Xuejun Huang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Xuejun Huang has a total of 17 co-authors. Here are the top co-authors by number of shared publications:
- Jinshan Ding
National Key Laboratory of Radar Signal Processing, Xidian University, Xian, China
11 shared publications - Liwu Wen
National Key Laboratory of Radar Signal Processing, Xidian University, Xi’an 710071, China
6 shared publications - Zhong Xu
National Key Laboratory of Radar Signal Processing, Xidian University, Xian, China
3 shared publications - Chao Zhong
National Key Laboratory of Radar Signal Processing, Xidian University, Xi’an 710071, China
2 shared publications - Yan Zhang
National Key Laboratory of Radar Signal Processing, Xidian University, Xi’an 710071, China
2 shared publications - Chao Xue
Harbin Engineering University
1 shared publication
Latest Publications
- Open Access
- 24 February 2026
- Published by MDPI AG in Remote Sensing
- Vol. 18 (5) , 670
Video SAR Enhanced Imaging Using a Self-Supervised Super-Resolution Reconstruction Network
Video synthetic aperture radar (SAR) enables observation of moving targets by leveraging temporal information across successive frames. In particular, dynamic shadows in video SAR image sequences provide critical cues for detecting moving objects whose energy is smeared or Doppler-shifted. To achieve high-resolution imaging at a high frame rate for effective dynamic scene monitoring, video SAR systems typically operate at extremely high frequencies or even in the terahertz band, rather than the microwave band. However, terahertz video SAR suffers from significant signal attenuation due to atmospheric absorption. We present a deep learning framework to achieve high-frame-rate and high-resolution imaging for microwave video SAR systems. In this framework, the problem of microwave video SAR imaging is formulated as an image super-resolution reconstruction task for low-resolution yet high-frame-rate image sequences from microwave video SAR. We develop a simple yet effective image super-resolution reconstruction network that is completely built upon convolutional neural networks. The designed network takes a low-resolution image sequence and the corresponding high-resolution image with blurred shadows as input, and then produces a high-resolution image sequence where shadows are clearly visible. Furthermore, the network is trained in a self-supervised manner and thus does not require high-resolution image sequences with unblurred shadows as ground truth, which is appealing to practical applications. Processing results of real data from two different video SAR systems have shown good performance of the proposed approach with convincing generalization ability.
- 24 February 2026
- Published by MDPI AG in Remote Sensing
- Vol. 18 (5) , 670
Open AccessVideo SAR Enhanced Imaging Using a Self-Supervised Super-Resolution Reconstruction Network
Video synthetic aperture radar (SAR) enables observation of moving targets by leveraging temporal information across successive frames. In particular, dynamic shadows in video SAR image sequences provide critical cues for detecting moving objects whose energy is smeared or Doppler-shifted. To achieve high-resolution imaging at a high frame rate for effective dynamic scene monitoring, video SAR systems typically operate at extremely high frequencies or even in the terahertz band, rather than the microwave band. However, terahertz video SAR suffers from significant signal attenuation due to atmospheric absorption. We present a deep learning framework to achieve high-frame-rate and high-resolution imaging for microwave video SAR systems. In this framework, the problem of microwave video SAR imaging is formulated as an image super-resolution reconstruction task for low-resolution yet high-frame-rate image sequences from microwave video SAR. We develop a simple yet effective image super-resolution reconstruction network that is completely built upon convolutional neural networks. The designed network takes a low-resolution image sequence and the corresponding high-resolution image with blurred shadows as input, and then produces a high-resolution image sequence where shadows are clearly visible. Furthermore, the network is trained in a self-supervised manner and thus does not require high-resolution image sequences with unblurred shadows as ground truth, which is appealing to practical applications. Processing results of real data from two different video SAR systems have shown good performance of the proposed approach with convincing generalization ability.
- 16 February 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Antennas and Wireless Propagation Letters
- Vol. 25 (4) , 1792-1796
Joint Multiframe Superresolution Reconstruction Approach for Video Imaging Based on Information Metasurfaces
Information metasurfaces offer a low-cost solution for microwave forward-looking imaging. However, imaging dynamic and non-sparse scenes using information metasurfaces remains challenging. To address this issue, we present a joint multi-frame super-resolution reconstruction approach for video imaging using information metasurfaces. By leveraging the ultra-fast code-switching capability of information metasurfaces, we formulate the video imaging problem as a joint multi-frame super-resolution reconstruction task. Furthermore, by exploiting the temporal correlation of the background and the sparsity of dynamic variation across multiple frames, we introduce a low-rank and sparse decomposition with inter-frame smoothness model for the super-resolution reconstruction. An effective algorithm based on the proximal gradient method is developed to solve the proposed model, enabling high-resolution imaging of dynamic and non-sparse scenes. Simulation results demonstrate that the proposed approach achieves superior reconstruction performance compared to the existing computational imaging methods.
- 16 February 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Antennas and Wireless Propagation Letters
- Vol. 25 (4) , 1792-1796
Joint Multiframe Superresolution Reconstruction Approach for Video Imaging Based on Information Metasurfaces
Information metasurfaces offer a low-cost solution for microwave forward-looking imaging. However, imaging dynamic and non-sparse scenes using information metasurfaces remains challenging. To address this issue, we present a joint multi-frame super-resolution reconstruction approach for video imaging using information metasurfaces. By leveraging the ultra-fast code-switching capability of information metasurfaces, we formulate the video imaging problem as a joint multi-frame super-resolution reconstruction task. Furthermore, by exploiting the temporal correlation of the background and the sparsity of dynamic variation across multiple frames, we introduce a low-rank and sparse decomposition with inter-frame smoothness model for the super-resolution reconstruction. An effective algorithm based on the proximal gradient method is developed to solve the proposed model, enabling high-resolution imaging of dynamic and non-sparse scenes. Simulation results demonstrate that the proposed approach achieves superior reconstruction performance compared to the existing computational imaging methods.
- Open Access
- 20 January 2026
- Published by MDPI AG in Remote Sensing
- Vol. 18 (2) , 343
Shadow Spatiotemporal Track-Before-Detect Approach for Distributed UAV-Borne Video SAR
Shadow detection has become a key technology for ground-based moving target indication in video synthetic aperture radar (SAR). However, single-platform video SAR faces the issue of moving-target shadows being occluded. This paper proposes a new dynamic programming-based spatiotemporal track-before-detect (DP-ST-TBD) algorithm for moving-target shadow indication based on a distributed unmanned aerial vehicle (UAV)-borne video SAR system. First, this approach establishes a spatiotemporal cooperative shadow detection model, which extends the temporal accumulation of traditional DP-TBD to spatiotemporal accumulation by state temporal transition and spatial mapping. Second, an adaptive state transition method is proposed to address the challenge in which the fixed-state transition of traditional DP-TBD struggles with maneuvering target detection. It utilizes target’s Doppler features from heterogeneous-view range-Doppler (RD) spectra to assist in target’s shadow search within the image domain. Finally, a state shrinking–sparseness strategy is used to reduce the computational burden caused by dense states in spatiotemporal search; thus, multi-platform, multi-frame accumulation of moving-target shadows can be realized based on sparse states. The comparative experiments demonstrate that the proposed DP-ST-TBD improves shadow-detection performance through heterogeneous-view measurements while reducing the required number of frames for reliable detection compared to the conventional two-step detection method (single-platform shadow detection followed by multi-platform track fusion).
- 20 January 2026
- Published by MDPI AG in Remote Sensing
- Vol. 18 (2) , 343
Open AccessShadow Spatiotemporal Track-Before-Detect Approach for Distributed UAV-Borne Video SAR
Shadow detection has become a key technology for ground-based moving target indication in video synthetic aperture radar (SAR). However, single-platform video SAR faces the issue of moving-target shadows being occluded. This paper proposes a new dynamic programming-based spatiotemporal track-before-detect (DP-ST-TBD) algorithm for moving-target shadow indication based on a distributed unmanned aerial vehicle (UAV)-borne video SAR system. First, this approach establishes a spatiotemporal cooperative shadow detection model, which extends the temporal accumulation of traditional DP-TBD to spatiotemporal accumulation by state temporal transition and spatial mapping. Second, an adaptive state transition method is proposed to address the challenge in which the fixed-state transition of traditional DP-TBD struggles with maneuvering target detection. It utilizes target’s Doppler features from heterogeneous-view range-Doppler (RD) spectra to assist in target’s shadow search within the image domain. Finally, a state shrinking–sparseness strategy is used to reduce the computational burden caused by dense states in spatiotemporal search; thus, multi-platform, multi-frame accumulation of moving-target shadows can be realized based on sparse states. The comparative experiments demonstrate that the proposed DP-ST-TBD improves shadow-detection performance through heterogeneous-view measurements while reducing the required number of frames for reliable detection compared to the conventional two-step detection method (single-platform shadow detection followed by multi-platform track fusion).
- Open Access
- 6 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Radar Systems
- Vol. 4, 280-295
Efficient Multitarget Tracking Approach for UAV-Borne Video SAR
Shadow-based moving target tracking is essential for unmanned aerial vehicle (UAV)-borne video synthetic aperture radar (ViSAR). Traditional detect-before-track (DBT) algorithms are ineffective for dim shadows, while track-before-detect (TBD) algorithms incur prohibitive computational costs. Moreover, improving tracking accuracy for dim shadows often increases false alarms in cluttered areas. This paper presents an adaptive shadow-aware tracking (ASAT) algorithm and its implementation for low-power edge deployment. The algorithm adaptively extracts the moving target shadow aggregation area (MSAA) from current frames using prior tracking results. Shadows within the MSAA are classified into strong and dim categories using multi-value constant false alarm rate detection and morphological processing. Dim shadow traces are then processed by a TBD-based algorithm, while strong shadow traces are handled by a DBT-based algorithm. Together, adaptive MSAA extraction and subsequent classification-based processing form a joint strategy that enhances tracking accuracy and improves real-time performance on low-power edge platforms. The ASAT algorithm is deployed on a low-power edge system-on-module featuring a multi-core central processing unit and an embedded graphics processing unit. To enable low-power edge–oriented algorithm design and lightweight implementation, an algorithm–hardware co-design methodology is introduced. Experiments on UAV-borne ViSAR data demonstrate that the proposed approach effectively outperforms traditional approaches in both tracking accuracy and real-time performance on UAV-borne low-power edge platforms.
- 6 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Radar Systems
- Vol. 4, 280-295
Open AccessEfficient Multitarget Tracking Approach for UAV-Borne Video SAR
Shadow-based moving target tracking is essential for unmanned aerial vehicle (UAV)-borne video synthetic aperture radar (ViSAR). Traditional detect-before-track (DBT) algorithms are ineffective for dim shadows, while track-before-detect (TBD) algorithms incur prohibitive computational costs. Moreover, improving tracking accuracy for dim shadows often increases false alarms in cluttered areas. This paper presents an adaptive shadow-aware tracking (ASAT) algorithm and its implementation for low-power edge deployment. The algorithm adaptively extracts the moving target shadow aggregation area (MSAA) from current frames using prior tracking results. Shadows within the MSAA are classified into strong and dim categories using multi-value constant false alarm rate detection and morphological processing. Dim shadow traces are then processed by a TBD-based algorithm, while strong shadow traces are handled by a DBT-based algorithm. Together, adaptive MSAA extraction and subsequent classification-based processing form a joint strategy that enhances tracking accuracy and improves real-time performance on low-power edge platforms. The ASAT algorithm is deployed on a low-power edge system-on-module featuring a multi-core central processing unit and an embedded graphics processing unit. To enable low-power edge–oriented algorithm design and lightweight implementation, an algorithm–hardware co-design methodology is introduced. Experiments on UAV-borne ViSAR data demonstrate that the proposed approach effectively outperforms traditional approaches in both tracking accuracy and real-time performance on UAV-borne low-power edge platforms.
- Open Access
- 30 December 2025
- Published by MDPI AG in Preprints.org
Video SAR Enhanced Imaging Using Self-Supervised Super-Resolution Reconstruction Network
Video synthetic aperture radar (SAR) enables observation of moving targets by leveraging temporal information across successive frames. In particular, dynamic shadows in video SAR image sequences provide critical cues for detecting moving objects whose energy is smeared or Doppler-shifted. To achieve high-resolution imaging at a high frame rate for effective dynamic scene monitoring, video SAR systems typically operate at extremely high frequencies or even in the terahertz band, rather than the microwave band. However, terahertz video SAR suffers from significant signal attenuation due to atmospheric absorption. We present a deep learning framework for high-frame-rate and high-resolution imaging with microwave video SAR system. In this framework, the problem of microwave video SAR imaging is formulated as an image super-resolution reconstruction task for low-resolution yet high-frame-rate image sequences from microwave video SAR. We develop a simple yet effective image super-resolution reconstruction network that is completely built upon convolutional neural networks. The designed network takes a low-resolution image sequence and the corresponding high-resolution image with blurred shadows as input, and then produces a high-resolution image sequence where shadows are clearly visible. Furthermore, the network is trained in a self-supervised manner and thus does not require desired high-resolution image sequences as ground truth, which is appealing to practical applications. Processing results of real data from two different video SAR systems have shown good performance of the proposed approach with convincing generalization ability.
- 30 December 2025
- Published by MDPI AG in Preprints.org
Open AccessVideo SAR Enhanced Imaging Using Self-Supervised Super-Resolution Reconstruction Network
Video synthetic aperture radar (SAR) enables observation of moving targets by leveraging temporal information across successive frames. In particular, dynamic shadows in video SAR image sequences provide critical cues for detecting moving objects whose energy is smeared or Doppler-shifted. To achieve high-resolution imaging at a high frame rate for effective dynamic scene monitoring, video SAR systems typically operate at extremely high frequencies or even in the terahertz band, rather than the microwave band. However, terahertz video SAR suffers from significant signal attenuation due to atmospheric absorption. We present a deep learning framework for high-frame-rate and high-resolution imaging with microwave video SAR system. In this framework, the problem of microwave video SAR imaging is formulated as an image super-resolution reconstruction task for low-resolution yet high-frame-rate image sequences from microwave video SAR. We develop a simple yet effective image super-resolution reconstruction network that is completely built upon convolutional neural networks. The designed network takes a low-resolution image sequence and the corresponding high-resolution image with blurred shadows as input, and then produces a high-resolution image sequence where shadows are clearly visible. Furthermore, the network is trained in a self-supervised manner and thus does not require desired high-resolution image sequences as ground truth, which is appealing to practical applications. Processing results of real data from two different video SAR systems have shown good performance of the proposed approach with convincing generalization ability.
- 18 August 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Antennas and Wireless Propagation Letters
- Vol. 24 (10) , 3634-3638
Efficient Attributed Scattering Center Extraction Using Gradient-Based Optimization for SAR Image
The attributed scattering center (ASC) model provides a concise representation of high-frequency electromagnetic scattering of targets, making it widely applicable to synthetic aperture radar target characteristic analysis and recognition. However, ASC extraction, as a non-convex parameter estimation problem, is computationally time-consuming in global optimization due to the non-convexity of loss function. While gradient-based optimization methods in deep learning frameworks can significantly improve efficiency, applying them directly to non-convex optimization remains challenging. To solve this issue, this letter proposes an efficient gradient-based ASC extraction method. First, A more convex-like loss function is designed using windowing techniques to broaden the convergence region and reduce the difficulty of non-convex optimization. Then, an image-frequency domain joint parameter initialization and estimation strategy are developed, achieving fully gradient-based parameter estimation. Validation on simulated and measured data confirms the effectiveness of the proposed method. While maintaining accuracy, it improves ASC extraction efficiency for tank targets from a minute-level to a second-level.
- 18 August 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Antennas and Wireless Propagation Letters
- Vol. 24 (10) , 3634-3638
Efficient Attributed Scattering Center Extraction Using Gradient-Based Optimization for SAR Image
The attributed scattering center (ASC) model provides a concise representation of high-frequency electromagnetic scattering of targets, making it widely applicable to synthetic aperture radar target characteristic analysis and recognition. However, ASC extraction, as a non-convex parameter estimation problem, is computationally time-consuming in global optimization due to the non-convexity of loss function. While gradient-based optimization methods in deep learning frameworks can significantly improve efficiency, applying them directly to non-convex optimization remains challenging. To solve this issue, this letter proposes an efficient gradient-based ASC extraction method. First, A more convex-like loss function is designed using windowing techniques to broaden the convergence region and reduce the difficulty of non-convex optimization. Then, an image-frequency domain joint parameter initialization and estimation strategy are developed, achieving fully gradient-based parameter estimation. Validation on simulated and measured data confirms the effectiveness of the proposed method. While maintaining accuracy, it improves ASC extraction efficiency for tank targets from a minute-level to a second-level.
- 8 February 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Geoscience and Remote Sensing
- Vol. 62 (01962892) , 1-12
High Frame-Rate Imaging Using Swarm of UAV-Borne Radars
High frame-rate imaging of synthetic aperture radar (SAR), known as video SAR, has received much research interest these years. It usually operates at extremely high frequency and even THz band as a technical tradeoff between high frame rate and high resolution. As a result, video SAR system always suffers from limited functional range due to strong atmospheric attenuation of signals. This article attempts to present a new high frame-rate collaborative imaging regime in microwave frequency band based on swarm of unmanned aerial vehicles (UAVs). The spatial degrees of freedom are employed to shorten the synthetic time and thus improve the frame rate. More specifically, the long synthetic aperture is split into multiple short sub-apertures, and each UAV-borne radar implements short sub-aperture imaging in a short time. Then, the accelerated fast back-projection algorithm is employed to fuse multiple sub-images to produce an image with high azimuth resolution.To implement the collaborative working of swarm of UAV-borne radars, a suitable orthogonal waveform is selected and a useful spatial configuration of the swarm is designed to compensate for the effect of the orthogonal waveform on imaging. Simulation results have been presented to highlight the advantages of collaborative imaging using swarm of UAV-borne radars.
- 8 February 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Geoscience and Remote Sensing
- Vol. 62 (01962892) , 1-12
High Frame-Rate Imaging Using Swarm of UAV-Borne Radars
High frame-rate imaging of synthetic aperture radar (SAR), known as video SAR, has received much research interest these years. It usually operates at extremely high frequency and even THz band as a technical tradeoff between high frame rate and high resolution. As a result, video SAR system always suffers from limited functional range due to strong atmospheric attenuation of signals. This article attempts to present a new high frame-rate collaborative imaging regime in microwave frequency band based on swarm of unmanned aerial vehicles (UAVs). The spatial degrees of freedom are employed to shorten the synthetic time and thus improve the frame rate. More specifically, the long synthetic aperture is split into multiple short sub-apertures, and each UAV-borne radar implements short sub-aperture imaging in a short time. Then, the accelerated fast back-projection algorithm is employed to fuse multiple sub-images to produce an image with high azimuth resolution.To implement the collaborative working of swarm of UAV-borne radars, a suitable orthogonal waveform is selected and a useful spatial configuration of the swarm is designed to compensate for the effect of the orthogonal waveform on imaging. Simulation results have been presented to highlight the advantages of collaborative imaging using swarm of UAV-borne radars.
- 16 November 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Geoscience and Remote Sensing Letters
- Vol. 19 (1545598X) , 1-5
Real-Time Super-Resolution ISAR Imaging Using Unsupervised Learning
Compressive sensing (CS) enables high-resolution inverse synthetic aperture radar (ISAR) imaging with limited measurements. However, these methods reconstruct images via iterative optimization, resulting in a high computational load. Recently, convolutional neural networks (CNNs) have been used to perform super-resolution ISAR imaging in real time, where high-resolution images are necessarily used as ground truth. However, the desired high-resolution images are not reliable in practice. This letter presents an unsupervised CNN-based framework for super-resolution ISAR imaging. The well-trained CNN can directly produce high-resolution ISAR images in real time. Moreover, the network is trained in an unsupervised manner, which is suitable for practical applications. Furthermore, a pseudo $\ell _{0}$ -norm has been used as the sparse constraint for the exact image reconstruction. The proposed approach has been used to process the real ISAR data, and the experimental results are convincing.
- 16 November 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Geoscience and Remote Sensing Letters
- Vol. 19 (1545598X) , 1-5
Real-Time Super-Resolution ISAR Imaging Using Unsupervised Learning
Compressive sensing (CS) enables high-resolution inverse synthetic aperture radar (ISAR) imaging with limited measurements. However, these methods reconstruct images via iterative optimization, resulting in a high computational load. Recently, convolutional neural networks (CNNs) have been used to perform super-resolution ISAR imaging in real time, where high-resolution images are necessarily used as ground truth. However, the desired high-resolution images are not reliable in practice. This letter presents an unsupervised CNN-based framework for super-resolution ISAR imaging. The well-trained CNN can directly produce high-resolution ISAR images in real time. Moreover, the network is trained in an unsupervised manner, which is suitable for practical applications. Furthermore, a pseudo $\ell _{0}$ -norm has been used as the sparse constraint for the exact image reconstruction. The proposed approach has been used to process the real ISAR data, and the experimental results are convincing.
- 26 July 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 3570-3575
Robust Elliptically Contoured Distribution Based Kalman Filter for INS/DVL Integrated Navigation
In complex underwater environment, the outputs of Doppler velocity log (DVL) often contain outliers induced by external disturbance or soft failure, which will degrade the performance of the traditional Kalman filter based INS/DVL integrated navigation system. To address this problem, a novel robust Kalman filter with outlier-contaminated measurement noise modelled as elliptically contoured distributed is proposed. Firstly, an analytical update form of the joint posterior probability density function is derived via introducing an unknown parameter. Then the fixed-point iteration method is exploited to achieve an approximate estimate of the unknown parameter, based on which the required joint posterior probability density function is analytically updated. Finally, simulation results of INS/DVL integrated navigation are used to verify the validity and advantages of the proposed method.
- 26 July 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 3570-3575
Robust Elliptically Contoured Distribution Based Kalman Filter for INS/DVL Integrated Navigation
In complex underwater environment, the outputs of Doppler velocity log (DVL) often contain outliers induced by external disturbance or soft failure, which will degrade the performance of the traditional Kalman filter based INS/DVL integrated navigation system. To address this problem, a novel robust Kalman filter with outlier-contaminated measurement noise modelled as elliptically contoured distributed is proposed. Firstly, an analytical update form of the joint posterior probability density function is derived via introducing an unknown parameter. Then the fixed-point iteration method is exploited to achieve an approximate estimate of the unknown parameter, based on which the required joint posterior probability density function is analytically updated. Finally, simulation results of INS/DVL integrated navigation are used to verify the validity and advantages of the proposed method.
- 25 December 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Geoscience and Remote Sensing
- Vol. 59 (12) , 10151-10160
Video SAR Image Despeckling by Unsupervised Learning
It has long been recognized that synthetic aperture radar (SAR) images suffer in many applications from the speckle noise. Video SAR has a high frame rate of imaging and contains redundant information among frames. The temporal redundancy in video SAR images has been found useful for suppressing the speckle noise. However, the motion and the local differences between frames make it difficult to employ the temporal redundancy to suppress speckle noise. This article presents a video SAR image despeckling framework based on a new unsupervised training strategy referred to as DualNoise2Noise. This developed framework consists of a registration network and a denoising network. The registration network first compensates for the motion between two adjacent frames in video SAR in real time. After image registration, two adjacent frames with random speckle noise can be considered as the observations of the same region with local differences. The denoising network adopts the DualNoise2Noise training strategy to suppress speckle noise by using the temporal redundancy and to remove the negative impact caused by the local differences. The proposed approach has been used to process the real video SAR data, and the experimental results are convincing.
- 25 December 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Geoscience and Remote Sensing
- Vol. 59 (12) , 10151-10160
Video SAR Image Despeckling by Unsupervised Learning
It has long been recognized that synthetic aperture radar (SAR) images suffer in many applications from the speckle noise. Video SAR has a high frame rate of imaging and contains redundant information among frames. The temporal redundancy in video SAR images has been found useful for suppressing the speckle noise. However, the motion and the local differences between frames make it difficult to employ the temporal redundancy to suppress speckle noise. This article presents a video SAR image despeckling framework based on a new unsupervised training strategy referred to as DualNoise2Noise. This developed framework consists of a registration network and a denoising network. The registration network first compensates for the motion between two adjacent frames in video SAR in real time. After image registration, two adjacent frames with random speckle noise can be considered as the observations of the same region with local differences. The denoising network adopts the DualNoise2Noise training strategy to suppress speckle noise by using the temporal redundancy and to remove the negative impact caused by the local differences. The proposed approach has been used to process the real video SAR data, and the experimental results are convincing.