BitC-3DGS: High-Capacity 3D Gaussian Splatting Watermarking via Bit Compression

2026-05-28

AI Summary

This work addresses the limited information capacity of existing 3D Gaussian splatting watermarking methods, which are constrained by the 77-token context window of CLIP encoders and thus struggle to embed high-capacity data such as ownership identifiers, provenance metadata, or authentication codes. To overcome this bottleneck, the authors propose a bit-compressed watermarking framework that encodes multiple message bits into a single semantic token and employs a dual-branch neural network to jointly perform block decompression and bit decoding. The approach introduces a novel tokenization mechanism based on bit compression and a hard message sampling strategy, enabling successful embedding of 128-bit watermarks on the Blender and LLFF datasets. The method achieves recovery accuracy comparable to state-of-the-art techniques operating at only 64 bits, while preserving high rendering fidelity and substantially surpassing conventional capacity limits.

Abstract

High-capacity watermarking is necessary for 3D Gaussian Splatting (3DGS) assets to embed rich information (e.g., ownership, provenance, and authentication codes), enabling reliable identification and integrity verification in large-scale 3D asset pipelines. Existing bit-to-token watermarking methods based on a pre-trained text encoder are limited to 77-bit messages due to CLIP's fixed 77-token context length, as tokens beyond this limit are unsupported by learned positional embeddings. To address this limitation, we introduce BitC-3DGS, a bit-compression framework that encodes multiple message bits per token. It employs a bit-compressed tokenization scheme that encodes multiple bits within the same chunk into a single semantic token. To enable recovery of the compressed information, it further introduces a dual-branch architecture for joint chunk decompression and bit decoding, along with a hard-message sampling strategy to improve combinatorial coverage during decoder training. Extensive experiments on the Blender and LLFF datasets demonstrate the effectiveness of BitC-3DGS for high-capacity watermarking, achieving high message recovery accuracy and rendering fidelity. For example, it supports 128-bit message capacity with recovery accuracy comparable to that of 64-bit messages in recent state-of-the-art methods.

Problem

• 3D Gaussian Splatting
• high-capacity watermarking
• bit compression
• message embedding
• CLIP token limit

Innovation

• bit compression
• 3D Gaussian Splatting
• high-capacity watermarking
• dual-branch decoding
• tokenization