Impact of Point Cloud Normals Compression on Objective Quality Assessment

Amar Tious, Toinon Vigier, Vincent Ricordel

Abstract: With the emergence of immersive technologies and new use cases, point cloud (PC) compression and quality assessment have become very active research fields, which led to tremendous progress and standardization challenges in the last 5 years. But these methods only propose to compress the PC's geometry and color properties, and very few studies also consider other attributes that could improve the rendering of point clouds in 3D scenes. In this paper, we focus on the surface normal properties that allow surface shading of the object, and add these attributes too in the video-based point cloud compression (V-PCC) pipeline considering two approaches: either by using attribute coding to transmit the normals, or by recomputing them from decoded PC geometry. We compare these approaches with objective metrics that take distortion of the normals and shaded rendering into account, and show the strategic advantages of normal attribute compression. Finally we discuss the challenges that point cloud compression brings up in order to optimize the physically-based rendering.