Multiresolution Point Cloud Compression for Real-Time Visualization and Streaming of Large 3D Datasets

In recent years, the integration of point cloud data has been observed across diverse domains such as immersive media, autonomous driving, and healthcare. However, processing and visualizing 3D point cloud data requires more computational power and processing time than typical 2D images or videos. In this paper, we present a multiresolution point cloud visualization approach that enables the encoding and reconstruction of human body point clouds at multiple quality/bitrate levels for streaming or real-time visualization applications. Specifically, we partition the 3D model into cells and build an octree structure to store the point cloud at multiple levels of detail. The hierarchical structure is encoded efficiently for fast loading times. We optimize the resolution selection for each visible node based on the camera viewpoint and the geometric error perceived by the user. We validate the proposed method using a widely known point cloud dataset from the Moving Picture Experts Group. The experiments demonstrate that the method can dynamically adapt the resolution of point clouds and provide faster rendering times compared to well-known point cloud visualization software.