Interactive Rendering for Large-Scale Mesh Based on MapReduce

We present an interactive rendering solution that aims at visualizing large-scale mesh with hundred millions of triangles up to extremely high resolution on cloud computing platform. Firstly, we propose a novel adaptive parallel rasterization method based on MapReduce, whose results are stored in a data format called enhanced layered depth image (ELDI). In order to fully apply the powerful capabilities of cloud computing, our proposed method integrates pixel and triangle based strategies in one processing pipeline so as to significantly reduce data transfer between Map and Reduce steps. Consequently, a light-weight web service for interactive rendering is also proposed to couple with the rasterization step, which enables user to interactively visualize large-scale mesh via browser on both mobile devices and work stations. According to demonstrated examples in the paper, once the rasterization results are obtained, users can freely adjust visualizing effects of complex mesh. Our rendering approach is promising for big data exploration, analysis or authoring.