内容感知的环境光采样算法

In this paper, we present a novel approach for environment map sampling, which is an effective and pragmatictechnique to reduce the computational cost of realistic rendering. The proposed approach exploits the advantage of adaptivemean-shift image clustering, yielding content-aware strata. The resulting strata, however, have unbalanced importances for therendering, and the strata number is not user-controlled. To handle these issues, we further develop an adaptive split-and-mergescheme that refines the strata and obtains a more balanced strata distribution. Compared with the state-of-the-art methods, ourapproach achieves comparable and even better rendering accuracy. What’s more, experimental results show that our approachis more robust to the variation of the sample number.