发表机构
United Imaging Intelligence（联影智能）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
FactorSplat提出外观可控的高斯代理，支持推理时传递函数编辑，在CT/MR体绘制中PSNR提升1.10-1.52 dB，无需重训。
AI 中文摘要
传递函数（TFs）控制医学体绘制中的颜色和可见性，但图像训练的高斯代理通常将一种传递函数烘焙到其外观中。我们提出FactorSplat，一种每场景N维高斯泼溅（N-DGS）代理，在推理时接受特定区域的强度到RGBA曲线。局部查找应用作者定义的颜色和不透明度变化，而共享功能编码器和低秩每高斯因子学习残差外观响应。几何和方向外观在预设间保持共享，可见性控制和TF感知剪枝保留隐藏和揭示结构的能力。在七次CT和MR扫描中，FactorSplat在验证、插值、未见组合和分布外（OOD）编辑上，相比区域感知VEG提高了平均PSNR和变化区域误差。在这四个分割中，七扫描平均PSNR增益相对于VEG范围为1.10至1.52 dB。每个扫描一个检查点支持未见编辑而无需重新训练。在$1600^2$分辨率下，缓存的快速渲染器平均524 FPS，TF切换时间为1.17毫秒。项目页面：此https URL。
英文摘要
Transfer functions (TFs) control color and visibility in medical volume rendering, but image-trained Gaussian proxies typically bake one transfer function into their appearance. We present FactorSplat, a per-scene N-dimensional Gaussian splatting (N-DGS) proxy that accepts region-specific intensity-to-RGBA curves at inference. A local lookup applies the authored color and opacity change, while a shared functional encoder and low-rank per-Gaussian factors learn the residual appearance response. Geometry and directional appearance remain shared across presets, with visibility control and TF-aware pruning preserving the ability to hide and reveal structures. On seven CT and MR scans, FactorSplat improves mean PSNR and changed-region error over region-aware VEG across validation, interpolation, unseen composition, and out-of-distribution (OOD) edits. Across these four splits, seven-scan mean PSNR gains over VEG range from 1.10 to 1.52 dB. One checkpoint per scan supports unseen edits without retraining. At $1600^2$, the cached fast renderer averages 524 FPS with 1.17 ms TF switches. Project page: https://gaozhongpai.github.io/FactorSplat/.
