点渲染方式对量化点云主观质量的影响及算法优化

To investigate the impact of point rendering methods on the subjective quality of quantization point clouds and optimize the rendering algorithm, a subjective quality evaluation experiment is designed, considering different basic geometric types and calculation methods for rendering radius. The effects of the basic geometric types and nearest-neighbor rendering algorithm on the subjective quality of geometric and attribute quantization distortion are analyzed through the Wilcoxon signed-rank test and independent samples T-test. The experimental results show that the subjective quality is affected differently by the basic geometric types due to variations in the degree of overlap. A larger overlapping area of the basic geometric types shall result in lower subjective quality. The nearest-neighbor rendering algorithm does not have any significant impact on the subjective quality. This algorithm reduces the occurrence of holes in the rendered point cloud but increases the level of overlap. When the attribute distortion is minor, the nearest-neighbor rendering algorithm outperforms the fixed rendering radius method. However, when the attribute distortion is significant, it underperforms the fixed rendering radius method. Based on the subjective experimental results, a mathematical model is established to calculate the rendering radius of the basic geometric types, considering overlap and the distortion of holes. This model has employed geometric quantization with an octree pruning and simplified the model parameters using spatial similarity. An algorithm is proposed to calculate the rendering radius of the basic geometric types using geometric quantization parameters. Compared with the nearest-neighbor rendering algorithm, the proposed method reduces the time complexity by 52% and improves the peak signal-to-noise ratio of the rendered point cloud by 12. 3%. Additionally, it enhances the subjective quality by 0. 5 points. The research, featuring the reduction of computational resources and the improvement of rendering efficiency, can provide references for the design and optimization of Tenderers.