Robust3DGSW: Toward Robust Watermarking for Quantization-Aware 3D Gaussian Splatting

Although current watermarking techniques for 3D Gaussian Splatting (3DGS) are promising in protecting the copyrights of both 3DGS models and their rendered images, they suffer from low watermark robustness and poor rendering quality when quantizing large 3DGS models to accommodate resource-limited devices. To address these problems, this paper introduces a novel twostage quantization-aware 3DGS watermarking approach called Robust3DGSW. By properly embedding watermarks in the mid-frequency bands of both the 3D Gaussian parameters and the 2D rendered images, the first stage of Robust3DGSW can effectively counteract quantizationinduced signal loss and mitigate the adverse effects of watermarks on rendered images. In the second stage, Ro-bust3DGSW trains both 2D and 3D decoders using our proposed multi-scale adversarial perturbation approach, alongside a gradual quantization process, which enables robust watermark extraction even under excessive quantization. Comprehensive experimental results on the wellknown Blender, LLFF, and MipNeRF-360 datasets demonstrate that, compared to state-of-the-art 3DGS watermarking techniques, Robust3DGSW not only mitigates the negative effects of quantization on watermarks but also enables fast, high-quality rendering.