🤖 AI Summary
Existing 2D Gaussian Splatting (2DGS) suffers from blurred edges and boundary distortions under low Gaussian counts (<1k) due to insufficient contour awareness. This work proposes semantic-guided contour-aware 2DGS: it is the first to incorporate segmentation priors into the 2DGS framework, constraining Gaussian distributions within semantic regions and suppressing cross-boundary mixing. We design region-constrained rasterization and a warm-up progressive training strategy to enhance edge convergence stability. Evaluated on benchmarks including DAVIS, our method significantly improves edge PSNR and SSIM, preserves sharp contours even with extremely sparse Gaussians, achieves real-time rendering (>60 FPS), and maintains memory overhead comparable to vanilla 2DGS. Key contributions include: (i) contour-aware geometric modeling, (ii) semantic-driven Gaussian allocation, and (iii) an efficient boundary optimization paradigm.
📝 Abstract
Image representation is a fundamental task in computer vision. Recently, Gaussian Splatting has emerged as an efficient representation framework, and its extension to 2D image representation enables lightweight, yet expressive modeling of visual content. While recent 2D Gaussian Splatting (2DGS) approaches provide compact storage and real-time decoding, they often produce blurry or indistinct boundaries when the number of Gaussians is small due to the lack of contour awareness. In this work, we propose a Contour Information-Aware 2D Gaussian Splatting framework that incorporates object segmentation priors into Gaussian-based image representation. By constraining each Gaussian to a specific segmentation region during rasterization, our method prevents cross-boundary blending and preserves edge structures under high compression. We also introduce a warm-up scheme to stabilize training and improve convergence. Experiments on synthetic color charts and the DAVIS dataset demonstrate that our approach achieves higher reconstruction quality around object edges compared to existing 2DGS methods. The improvement is particularly evident in scenarios with very few Gaussians, while our method still maintains fast rendering and low memory usage.