In the fields of film, architecture, and design, global illumination is actively used due to the demand of photorealistic computer graphics images. However, monte carlo ray tracing, which is commonly used algorithm for it, is known to be computationally inefficient because it take a large number of samples until convergence of its output image. Therefore, increasing the computational efficiency of ray tracing has been a major research topic for many years.
Path guiding is one of its optimization method that reduces the number of samples by constructing an efficient light path. It has been shown that path guiding can significantly reduce dispersion compared to existing methods, especially in scenes that include caustics.
However, in scenes with many light sources, it is difficult to obtain valid contributions from each light source, and the number of samples required for estimating the radiance distribution tends to be large. Therefore, we propose optimization method of path guiding on the multi-source illumination problem, which improves the dispersion reduction by using appropriate sampling method for light source selection.
|Shumpei Sugita||Keio University|
[J-1] Shumpei Sugita, Issei Fujishiro: “GPU-based Acceleration of Path Guiding for Many-Light Global Illumination,” The Journal of the Institute of Image Electronics Engineers of Japan, Vol 52, No. 1, pp. 193-204, January 2023 (in Japanese).
[DP-1] Shumpei Sugita, Issei Fujishiro: “Efficient many-light global illumination using path guiding,” in Proceedings of the 84th National Convention of International Processing Society of Japan, Vol. 4, pp. 249―250 (6ZF-02), Online, March 3―5, 2022 (in Japanese).
[DP-2] Shumpei Sugita, Issei Fujishiro: “Efficient many-light global illumination using path guiding,” in Proceedings of SIG Technical Reports, Vol. 2022–CG–186, No. 5, pp. 1-6, June 27, 2022 (in Japanese).
[DP-3] Shumpei Sugita, Issei Fujishiro: “Improving Path Guiding for Many-Light Global Illumination and
its GPU-based Acceleration,” in Proceedings of the Media Computing Conference, pp. 1-4, August 31― September 2, 2022 (in Japanese).
- Grant-in-Aid for Scientific Research (A): 21H04916 (2021－)