GaussianHead: High-Fidelity Head Avatars With Learnable Gaussian Derivation

• Published in: IEEE transactions on visualization and computer graphics Vol. 31; no. 7; pp. 4141 - 4154
• Main Authors: Wang, Jie, Xie, Jiu-Cheng, Li, Xianyan, Xu, Feng, Pun, Chi-Man, Gao, Hao
• Format: Journal Article
• Language: English
• Published: United States IEEE 01.07.2025
• Subjects: 3D Gaussian splatting; 3D reconstruction; Anisotropic; Avatars; Covariance matrices; Head; Head avatar; hybrid neural network; Image reconstruction; Magnetic heads; Rendering (computer graphics); Solid modeling; Three-dimensional displays; Training
• ISSN: 1077-2626; 1941-0506; 1941-0506
• DOI: 10.1109/TVCG.2025.3561794

Creating lifelike 3D head avatars and generating compelling animations for diverse subjects remain challenging in computer vision. This paper presents GaussianHead, which models the active head based on anisotropic 3D Gaussians. Our method integrates a motion deformation field and a single-resolution tri-plane to capture the head's intricate dynamics and detailed texture. Notably, we introduce a customized derivation scheme for each 3D Gaussian, facilitating the generation of multiple "doppelgangers" through learnable parameters for precise position transformation. This approach enables efficient representation of diverse Gaussian attributes and ensures their precision. Additionally, we propose an inherited derivation strategy for newly added Gaussians to expedite training. Extensive experiments demonstrate GaussianHead's efficacy, achieving high-fidelity visual results with a remarkably compact model size (<inline-formula><tex-math notation="LaTeX">\approx 12</tex-math> <mml:math><mml:mrow><mml:mo>≈</mml:mo><mml:mn>12</mml:mn></mml:mrow></mml:math><inline-graphic xlink:href="xie-ieq1-3561794.gif"/> </inline-formula> MB). Our method outperforms state-of-the-art alternatives in tasks such as reconstruction, cross-identity reenactment, and novel view synthesis.