Woven Fabric Capture with a Reflection-Transmission Photo Pair

TL;DR

This work tackles the challenge of recovering high-fidelity woven fabric parameters by moving beyond single-image reflection capture to a reflection–transmission pair captured with a simple phone setup. It introduces a novel two-layer BSDF model that combines a two-layer SpongeCake representation with an azimuthally-invariant ASGGX phase function to model both reflection and transmission, including thickness-dependent effects through a tension-aware thickness modulation. The parameter estimation pipeline combines a lightweight neural predictor with differentiable rendering, enabling accurate matches to both reflection and transmission images and producing plausible draped cloth renderings. The results on synthetic and real data demonstrate improved fidelity over prior single-image methods and highlight the practical potential for fabric design, digital humans, and interior visualization, while also outlining limitations related to yarn-level variations and unseen weave patterns.

Abstract

Digitizing woven fabrics would be valuable for many applications, from digital humans to interior design. Previous work introduces a lightweight woven fabric acquisition approach by capturing a single reflection image and estimating the fabric parameters with a differentiable geometric and shading model. The renderings of the estimated fabric parameters can closely match the photo; however, the captured reflection image is insufficient to fully characterize the fabric sample reflectance. For instance, fabrics with different thicknesses might have similar reflection images but lead to significantly different transmission. We propose to recover the woven fabric parameters from two captured images: reflection and transmission. At the core of our method is a differentiable bidirectional scattering distribution function (BSDF) model, handling reflection and transmission, including single and multiple scattering. We propose a two-layer model, where the single scattering uses an SGGX phase function as in previous work, and multiple scattering uses a new azimuthally-invariant microflake definition, which we term ASGGX. This new fabric BSDF model closely matches real woven fabrics in both reflection and transmission. We use a simple setup for capturing reflection and transmission photos with a cell phone camera and two point lights, and estimate the fabric parameters via a lightweight network, together with a differentiable optimization. We also model the out-of-focus effects explicitly with a simple solution to match the thin-lens camera better. As a result, the renderings of the estimated parameters can agree with the input images on both reflection and transmission for the first time. The code for this paper is at https://github.com/lxtyin/FabricBTDF-Recovery.

Figures (16)

• Figure 1: Given an incoming ray, we simulate the scattering among the fibers with different bounces using the SGGX phase function and visualize the distribution of the outgoing ray. The outgoing distribution becomes uniform along the azimuth angle. Representing such an azimuthally invariant distribution is beyond the capability of the SGGX phase function.
• Figure 2: Configurations of the SGGX (left) and our ASGGX (right). In our ASGGX, $\omega_o$ is rotated to the same longitudinal plane as $\omega_i$, leading to a new vector $\omega_o'$. Then, $\omega_i$ and $\omega_o'$ form a half-vector $\omega_h'$, which is used to look up the microflake density.
• Figure 3: Multiple scattering comparison among SGGX, ASGGX (ours), and Monte-Carlo simulation Guo:2018:Layered, which serves as a synthetic reference on two different sets of parameters (roughness $\alpha$ and thickness $T$). Our method can closely match the reference on both reflection and transmission, while the results by SGGX match the GT on reflection only.
• Figure 4: Macro appearance and its zoom-in. Yarns of the bottom layer are also visible, which have a significant effect on the macro appearance.
• Figure 5: The yarn becomes thinner at the intersection of two yarns because of the increased tension.