Inverse transform sampling by ray tracing

By using ray traversal search and importance sampling, the texture rendering process is optimized using ray tracing hardware, which solves the problem of unreliable performance of traditional inverse transform sampling and achieves improved rendering performance and reduced noise.

CN122415339APending Publication Date: 2026-07-17NVIDIA CORP

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
NVIDIA CORP
Filing Date
2022-08-12
Publication Date
2026-07-17

AI Technical Summary

Technical Problem

Existing technologies, when rendering images or videos in real time, rely on traditional inverse transform sampling methods that are unreliable in performance and cannot effectively control variance reduction or performance priority, resulting in high noise levels and making it difficult to provide high-quality content with limited performance budgets.

Method used

Ray traversal search is used to replace traditional binary search. Importance sampling is performed using ray tracing hardware. By determining the cumulative distribution function and geometric representation of the texture, rays are traced to improve sampling performance. Furthermore, the geometric representation of the intersecting cumulative probability is modified to reduce variance.

Benefits of technology

It significantly improves rendering performance, reduces image noise, and achieves up to 70% performance improvement while maintaining a similar variance reduction effect.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122415339A_ABST
    Figure CN122415339A_ABST
Patent Text Reader

Abstract

本公开涉及通过射线追踪进行逆变换采样。通过使用逆变换采样将采样指向更重要的区域(例如具有更高亮度值的区域)以减少噪声并提高收敛性,可以部分实现高质量的图像渲染。通过使用可以加速的树遍历单元重新制定为射线追踪问题,可以更有效地实现逆变换采样。可以基于纹理中像素的各个行和列的一组累积分布函数(CDF)生成几何网格,并且可以针对该网格追踪单独的射线,这些射线在具有更大的重要性(例如更高的亮度值)的点处具有更高的相交概率。可以通过分析相交位置处的CDF几何的偏导数来导出用于重要性采样的概率分布函数。
Need to check novelty before this filing date? Find Prior Art