3DGS-HPC: Distractor-free 3D Gaussian Splatting with Hybrid Patch-wise Classification

3D Gaussian Splatting (3DGS) has demonstrated remarkable performance in novel view synthesis and 3D scene reconstruction, but its quality often degrades in real-world environments due to transient distractors, such as moving objects and varying shadows. Existing methods commonly introduce semantic priors from pre-trained vision models either to group pixels into coherent regions or to define perceptual error metrics. However, semantic grouping is often misaligned with the binary static/transient distinction, while perceptual features can be fragile under appearance perturbations introduced during 3DGS optimization. We propose 3DGS-HPC, a framework that addresses these issues by combining two complementary principles: a patch-wise classification strategy that leverages local spatial consistency for robust region-level decisions, and a hybrid classification metric that adaptively integrates photometric and perceptual cues for more reliable separation. Extensive experiments demonstrate the superiority and robustness of our method in mitigating distractors to improve 3DGS-based novel view synthesis. Our project page is https://cnhaox.github.io/3DGS-HPC/.

Many tools can build a 3D scene from photos, but they often assume every photo shows the same unchanged world. In real captures, people walk by, cars move, and shadows shift, so temporary elements can get baked into the final 3D model as blur, holes, or ghostly artifacts. We study this problem for 3D Gaussian Splatting, a fast technique for creating new views of a scene from photos. Our method, 3DGS-HPC, teaches the system to ignore temporary distractions while keeping scene parts that are truly permanent. Instead of asking a general-purpose vision model to decide what each object is, we look at small image patches and ask a simpler question: does this region behave like stable scene content or a one-time distraction? We answer using two complementary clues: how different the region looks in color and how differently a vision model interprets it. Combining these clues makes the decision more reliable than using either alone. Across real-world benchmarks, our method produces cleaner 3D reconstructions and preserves more static detail than previous approaches. This can make 3D capture more useful for AR/VR content creation, robotics, and mapping in everyday dynamic environments.