SRM-Hair: Single Image Head Mesh Reconstruction via 3D Morphable Hair

TL;DR

SRM-Hair tackles single-image head mesh reconstruction by introducing semantic-consistent ray modeling to produce an ordered, morphable hair representation. A high-fidelity hair dataset paired with 3D faces enables PCA-based hair priors, while semantically aligned scalp rays yield consistent hair vertex statistics, enabling additivity, adaptability, and thickness control. The framework predicts morphable hair coefficients from in-the-wild images, reconstructs hair geometry from scalp ray distances, and uses a refinement network to remove outliers, achieving state-of-the-art accuracy for both hair region and full-head geometry. This approach enables efficient, independent hair mesh reconstruction suitable for realistic avatars and hair rendering, and adds a valuable real-data hair mesh resource for the community.

Abstract

3D Morphable Models (3DMMs) have played a pivotal role as a fundamental representation or initialization for 3D avatar animation and reconstruction. However, extending 3DMMs to hair remains challenging due to the difficulty of enforcing vertex-level consistent semantic meaning across hair shapes. This paper introduces a novel method, Semantic-consistent Ray Modeling of Hair (SRM-Hair), for making 3D hair morphable and controlled by coefficients. The key contribution lies in semantic-consistent ray modeling, which extracts ordered hair surface vertices and exhibits notable properties such as additivity for hairstyle fusion, adaptability, flipping, and thickness modification. We collect a dataset of over 250 high-fidelity real hair scans paired with 3D face data to serve as a prior for the 3D morphable hair. Based on this, SRM-Hair can reconstruct a hair mesh combined with a 3D head from a single image. Note that SRM-Hair produces an independent hair mesh, facilitating applications in virtual avatar creation, realistic animation, and high-fidelity hair rendering. Both quantitative and qualitative experiments demonstrate that SRM-Hair achieves state-of-the-art performance in 3D mesh reconstruction. Our project is available at https://github.com/wang-zidu/SRM-Hair

Figures (8)

• Figure 1: We propose SRM-Hair to advance head mesh reconstruction with a dataset of over 250 high-fidelity real hair mesh scans paired with 3D face data (a), along with a novel method for reconstructing a 3D head mesh with independent hair region from a single image (c) using semantic-consistent ray modeling (b) for 3D morphable hair.
• Figure 2: Overview of our semantic-consistent ray modeling. The core idea is to emit sufficient semantically consistent rays with specific origins and directions to intersect with the hair surface, enabling statistical analysis of the hair vertices.
• Figure 3: The properties of semantic-consistent ray modeling.
• Figure 4: Overview of SRM-Hair for reconstructing a 3D hair mesh from a single image.
• Figure 5: Qualitative comparison with the other methods. Our method (SRM-Hair) achieves the best results, accurately reflecting the full head region. Furthermore, our reconstructed hair (shown in dark gray) and face are separated, enhancing the potential for application.