Incident Radiance Reuse Filter And Improvement Of Radiance Estimation

Global illumination algorithms can synthesize realistic images, and are increasingly used in areas such as architecture, design, entertainment and so on. Scenes in these areas are becoming increasingly complex, and Monte Carlo ray tracing based illumination techniques are the best choice to handle this complexity. However, the main problem of these techniques is the variance in integral estimation seen as noise in the rendered images. Noise can be reduced by increasing samples, but it will take quite a long time. Noise can also be reduced by filtering, so it is necessary to develop filter algorithms in order to render pleasant images fast.Incident radiance reuse filter was presented based on previous work, and classic radiance estimation algorithm was improved. Incident radiance reuse filter was developed based on Monte Carlo path tracing. Incident radiance samples were saved in a data structure, and then reused near rendering points to calculate illumination according to several reuse standards. Distance information around the intersect point was considered in the reuse standards, so large reuse window could be used to reduce noise without introducing much bias to illumination of details.Incident radiance reuse filter could handle only parts of light transport paths efficiently, so caustic was rendered with photon map algorithm, but classic radiance estimation was biased especially in details. In this paper classic radiance estimation algorithm was improved and light flux calculation was revised, resulting in better representation of details. The algorithm presented in this paper could render pleasant images of scenes with many kinds of materials in a short time, resulting in wide applicability in many areas.