Junqi Huang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Junqi Huang has a total of 6 co-authors. Here are the top co-authors by number of shared publications:
- Xianfeng Gu
Department of Computer Science, Stony Brook University, Stony Brook, USA
2 shared publications - Xiang Gao
Department of Computer Science, Stony Brook University, Stony Brook, USA
2 shared publications - Xinmu Wang
Department of Computer Science, Stony Brook University, Stony Brook, USA
2 shared publications - Fumi Honda
Department of Computer Science, Stony Brook University, Stony Brook, USA
1 shared publication - Wei Chen
Academy for Multidisciplinary Studies Capital Normal University Beijing China
1 shared publication - Zhou Zhao
Department of Computer Science, Stony Brook University, NY, USA
1 shared publication
Latest Publications
- 1 July 2026
- Published by Wiley in Computer Animation and Virtual Worlds
- Vol. 37 (4) , e70173
HoloDiff: Holistic Talking Human Animation Via Latent Diffusion
Holistic talking human animation aims to synthesize coordinated facial expressions and body poses that are consistent with spoken content and human communication patterns. This task requires both realistic motion generation and accurate temporal alignment between speech and motion, which remain challenging for existing methods that model facial and bodily motion separately or operate in high‐dimensional motion spaces. We present HoloDiff, a latent diffusion framework for holistic talking human animation. Given an input speech signal, HoloDiff encodes facial and bodily dynamics into a compact and structured latent representation that preserves anatomical structure and cross‐part interactions. A speech‐conditioned diffusion process is then applied in this latent space to model long‐range temporal dynamics and generate motion sequences aligned with audio input. By performing diffusion in a structured latent space, HoloDiff reduces modeling complexity while improving temporal coherence and audio–motion alignment. Extensive experiments on the SHOW dataset demonstrate that HoloDiff produces more realistic and expressive talking human animation and achieves better multimodal correspondence between speech and motion compared to baselines.
- 1 July 2026
- Published by Wiley in Computer Animation and Virtual Worlds
- Vol. 37 (4) , e70173
HoloDiff: Holistic Talking Human Animation Via Latent Diffusion
Holistic talking human animation aims to synthesize coordinated facial expressions and body poses that are consistent with spoken content and human communication patterns. This task requires both realistic motion generation and accurate temporal alignment between speech and motion, which remain challenging for existing methods that model facial and bodily motion separately or operate in high‐dimensional motion spaces. We present HoloDiff, a latent diffusion framework for holistic talking human animation. Given an input speech signal, HoloDiff encodes facial and bodily dynamics into a compact and structured latent representation that preserves anatomical structure and cross‐part interactions. A speech‐conditioned diffusion process is then applied in this latent space to model long‐range temporal dynamics and generate motion sequences aligned with audio input. By performing diffusion in a structured latent space, HoloDiff reduces modeling complexity while improving temporal coherence and audio–motion alignment. Extensive experiments on the SHOW dataset demonstrate that HoloDiff produces more realistic and expressive talking human animation and achieves better multimodal correspondence between speech and motion compared to baselines.
- Open Access
- 9 May 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Open Journal of Signal Processing
- Vol. 6, 546-554
Hierarchical GraphCut Phase Unwrapping Based on Invariance of Diffeomorphisms Framework
Recent years have witnessed rapid advancements in 3D scanning technologies, with diverse applications spanning VR/AR, digital human creation, and medical imaging. Structured-light scanning with phase-shifting techniques is preferred for its use of non-radiative, low-intensity visible light and high accuracy, making it well suited for human-centric applications such as capturing 4D facial dynamics. A key step in these systems is phase unwrapping, which recovers continuous phase values from measurements that are inherently wrapped modulo $2\pi$. The goal is to estimate the unwrapped phase count $k$, an integer-valued variable in the equation $\Phi = \phi + 2\pi k$, where $\phi$ is the wrapped phase and $\Phi$ is the true phase. However, the presence of noise, occlusions, and piecewise continuous phase functions induced by complex 3D surface geometry makes the inverse reconstruction of the true phase extremely challenging. This is because phase unwrapping is an inherently ill-posed problem: measurements only provide modulo $2\pi$ values, and recovering the correct unwrapped phase count requires strong assumptions about the smoothness or continuity of the underlying 3D surface. Existing methods typically involve a trade-off between speed and accuracy: Fast approaches lack precision, while accurate algorithms are too slow for real-time use. To overcome these limitations, this work proposes a novel phase unwrapping framework that reformulates GraphCut-based unwrapping as a pixel-labeling problem. This framework helps significantly improve the estimation of the unwrapped phase count $k$ through the invariance property of diffeomorphisms applied in image space via conformal and optimal transport (OT) maps. An odd number of diffeomorphisms are precomputed from the input phase data, and a hierarchical GraphCut algorithm is applied in each corresponding domain. The resulting label maps are fused via majority voting to efficiently and robustly estimate the unwrapped phase count $k$ at each pixel, using an odd number of votes to break ties. Experimental results demonstrate a 45.5× speedup and lower $L^{2}$ error in both real experiments and simulations, showing potential for real-time applications.
- 9 May 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Open Journal of Signal Processing
- Vol. 6, 546-554
Open AccessHierarchical GraphCut Phase Unwrapping Based on Invariance of Diffeomorphisms Framework
Recent years have witnessed rapid advancements in 3D scanning technologies, with diverse applications spanning VR/AR, digital human creation, and medical imaging. Structured-light scanning with phase-shifting techniques is preferred for its use of non-radiative, low-intensity visible light and high accuracy, making it well suited for human-centric applications such as capturing 4D facial dynamics. A key step in these systems is phase unwrapping, which recovers continuous phase values from measurements that are inherently wrapped modulo $2\pi$. The goal is to estimate the unwrapped phase count $k$, an integer-valued variable in the equation $\Phi = \phi + 2\pi k$, where $\phi$ is the wrapped phase and $\Phi$ is the true phase. However, the presence of noise, occlusions, and piecewise continuous phase functions induced by complex 3D surface geometry makes the inverse reconstruction of the true phase extremely challenging. This is because phase unwrapping is an inherently ill-posed problem: measurements only provide modulo $2\pi$ values, and recovering the correct unwrapped phase count requires strong assumptions about the smoothness or continuity of the underlying 3D surface. Existing methods typically involve a trade-off between speed and accuracy: Fast approaches lack precision, while accurate algorithms are too slow for real-time use. To overcome these limitations, this work proposes a novel phase unwrapping framework that reformulates GraphCut-based unwrapping as a pixel-labeling problem. This framework helps significantly improve the estimation of the unwrapped phase count $k$ through the invariance property of diffeomorphisms applied in image space via conformal and optimal transport (OT) maps. An odd number of diffeomorphisms are precomputed from the input phase data, and a hierarchical GraphCut algorithm is applied in each corresponding domain. The resulting label maps are fused via majority voting to efficiently and robustly estimate the unwrapped phase count $k$ at each pixel, using an odd number of votes to break ties. Experimental results demonstrate a 45.5× speedup and lower $L^{2}$ error in both real experiments and simulations, showing potential for real-time applications.