Pu Huang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Pu Huang has a total of 27 co-authors. Here are the top co-authors by number of shared publications:
- Zhangjing Yang
School of Computer Science, Najing Audit University, Nanjing, China
5 shared publications Fanlong Zhang
School of Computer Science, Najing Audit University, Nanjing, China
3 shared publicationsGeng Yang
Nanjing University of Posts and Telecommunications
3 shared publicationsMinghua Wan
Nanjing Audit Univ, Sch Comp Sci, Sch Intelligent Auditing, Nanjing 211815, Peoples R China
3 shared publicationsZhangjing Yang
3 shared publicationsChengshan Qian
School of Computer Software, Nanjing University of Information Science and Technology, Nanjing, China
2 shared publications
Latest Publications
- 3 March 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Computational Imaging
- Vol. 12 (25730436) , 630-645
Gradient-Aware Gamma Adjustment for Robust Diffusion-Based Computational Imaging
Restoring images degraded by spatially variant distortions remains a fundamental challenge in computational imaging. While diffusion-based methods like Denoising Diffusion with Preconditioned Guidance (DDPG) offer efficient solvers, they typically rely on a static, globally fixed trade-off between Back-Projection (BP) and Least-Squares (LS) guidance. This static scheduling prevents the sampling process from adapting to heterogeneous local signal-to-noise characteristics, often leading to suboptimal restoration in complex or mixed-content regions. We propose Gradient-Aware Gamma Adjustment (GAGA), a spatially adaptive framework that modulates the preconditioned guidance in a pixel-wise manner. By leveraging multi-scale gradient cues extracted from intermediate reconstructions, GAGA dynamically adjusts the BP–LS balance to better accommodate local structural variations. From a theoretical perspective, we refine the LS guidance by explicitly aligning it with the Gaussian likelihood score via the noise-dependent scaling factor $\sigma _{e}^{-2}$. We further show that the resulting spatially adaptive preconditioner remains strictly positive definite under bounded modulation, guaranteeing well-posed and numerically stable guidance for ill-posed inverse problems. Extensive experiments on CelebA-HQ and ImageNet, as well as on spatially variant degradations and computational imaging tasks including super-resolution, inpainting, CT, and MRI reconstruction, demonstrate consistent improvements in both fidelity (PSNR) and perceptual quality (LPIPS) over static preconditioned baselines, without introducing noticeable computational overhead.
- 3 March 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Computational Imaging
- Vol. 12 (25730436) , 630-645
Gradient-Aware Gamma Adjustment for Robust Diffusion-Based Computational Imaging
Restoring images degraded by spatially variant distortions remains a fundamental challenge in computational imaging. While diffusion-based methods like Denoising Diffusion with Preconditioned Guidance (DDPG) offer efficient solvers, they typically rely on a static, globally fixed trade-off between Back-Projection (BP) and Least-Squares (LS) guidance. This static scheduling prevents the sampling process from adapting to heterogeneous local signal-to-noise characteristics, often leading to suboptimal restoration in complex or mixed-content regions. We propose Gradient-Aware Gamma Adjustment (GAGA), a spatially adaptive framework that modulates the preconditioned guidance in a pixel-wise manner. By leveraging multi-scale gradient cues extracted from intermediate reconstructions, GAGA dynamically adjusts the BP–LS balance to better accommodate local structural variations. From a theoretical perspective, we refine the LS guidance by explicitly aligning it with the Gaussian likelihood score via the noise-dependent scaling factor $\sigma _{e}^{-2}$. We further show that the resulting spatially adaptive preconditioner remains strictly positive definite under bounded modulation, guaranteeing well-posed and numerically stable guidance for ill-posed inverse problems. Extensive experiments on CelebA-HQ and ImageNet, as well as on spatially variant degradations and computational imaging tasks including super-resolution, inpainting, CT, and MRI reconstruction, demonstrate consistent improvements in both fidelity (PSNR) and perceptual quality (LPIPS) over static preconditioned baselines, without introducing noticeable computational overhead.
- 17 December 2025
- Published by Elsevier in Expert Systems with Applications
- Vol. 305
Structural enhanced robust discriminative least squares regression for image classification
- 17 December 2025
- Published by Elsevier in Expert Systems with Applications
- Vol. 305
Structural enhanced robust discriminative least squares regression for image classification
- 26 February 2025
- Published by Elsevier in Expert Systems with Applications
- Vol. 275
Mixed granularity network for person re-identification
- 26 February 2025
- Published by Elsevier in Expert Systems with Applications
- Vol. 275
Mixed granularity network for person re-identification
- 1 July 2024
- Published by Elsevier in Information Sciences
- Vol. 680
Adversarial compact wrapping classifier learning for open set recognition
- 1 July 2024
- Published by Elsevier in Information Sciences
- Vol. 680
Adversarial compact wrapping classifier learning for open set recognition
- 18 May 2024
- Published by Elsevier in Expert Systems with Applications
- Vol. 252
Regularisation constrained denoising discriminant least squares regression for image classification
- 18 May 2024
- Published by Elsevier in Expert Systems with Applications
- Vol. 252
Regularisation constrained denoising discriminant least squares regression for image classification
- 9 May 2024
- Published by Elsevier in Information Sciences
- Vol. 673
Cascaded maximum median-margin discriminant projection with its application to face recognition
- 9 May 2024
- Published by Elsevier in Information Sciences
- Vol. 673
Cascaded maximum median-margin discriminant projection with its application to face recognition
- 4 November 2023
- Published by Elsevier in Computers and Electrical Engineering
- Vol. 112
Two phase representation discriminant projection for face recognition
- 4 November 2023
- Published by Elsevier in Computers and Electrical Engineering
- Vol. 112
Two phase representation discriminant projection for face recognition
- Open Access
- 7 July 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
- Vol. 14 (19391404) , 7310-7323
Triple-Regularized Latent Subspace Discriminative Regression for Hyperspectral Image Classification
There are various types and distributions of noise in hyperspectral images. However, the existing classification models are vulnerable to noise in the data. To improve the robustness of the classification models, a novel hyperspectral image classification model is proposed, named triple-regularized latent subspace discriminative regression (TRLSDR). The core idea of TRLSDR is to add a latent subspace to the standard discriminative least squares regression model to extract high-order features from the visual space by undercomplete autoencoder, and then use clean data for classification. Three regularizers are introduced in this process: Tikhonov regularizer is used to avoid overfitting; Laplacian regularizer is used to capture the neighborhood relationship; and low-rank regularizer is used to alleviate the error of Laplacian matrix construction caused by noise in original samples. We designed experiments on five hyperspectral image datasets and the results show that the proposed model is superior to the existing regression models.
- 7 July 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
- Vol. 14 (19391404) , 7310-7323
Open AccessTriple-Regularized Latent Subspace Discriminative Regression for Hyperspectral Image Classification
There are various types and distributions of noise in hyperspectral images. However, the existing classification models are vulnerable to noise in the data. To improve the robustness of the classification models, a novel hyperspectral image classification model is proposed, named triple-regularized latent subspace discriminative regression (TRLSDR). The core idea of TRLSDR is to add a latent subspace to the standard discriminative least squares regression model to extract high-order features from the visual space by undercomplete autoencoder, and then use clean data for classification. Three regularizers are introduced in this process: Tikhonov regularizer is used to avoid overfitting; Laplacian regularizer is used to capture the neighborhood relationship; and low-rank regularizer is used to alleviate the error of Laplacian matrix construction caused by noise in original samples. We designed experiments on five hyperspectral image datasets and the results show that the proposed model is superior to the existing regression models.
- 22 June 2018
- Published by Springer Nature in Soft Computing
- Vol. 23 (16) , 7015-7028
Feature extraction based on graph discriminant embedding and its applications to face recognition
- 22 June 2018
- Published by Springer Nature in Soft Computing
- Vol. 23 (16) , 7015-7028
Feature extraction based on graph discriminant embedding and its applications to face recognition
- Open Access
- 9 March 2017
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 5, 4340-4349
Fuzzy Linear Regression Discriminant Projection for Face Recognition
How to capture distinctive features from facial images when there are large variations in illumination, poses, and expressions is important for the face recognition problems. This paper introduces a novel algorithm called fuzzy linear regression discriminant projection (FLRDP) for face recognition. The proposed algorithm FLRDP seeks to generate an efficient subspace for the LRC method and could effectively handle variations between facial images. To be specific, FLRDP first computes the gradual membership degrees of each sample to corresponding classes, and then incorporates such membership degree information into the construction of the fuzzy between-class and within-class reconstruction errors. Finally, the criterion function is derived via maximizing the ratio of the fuzzy between-class reconstruction error to the fuzzy within-class reconstruction error. Experimental results carried out on the ORL, CMU PIE, and FERET face databases show the superiority of our proposed method over other state-of-the-art algorithms.
- 9 March 2017
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 5, 4340-4349
Open AccessFuzzy Linear Regression Discriminant Projection for Face Recognition
How to capture distinctive features from facial images when there are large variations in illumination, poses, and expressions is important for the face recognition problems. This paper introduces a novel algorithm called fuzzy linear regression discriminant projection (FLRDP) for face recognition. The proposed algorithm FLRDP seeks to generate an efficient subspace for the LRC method and could effectively handle variations between facial images. To be specific, FLRDP first computes the gradual membership degrees of each sample to corresponding classes, and then incorporates such membership degree information into the construction of the fuzzy between-class and within-class reconstruction errors. Finally, the criterion function is derived via maximizing the ratio of the fuzzy between-class reconstruction error to the fuzzy within-class reconstruction error. Experimental results carried out on the ORL, CMU PIE, and FERET face databases show the superiority of our proposed method over other state-of-the-art algorithms.