CompSplat: Compression-aware 3D Gaussian Splatting for Real-world Video

2026-02-10

AI Summary

This work addresses geometric distortions and inter-frame inconsistencies in novel view synthesis from real-world videos, which arise due to compression artifacts, irregular camera trajectories, and unknown poses. To tackle these challenges, the authors propose a compression-aware 3D Gaussian splatting training framework that explicitly models diverse compression patterns in long, pose-free video sequences for the first time. The method jointly optimizes geometric structure and appearance consistency through a compression-aware frame weighting mechanism and an adaptive Gaussian pruning strategy. Evaluated on challenging benchmarks including Tanks and Temples, Free, and Hike, the approach significantly outperforms existing methods, achieving superior novel view synthesis quality and pose estimation accuracy—particularly under severe compression conditions.

Technology Category

• Computer Vision: Computational Photography, Image & Video Synthesis
• Machine Learning: Multi-instance/Multi-view Learning
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets

Abstract

High-quality novel view synthesis (NVS) from real-world videos is crucial for applications such as cultural heritage preservation, digital twins, and immersive media. However, real-world videos typically contain long sequences with irregular camera trajectories and unknown poses, leading to pose drift, feature misalignment, and geometric distortion during reconstruction. Moreover, lossy compression amplifies these issues by introducing inconsistencies that gradually degrade geometry and rendering quality. While recent studies have addressed either long-sequence NVS or unposed reconstruction, compression-aware approaches still focus on specific artifacts or limited scenarios, leaving diverse compression patterns in long videos insufficiently explored. In this paper, we propose CompSplat, a compression-aware training framework that explicitly models frame-wise compression characteristics to mitigate inter-frame inconsistency and accumulated geometric errors. CompSplat incorporates compression-aware frame weighting and an adaptive pruning strategy to enhance robustness and geometric consistency, particularly under heavy compression. Extensive experiments on challenging benchmarks, including Tanks and Temples, Free, and Hike, demonstrate that CompSplat achieves state-of-the-art rendering quality and pose accuracy, significantly surpassing most recent state-of-the-art NVS approaches under severe compression conditions.

Problem

• novel view synthesis
• compression artifacts
• pose drift
• geometric distortion
• real-world video

Innovation

• compression-aware
• 3D Gaussian Splatting
• novel view synthesis
• frame-wise weighting
• adaptive pruning