360-Degree Point Cloud Compression with Adaptive Rate Control Optimisation for Regions of Interest

The growing use of point cloud technology across various sectors has emerged as a significant research area, focusing on compressing point clouds for improved storage and transmission. Geometry-based point cloud compression (G-PCC), one of the latest standards for point cloud compression, has been considered to compress any static point clouds. G-PCC is primarily designed for general purposes, employing a uniform quantisation strategy that does not account for object positions. This is because most point clouds are captured from various angles, focusing on a central location with a limited number of objects. However, applications such as driverless cars use 360-degree cameras to capture more objects at different distances, making it essential to preserve the quality of objects in different regions, or Regions of Interest (ROIs). In this paper, we propose a non-uniform quantisation strategy within the G-PCC framework to maintain high-quality ROIs without compromising overall rate-distortion performance. Our experimental results show that the proposed method improves PSNR-D1 and PSNR-D2 of ROI by more than 2.5 dB while maintaining the overall rate-distortion performance of G-PCC.