Subjective Quality Assessment for Point Clouds of Digital Humans with Shaded Rendering

The emergence of use cases such as immersive telepresence and virtual entertainment, where real life dynamic scenes of humans are captured and reconstructed in 3D, has led to the development of new methods for point cloud compression and quality assessment. However, these methods only consider the coordinates and color attributes of the points. Other properties, such as the normals that describe the surface of the represented object and could improve decoded point cloud rendering, are ignored. In this paper, we add the coding of the normals as attributes in video-based point cloud compression (V-PCC). Through a subjective quality assessment study, we consider the impact of the induced coding distortions on shaded rendering and on the perceived quality of decoded point clouds of digital humans. We demonstrate the relevance of encoding these attributes, and the effect of this new type of distortion on the standard test methodology.