應用四元樹切割光子群於動態場景之全域照度估計

Global Illumination in Dynamic Scenes Using Quadtree to Cluster Photons

林欣靜(Shin-Ching Lin)

指導教授 ： 林聰武 ； 鍾斌賢

中原大學/工學院/資訊工程研究所/碩士(2004年)

https://doi.org/10.6840/cycu200400094

摘要

對於動態場景照度之更新，我們提出一個結合光子圖(photon map)與照度估計(illumination estimation)方法以及四元樹光子分群策略的新演算法。在光子分群方法上，我們改良Kirill的規則性光子分群，進而提出非規則性光子分群策略，所使用的方式為利用四元樹切割光源，根據動態物體的位置把全部光子適應性地分群，每一群光子群皆包含固定比例的導航光子以及矯正光子。這個方法利用框架與框架之間彼此的差異性不大的特性，在照度更新方面只需針對場景中變動部分進行重新估計即可，而這些變動的部分場景我們利用導航光子(pilot photons)偵測，判斷哪些導航光子必須被更新時，可以立即地打出同群的矯正光子(corrective photons)做更精確的光子圖資訊更新。
實驗結果證實，非規則性光子分群雖然無法很準確地更新的確需要更新的光子數，卻已經十分接近準確的數目。在龐大光子數場景狀況下，我們在切割光子與更新光子資訊時間，比規則性光子分群來得有效率。規則性光子分群為了能夠達到準確數目的更新光子數，使得在更新光子資訊時間上能夠達到最佳狀態，所以分群的數目必須較為細膩，但這卻又成為整體時間上的缺失，因為在如此多的光子群時，在導航光子偵測動態物體時，卻已在不知不覺中諸多時間於其間。而且，不同的場景適用的光子群數不盡相同，需要經過不斷的測試才能找到最佳的切割光子群數，而我們的非規則性光子分群，卻能適應性地找出最佳光子群數，使得時間花費上始終維持在一定水準，所以我們的非規則性光子分群法比規則性光子分群呈現出更有效率的且不失品質的成圖結果。

關鍵字

矯正光子 ； 導航光子 ； 光子圖 ； 照度估計

並列摘要

We present a method for global illumination in dynamic scene which is embedded photon map, density estimation techniques and cluster photons using quadtree. In the part of clustering photons, we improve Kirill’s method, and propose a non-uniform clustering photons method using quadtree to partition light area. All photons were partitioned into several groups according to the position of dynamic objects, and there are fixed ratio number of pilot photons and corrective photons in the group. The improved method exploits temporal coherence of illumination by tracing photons selectively to the scene regions that require illumination update. Such regions are identified by pilot photons. Based on the pilot photons which require updating, the corrective photons in the same group can be reshoot to update the information in the photon map.
Experiment results show that our method even although can’t update accurate number of photons but close perfectly the number. For the scene with huge photons, our method is efficiently than uniform clustering photons method according to measure partition time and update time. In order to get the accurate number of update photons to reach optimal update time, it needs a lot of groups for the uniform clustering photons method. But it will fail because so many groups spend much time on detecting dynamic objects. Different group number maybe used to different scenes, we will find the optimal group number through testing many times. Our non-uniform clustering photons method can cluster photons adaptive. So our method is more efficient than uniform clustering photons method, and has good quality rendering result.

並列關鍵字

pilot photon ； photon map ； illumination estimation ； corrective photon

參考文獻

[1] Henrik Wann Jensen, Realistic Image Synthesis Using Photon Mapping, A K Peters, Ltd., 2001.

[4] Henrik Wann Jensen, “Global illumination using photon maps,” Eurographics Rendering Workshop, pp.21-30, 1996.

[5] Turner Whitted, “An improved illumination model for shaded display,” Communication of the ACM, vol. 23, No. 4, pp. 343-349, June 1980.

[8] Gregory J. Ward, and Paul Heckbert, “Irradiance gradients,” In Third Eurographics Workshop on Rendering, pp. 85-98, May 1992.

[10] Harald Niederreiter, “Random number generation and quasi-Monte Carlo methods,” SIAM, chapter4, 1992.