Yixiao Yang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Yixiao Yang has a total of 19 co-authors. Here are the top co-authors by number of shared publications:
- Ran Tao
School of Information and Electronics, Beijing Institute of Technology, Beijing, China
4 shared publications - Xudong Zhao
School of Information and Electronics, Beijing Institute of Technology, Beijing, China
3 shared publications - Xin Zhang
School of Information and Electronics, Beijing Institute of Technology, Zhuhai, China
2 shared publications - Zhuang Zhou
School of Information and Electronics, Beijing Institute of Technology, Zhuhai, China
2 shared publications Binghua Su
School of Information and Electronics, Beijing Institute of Technology, Zhuhai, China
1 shared publication- Binghua Su
Beijing Institute of Technology, Beijing, China and Beijing Institute of Technology, Zhuhai, Zhuhai, China
1 shared publication
Latest Publications
- 3 February 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Signal Processing Letters
- Vol. 33 (10709908) , 1121-1125
Dual-Domain Fractional Fourier Transformer for Underwater Image Degradation Removal
In this letter, we propose DEFriT, a dual-domain framework for underwater image restoration targeting degradation caused by scattering and absorption. Supervised training is enabled by simulating underwater degradation from land images using a physics-based formation model. To address the spectrally non-stationary nature of underwater attenuation, we introduce the fractional Fourier transform (FrFT) to bridge spatial and spectral representations. A Fractional Transform Block (FrTB) with real-imaginary decomposition and Hybrid Time-Frequency Self-Attention (HTFSA) supports joint spatial-spectral modeling. Empirical analysis shows that degradation mainly affect the amplitude spectrum, with learned fractional orders concentrated in a narrow range ($p \in [0.48, 0.52]$). To the best of our knowledge, this is the first work to introduce FrFT into underwater image restoration and to systematically establish a fractional order prior. DEFriT achieves state-of-the-art performance on five benchmark datasets.
- 3 February 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Signal Processing Letters
- Vol. 33 (10709908) , 1121-1125
Dual-Domain Fractional Fourier Transformer for Underwater Image Degradation Removal
In this letter, we propose DEFriT, a dual-domain framework for underwater image restoration targeting degradation caused by scattering and absorption. Supervised training is enabled by simulating underwater degradation from land images using a physics-based formation model. To address the spectrally non-stationary nature of underwater attenuation, we introduce the fractional Fourier transform (FrFT) to bridge spatial and spectral representations. A Fractional Transform Block (FrTB) with real-imaginary decomposition and Hybrid Time-Frequency Self-Attention (HTFSA) supports joint spatial-spectral modeling. Empirical analysis shows that degradation mainly affect the amplitude spectrum, with learned fractional orders concentrated in a narrow range ($p \in [0.48, 0.52]$). To the best of our knowledge, this is the first work to introduce FrFT into underwater image restoration and to systematically establish a fractional order prior. DEFriT achieves state-of-the-art performance on five benchmark datasets.
- Open Access
- 14 December 2025
- Published by Association for Computing Machinery (ACM)
- p. 1-11
Self-supervised Underwater Color Restoration via Wavelet-Diffusion Model with Filtered Multi-Scale Feature Distillation
- 14 December 2025
- Published by Association for Computing Machinery (ACM)
- p. 1-11
Open AccessSelf-supervised Underwater Color Restoration via Wavelet-Diffusion Model with Filtered Multi-Scale Feature Distillation
- 1 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Selected Topics in Signal Processing
- Vol. 19 (8) , 1710-1723
Deep Physically Parameterized All-in-One Network for Lens-Free Microscopy Imaging
- 1 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Selected Topics in Signal Processing
- Vol. 19 (8) , 1710-1723
Deep Physically Parameterized All-in-One Network for Lens-Free Microscopy Imaging
- 27 November 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Geoscience and Remote Sensing
- Vol. 63 (01962892) , 1-14
Chirplet Fourier Analysis Network for Cross-Scene Classification of Multisource Remote Sensing Data
The joint application of multisource remote sensing (MSRS) data, such as hyperspectral image (HSI) and light detection and ranging (LiDAR), offers significant potential for accurate land cover classification. However, the existing applications often struggle with domain shifts across scenes caused by sensor, illumination, and phase variations. Focusing on this domain adaptation problem, a Chirplet Fourier analysis network (ChirpFAN) is proposed for cross-scene classification of MSRS data in this paper. Firstly, a fractional spatial-frequency-phase feature extraction module including the fractional Fourier transform and a learnable phase-aware weighting block is proposed to capture multi-domain features. Secondly, a Chirplet swin transformer (ChirpST) block integrates a Chirplet Fourier analysis (ChirpFA) layer within a Swin transformer is designed to analyze multi-scale textural and oscillatory patterns. Finally, a modality-shared network including ChirpST blocks is designed for inter-modal fusion and alignment. Extensive experiments demonstrate that the ChirpFAN framework achieves state-of-the-art performance with 3% average improvements on three challenging cross-scene MSRS datasets. Code will be released on GitHub.
- 27 November 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Geoscience and Remote Sensing
- Vol. 63 (01962892) , 1-14
Chirplet Fourier Analysis Network for Cross-Scene Classification of Multisource Remote Sensing Data
The joint application of multisource remote sensing (MSRS) data, such as hyperspectral image (HSI) and light detection and ranging (LiDAR), offers significant potential for accurate land cover classification. However, the existing applications often struggle with domain shifts across scenes caused by sensor, illumination, and phase variations. Focusing on this domain adaptation problem, a Chirplet Fourier analysis network (ChirpFAN) is proposed for cross-scene classification of MSRS data in this paper. Firstly, a fractional spatial-frequency-phase feature extraction module including the fractional Fourier transform and a learnable phase-aware weighting block is proposed to capture multi-domain features. Secondly, a Chirplet swin transformer (ChirpST) block integrates a Chirplet Fourier analysis (ChirpFA) layer within a Swin transformer is designed to analyze multi-scale textural and oscillatory patterns. Finally, a modality-shared network including ChirpST blocks is designed for inter-modal fusion and alignment. Extensive experiments demonstrate that the ChirpFAN framework achieves state-of-the-art performance with 3% average improvements on three challenging cross-scene MSRS datasets. Code will be released on GitHub.
- Open Access
- 1 January 2025
- Published by ScienceOpen
- p. e390
Can Computational Phase Retrieval Help 2D Classification in Cryo-EM?
Cryo-electron microscopy (Cryo-EM) has emerged as a powerful tool for determining the structures of biomolecules, especially in cases where traditional methods like X-ray crystallography and nuclear magnetic resonance (NMR) fall short [ 1]. Its ability to visualize biomolecules in near-native states at cryogenic temperatures has revolutionized structural biology, leading to the resolution of complex molecular machines such as the ribosome and membrane proteins, which are critical targets for drug development. The recent “resolution revolution” [ 2] in Cryo-EM has made it possible to achieve atomic or near-atomic resolution for many large macromolecular assemblies. In Cryo-EM, a substantial volume of particle data is often excluded due to inadequate phase contrast, which impedes full utilization of these particles. This limited insight into discarded structures constrains our capacity to attain the utmost resolution. Furthermore, the 2D classification task in single-particle reconstruction proves especially challenging for heterogeneous samples, low-concentration or low-symmetry structures, and small (<250 kDa) or flexible molecules. To enhance phase contrast, a conventional approach involves integrating a phase-shifting device—such as Zernike or Volta phase plates—within the back focal plane of the objective lens. However, these devices are prone to high-frequency signal attenuation, variability in fabrication quality, reliability issues, and shortened operational lifespans due to electrostatic charging. In recent years, computational phase retrieval algorithms have advanced markedly and are now widely applied in coherent diffraction imaging, holographic imaging, and Fourier ptychographic microscopy. Some studies in Cryo-EM have also begun investigating the potential of computational phase retrieval to extend imaging contrast to the limits allowable by our understanding of the physics of contrast formation [ 3].
- 1 January 2025
- Published by ScienceOpen
- p. e390
Open AccessCan Computational Phase Retrieval Help 2D Classification in Cryo-EM?
Cryo-electron microscopy (Cryo-EM) has emerged as a powerful tool for determining the structures of biomolecules, especially in cases where traditional methods like X-ray crystallography and nuclear magnetic resonance (NMR) fall short [ 1]. Its ability to visualize biomolecules in near-native states at cryogenic temperatures has revolutionized structural biology, leading to the resolution of complex molecular machines such as the ribosome and membrane proteins, which are critical targets for drug development. The recent “resolution revolution” [ 2] in Cryo-EM has made it possible to achieve atomic or near-atomic resolution for many large macromolecular assemblies. In Cryo-EM, a substantial volume of particle data is often excluded due to inadequate phase contrast, which impedes full utilization of these particles. This limited insight into discarded structures constrains our capacity to attain the utmost resolution. Furthermore, the 2D classification task in single-particle reconstruction proves especially challenging for heterogeneous samples, low-concentration or low-symmetry structures, and small (<250 kDa) or flexible molecules. To enhance phase contrast, a conventional approach involves integrating a phase-shifting device—such as Zernike or Volta phase plates—within the back focal plane of the objective lens. However, these devices are prone to high-frequency signal attenuation, variability in fabrication quality, reliability issues, and shortened operational lifespans due to electrostatic charging. In recent years, computational phase retrieval algorithms have advanced markedly and are now widely applied in coherent diffraction imaging, holographic imaging, and Fourier ptychographic microscopy. Some studies in Cryo-EM have also begun investigating the potential of computational phase retrieval to extend imaging contrast to the limits allowable by our understanding of the physics of contrast formation [ 3].
- 5 July 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Signal Processing
- Vol. 72 (1053587X) , 3303-3317
Single-Shot Phase Retrieval From a Fractional Fourier Transform Perspective
The realm of classical phase retrieval concerns itself with the arduous task of recovering a signal from its Fourier transform magnitude, which is fraught with inherent ambiguities. A single-exposure intensity measurement is commonly deemed insufficient for the reconstruction of the primal signal, given that the absent phase component is imperative for the inverse transformation. In this work, we present a novel single-shot phase retrieval paradigm from a fractional Fourier transform (FrFT) perspective, which involves integrating the FrFT-based physical measurement model within a self-supervised reconstruction scheme. Specifically, the proposed FrFT-based measurement model addresses the aliasing artifacts problem in the numerical calculation of Fresnel diffraction, featuring adaptability to both short-distance and long-distance propagation scenarios. Moreover, the intensity measurement in the FrFT domain proves highly effective in alleviating the ambiguities of phase retrieval and relaxing the previous conditions on oversampled or multiple measurements in the Fourier domain. Furthermore, the proposed self-supervised reconstruction approach harnesses the fast discrete algorithm of FrFT alongside untrained neural network priors, thereby attaining preeminent results. Through numerical simulations, we demonstrate that both amplitude and phase objects can be effectively retrieved from single intensity measurement using the proposed approach and provide a promising technique for support-free coherent diffraction imaging.
- 5 July 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Signal Processing
- Vol. 72 (1053587X) , 3303-3317
Single-Shot Phase Retrieval From a Fractional Fourier Transform Perspective
The realm of classical phase retrieval concerns itself with the arduous task of recovering a signal from its Fourier transform magnitude, which is fraught with inherent ambiguities. A single-exposure intensity measurement is commonly deemed insufficient for the reconstruction of the primal signal, given that the absent phase component is imperative for the inverse transformation. In this work, we present a novel single-shot phase retrieval paradigm from a fractional Fourier transform (FrFT) perspective, which involves integrating the FrFT-based physical measurement model within a self-supervised reconstruction scheme. Specifically, the proposed FrFT-based measurement model addresses the aliasing artifacts problem in the numerical calculation of Fresnel diffraction, featuring adaptability to both short-distance and long-distance propagation scenarios. Moreover, the intensity measurement in the FrFT domain proves highly effective in alleviating the ambiguities of phase retrieval and relaxing the previous conditions on oversampled or multiple measurements in the Fourier domain. Furthermore, the proposed self-supervised reconstruction approach harnesses the fast discrete algorithm of FrFT alongside untrained neural network priors, thereby attaining preeminent results. Through numerical simulations, we demonstrate that both amplitude and phase objects can be effectively retrieved from single intensity measurement using the proposed approach and provide a promising technique for support-free coherent diffraction imaging.