DynCam

We contribute a new library, DynCam, for real-time, low latency,streaming point cloud processing with a special focus on telep-resence in VR. Our library combines several RGBD-images frommultiple distributed sources to a single point cloud and transfersit through a network. This processing is organized as a pipelinethat supports implicit multithreading. The pipeline uses functionalreactive programming to describe transformations on the data in adeclarative way. In contrast to previous libraries, DynCam is plat-form independent, modular and lightweight. This makes it easyto extend and allows easy integration into existing applications.We have prototypically implemented a telepresence application inthe Unreal Engine. Our results show that DynCam outperformscompeting libraries concerning latency as well as network traffic.

The real-time rendering method for splatted point clouds proposed in the paper, was implemented in his master thesis by Valentin Kraft and is available here: UUE4_GPUPointCloudRenderer. Source code for the latency measurements presented in our paper are available upon request.