Perceptually Optimized Bit Allocation Between Attribute and Geometry for Video-Based Point Cloud Compression

Video-Based Point Cloud Compression (V-PCC) decomposes a dynamic point cloud into attribute, geometry and occupancy videos and then compresses them with video encoders separately. However, the visual importance of attribute and geometry videos are not equal and vary with the point clouds. In this paper, we propose a perceptually optimized bit allocation between attribute and geometry for V-PCC to exploit visual redundancies. Firstly, we propose a Point Cloud Quality Assessment based on Adaptive Visual Importance of Geometry and Attribute (PCQA-AVIGA) to accurately measure the perceptual quality of distorted point clouds, where the visual importance of geometry and attribute distortions is adaptively assigned. Secondly, we derive 2D and 3D perceptual distortion models and rate models to accurately model the relationships between PCQA-AVIGA, bit rate and the quantization steps of attribute and geometry videos. Finally, based on the PCQA-AVIGA, the predicted visual importance and the rate-distortion models, we propose a Perceptual Bit Allocation Optimization (PBAO) for V-PCC, where bits between the geometry and attribute videos are properly assigned to maximize the overall perceptual quality of point clouds with bit rate constraint. In addition, by considering the Intra- and Inter-frame characteristics in V-PCC, two PBAO models are developed to improve the coding performance. Experimental results show that the proposed PBAO reduces an average of 6.44% BDBR as compared with the latest V-PCC. Moreover, the perceptual quality is further improved.