrates a silicon-based metalens for the MWIR channel and a hybrid refractive-diffractive lens made of high-refractive-index chalcogenide glass for the LWIR channel. It achieves a compact total track length (TTL) of 11.31 mm. The MWIR channel features a 1.0 mm entrance pupil diameter, a 10° field of view (FOV), and achromatic imaging across the 3–4 µm spectral range with a focal length of 1.5 mm. The LWIR channel provides an 8.7 mm entrance pupil diameter, a 30° FOV, and broadband achromatic correction over the 8–12 µm spectral range with a focal length of 13 mm. To further enhance spatial resolution and recover fine image details, we employ low-rank adaptation (LoRA) fine-tuning within a physics-informed StableSR framework. This hybrid optical approach establishes, to our knowledge, a new paradigm in dual-band imaging systems by leveraging the complementary advantages of metalens dispersion engineering, diffractive phase modulation, and conventional refractive optics, delivering a lightweight, multispectral imaging solution with superior spectral discrimination and system compactness. The integration of mid-wave infrared (MWIR) and long-wave infrared (LWIR) imaging into a compact high-performance system remains a significant challenge in infrared optics. In this work, we present a dual-band infrared imaging system based on hybrid refractive-diffractive-metasurface optics. The system integrates a silicon-based metalens for the MWIR channel and a hybrid refractive-diffractive lens made of high-refractive-index chalcogenide glass for the LWIR channel. It achieves a compact total track length (TTL) of 11.31 mm. The MWIR channel features a 1.0 mm entrance pupil diameter, a 10° field of view (FOV), and achromatic imaging across the 3–4 µm spectral range with a focal length of 1.5 mm. The LWIR channel provides an 8.7 mm entrance pupil diameter, a 30° FOV, and broadband achromatic correction over the 8–12 µm spectral range with a focal length of 13 mm. To further enhance spatial resolution and recover fine image details, we employ low-rank adaptation (LoRA) fine-tuning within a physics-informed StableSR framework. This hybrid optical approach establishes, to our knowledge, a new paradigm in dual-band imaging systems by leveraging the complementary advantages of metalens dispersion engineering, diffractive phase modulation, and conventional refractive optics, delivering a lightweight, multispectral imaging solution with superior spectral discrimination and system compactness.
Chinese Optics Letters
- Publication Date: Jul. 31, 2025
- Vol. 23, Issue 8, 081105 (2025)
High-resolution non-line-of-sight imaging via polarization differential correlography
Lingfeng Liu, Shuo Zhu, Yi Wei, Jingye Miao, Wenjun Zhang, Lianfa Bai, Enlai Guo, and Jing Han
Non-line-of-sight (NLOS) imaging enables the detection and reconstruction of hidden objects around corners, offering promising applications in autonomous driving, remote sensing, and medical diagnosis. However, existing steady-state NLOS imaging methods face challenges in achieving high efficiency and precision due to the need for multiple diffuse reflections and incomplete Fourier amplitude sampling. This study proposes, to our knowledge, a novel steady-state NLOS imaging technique via polarization differential correlography (PDC-NLOS). By employing the polarization difference of the laser speckle, the method designs a single-shot polarized speckle illumination strategy. The fast and stable real-time encoding for hidden objects ensures stable imaging quality of the PDC-NLOS system. The proposed method demonstrates millimeter-level imaging resolution when imaging horizontally and vertically striped objects. Non-line-of-sight (NLOS) imaging enables the detection and reconstruction of hidden objects around corners, offering promising applications in autonomous driving, remote sensing, and medical diagnosis. However, existing steady-state NLOS imaging methods face challenges in achieving high efficiency and precision due to the need for multiple diffuse reflections and incomplete Fourier amplitude sampling. This study proposes, to our knowledge, a novel steady-state NLOS imaging technique via polarization differential correlography (PDC-NLOS). By employing the polarization difference of the laser speckle, the method designs a single-shot polarized speckle illumination strategy. The fast and stable real-time encoding for hidden objects ensures stable imaging quality of the PDC-NLOS system. The proposed method demonstrates millimeter-level imaging resolution when imaging horizontally and vertically striped objects.
Chinese Optics Letters
- Publication Date: Jul. 31, 2025
- Vol. 23, Issue 8, 081104 (2025)
Impact of frequency-domain filtering on facial expression recognition in spatial domain
Ju Li, Yifei Wang, Lixing Qian, Junjie Guo, and Yong Zhang
Deep learning-assisted facial expression recognition has been extensively investigated in sentiment analysis, human-computer interaction, and security surveillance. Generally, the recognition accuracy in previous reports requires high-quality images and powerful computational resources. In this work, we quantitatively investigate the impacts of frequency-domain filtering on spatial-domain facial expression recognition. Based on the Fer2013 dataset, we filter out 82.64% of high-frequency components, resulting in a decrease of 3.85% in recognition accuracy. Our findings well demonstrate the essential role of low-frequency components in facial expression recognition, which helps reduce the reliance on high-resolution images and improve the efficiency of neural networks. Deep learning-assisted facial expression recognition has been extensively investigated in sentiment analysis, human-computer interaction, and security surveillance. Generally, the recognition accuracy in previous reports requires high-quality images and powerful computational resources. In this work, we quantitatively investigate the impacts of frequency-domain filtering on spatial-domain facial expression recognition. Based on the Fer2013 dataset, we filter out 82.64% of high-frequency components, resulting in a decrease of 3.85% in recognition accuracy. Our findings well demonstrate the essential role of low-frequency components in facial expression recognition, which helps reduce the reliance on high-resolution images and improve the efficiency of neural networks.
Chinese Optics Letters
- Publication Date: Jul. 22, 2025
- Vol. 23, Issue 8, 081103 (2025)
Compact mechanical Alvarez lenses with a wide varifocal range using a cam-driver
Cancan Yao, Qun Hao, Lin Liu, Jie Cao, Haoyue Xing, Zhaohui Li, and Yang Cheng
This study proposes compact Alvarez varifocal lenses with a wide varifocal range, which consist of a set of Alvarez lenses and three sets of ordinary lenses. The Alvarez lenses have a double freeform surface and are driven by a cam-driven structure. The axial size of the proposed varifocal Alvarez lenses is only 30.50 mm. The experimental results show that the proposed varifocal lens can achieve a focal length range from 15 to 75 mm, and the imaging quality is still in an acceptable range for optical lens requirements. The compact varifocal Alvarez lenses are expected to be used in surveillance systems, industrial inspection, and machine vision. This study proposes compact Alvarez varifocal lenses with a wide varifocal range, which consist of a set of Alvarez lenses and three sets of ordinary lenses. The Alvarez lenses have a double freeform surface and are driven by a cam-driven structure. The axial size of the proposed varifocal Alvarez lenses is only 30.50 mm. The experimental results show that the proposed varifocal lens can achieve a focal length range from 15 to 75 mm, and the imaging quality is still in an acceptable range for optical lens requirements. The compact varifocal Alvarez lenses are expected to be used in surveillance systems, industrial inspection, and machine vision.
Chinese Optics Letters
- Publication Date: Jul. 22, 2025
- Vol. 23, Issue 8, 081102 (2025)
Photon-level single-pixel imaging of dynamic features in frequency domain
Shuxiao Wu, Jianyong Hu, Yaole Cao, Yuxing Jiang, Yanshan Fan, Zhixing Qiao, Guosheng Feng, Changgang Yang, Jianqiang Liu, Ruiyun Chen, Chengbing Qin, Guofeng Zhang, Liantuan Xiao, and Suotang Jia
Photon-level single-pixel imaging overcomes the reliance of traditional imaging techniques on large-scale array detectors, offering the advantages such as high sensitivity, high resolution, and efficient photon utilization. In this paper, we propose a photon-level dynamic feature single-pixel imaging method, leveraging the frequency domain sparsity of the object’s dynamic features to construct a compressed measurement system through discrete random photon detection. In the experiments, we successfully captured 167 and 200 Hz featured frequencies and achieved high-quality image reconstruction with a data compression ratio of 20%. Our approach introduces a new detection dimension, significantly expanding the applications of photon-level single-pixel imaging in practical scenarios. Photon-level single-pixel imaging overcomes the reliance of traditional imaging techniques on large-scale array detectors, offering the advantages such as high sensitivity, high resolution, and efficient photon utilization. In this paper, we propose a photon-level dynamic feature single-pixel imaging method, leveraging the frequency domain sparsity of the object’s dynamic features to construct a compressed measurement system through discrete random photon detection. In the experiments, we successfully captured 167 and 200 Hz featured frequencies and achieved high-quality image reconstruction with a data compression ratio of 20%. Our approach introduces a new detection dimension, significantly expanding the applications of photon-level single-pixel imaging in practical scenarios.
Chinese Optics Letters
- Publication Date: Jul. 03, 2025
- Vol. 23, Issue 8, 081101 (2025)
Experimental parameter error compensation in deep-learning-based coherent diffraction imaging
Shihong Huang, Yanxu Yang and Zizhong Liu
The integration of deep learning into computational imaging has driven substantial advancements in coherent diffraction imaging (CDI). While physics-driven neural networks have emerged as a promising approach through their unsupervised learning paradigm, their practical implementation faces critical challenges: measurement uncertainties in physical parameters (e.g., the propagation distance and the size of sample area) severely degrade reconstruction quality. To overcome this limitation, we propose a deep-learning-enabled spatial sample interval optimization framework that synergizes physical models with neural network adaptability. Our method embeds spatial sample intervals as trainable parameters within a PhysenNet architecture coupled with Fresnel diffraction physics, enabling simultaneous image reconstruction and system parameter calibration. Experimental validation demonstrates robust performance with structural similarity (SSIM) values consistently maintained at 0.6 across diffraction distances spanning of 10–200 mm, using a 1024 × 1024 region of interest (ROI) from a 1624 × 1440 CCD (pixel size: 4.5 μm) under 632.8 nm illumination. This framework has excellent fault tolerance, that is, it can still maintain high-quality image restoration even when the propagation distance measurement error is large. Compared to conventional iterative reconstruction algorithms, this approach can transform fixed parameters into learnable parameters, making almost all image restoration experiments easier to implement, enhancing system robustness against experimental uncertainties. This work establishes, to our knowledge, a new paradigm for adaptive diffraction imaging systems capable of operating in complex real scenarios. The integration of deep learning into computational imaging has driven substantial advancements in coherent diffraction imaging (CDI). While physics-driven neural networks have emerged as a promising approach through their unsupervised learning paradigm, their practical implementation faces critical challenges: measurement uncertainties in physical parameters (e.g., the propagation distance and the size of sample area) severely degrade reconstruction quality. To overcome this limitation, we propose a deep-learning-enabled spatial sample interval optimization framework that synergizes physical models with neural network adaptability. Our method embeds spatial sample intervals as trainable parameters within a PhysenNet architecture coupled with Fresnel diffraction physics, enabling simultaneous image reconstruction and system parameter calibration. Experimental validation demonstrates robust performance with structural similarity (SSIM) values consistently maintained at 0.6 across diffraction distances spanning of 10–200 mm, using a 1024 × 1024 region of interest (ROI) from a 1624 × 1440 CCD (pixel size: 4.5 μm) under 632.8 nm illumination. This framework has excellent fault tolerance, that is, it can still maintain high-quality image restoration even when the propagation distance measurement error is large. Compared to conventional iterative reconstruction algorithms, this approach can transform fixed parameters into learnable parameters, making almost all image restoration experiments easier to implement, enhancing system robustness against experimental uncertainties. This work establishes, to our knowledge, a new paradigm for adaptive diffraction imaging systems capable of operating in complex real scenarios.
Chinese Optics Letters
- Publication Date: Jun. 12, 2025
- Vol. 23, Issue 7, 071105 (2025)
Enhancing terahertz imaging with Rydberg atom-based sensors using untrained neural networks|Editors' Pick
Jun Wan, Bin Zhang, Xianzhe Li, Tao Li, Qirong Huang, Xinyu Yang, Kaiqing Zhang, Wei Huang, and Haixiao Deng
Terahertz (THz) imaging based on the Rydberg atom achieves high sensitivity and frame rates but faces challenges in spatial resolution due to diffraction, interference, and background noise. This study introduces a polarization filter and a deep learning-based method using a physically informed convolutional neural network to enhance resolution without pre-trained datasets. The technique reduces diffraction artifacts and achieves lens-free imaging with a resolution exceeding 1.25 lp/mm over a wide field of view. This advancement significantly improves the imaging quality of the Rydberg atom-based sensor, expanding its potential applications in THz imaging. Terahertz (THz) imaging based on the Rydberg atom achieves high sensitivity and frame rates but faces challenges in spatial resolution due to diffraction, interference, and background noise. This study introduces a polarization filter and a deep learning-based method using a physically informed convolutional neural network to enhance resolution without pre-trained datasets. The technique reduces diffraction artifacts and achieves lens-free imaging with a resolution exceeding 1.25 lp/mm over a wide field of view. This advancement significantly improves the imaging quality of the Rydberg atom-based sensor, expanding its potential applications in THz imaging.
Chinese Optics Letters
- Publication Date: Jun. 19, 2025
- Vol. 23, Issue 7, 071104 (2025)
Robust high-resolution imaging against translational motion by a Fourier-transform ghost diffraction technique
Long Zhang, Zongjun Li and Wenlin Gong
The relative motion between an imaging system and its target usually leads to image blurring. We propose a motion deblurring imaging system based on the Fourier-transform ghost diffraction (FGD) technique, which can overcome the spatial resolution degradation caused by both laterally and axially translational motion of the target. Both the analytical and experimental results demonstrate that when the effective transmission aperture of the receiving lens is larger than the target’s lateral motion amplitude and even if the target is located in the near-field region of the source, the amplitude and mode of the target’s motion have no effect on the quality of FGD, and high-resolution imaging in the spatial domain can be always achieved by the phase-retrieval method from the FGD patterns. Corresponding results based on the conventional Fourier diffraction system are also compared and discussed. The relative motion between an imaging system and its target usually leads to image blurring. We propose a motion deblurring imaging system based on the Fourier-transform ghost diffraction (FGD) technique, which can overcome the spatial resolution degradation caused by both laterally and axially translational motion of the target. Both the analytical and experimental results demonstrate that when the effective transmission aperture of the receiving lens is larger than the target’s lateral motion amplitude and even if the target is located in the near-field region of the source, the amplitude and mode of the target’s motion have no effect on the quality of FGD, and high-resolution imaging in the spatial domain can be always achieved by the phase-retrieval method from the FGD patterns. Corresponding results based on the conventional Fourier diffraction system are also compared and discussed.
Chinese Optics Letters
- Publication Date: Jun. 16, 2025
- Vol. 23, Issue 7, 071103 (2025)
Time-multiplexing non-line-of-sight imaging
Tailin Li, Xianmin Zheng, Kaiyuan Zhao, Min Li, Shiye Xia, Yaqing Liu, Ge Ren, and Yihan Luo
Non-line-of-sight (NLOS) imaging has potential in autonomous driving, robotic vision, and medical imaging, but it is hindered by extensive scans. In this work, we provide a time-multiplexing NLOS imaging scheme that is designed to reduce the number of scans on the relay surface. The approach introduces a time delay at the transmitting end, allowing two laser pulses with different delays to be sent per period and enabling simultaneous acquisition of data from multiple sampling points. Additionally, proof-of-concept experiments validate the feasibility of this approach, achieving reconstruction with half the scans. These results demonstrate a promising strategy for real-time NLOS imaging. Non-line-of-sight (NLOS) imaging has potential in autonomous driving, robotic vision, and medical imaging, but it is hindered by extensive scans. In this work, we provide a time-multiplexing NLOS imaging scheme that is designed to reduce the number of scans on the relay surface. The approach introduces a time delay at the transmitting end, allowing two laser pulses with different delays to be sent per period and enabling simultaneous acquisition of data from multiple sampling points. Additionally, proof-of-concept experiments validate the feasibility of this approach, achieving reconstruction with half the scans. These results demonstrate a promising strategy for real-time NLOS imaging.
Chinese Optics Letters
- Publication Date: Jun. 12, 2025
- Vol. 23, Issue 7, 071102 (2025)
Topics
© Copyright 2018-2021 | Chinese Laser Press. All Rights Reserved 沪ICP备15018463号-20