Best-effort projection based attribute compression for 3D point cloud

Abstract

According to the characteristics of exquisite presentation and rapid capture, 3D point cloud has been widely applied in immersive media industry. Aiming at vivid rendering of objects or dynamic scenarios, each point is associated with corresponding attributes like color, normal or intensity which induce massive data capacity. Consequently, efficient attribute compression methods are essential for effective immersive media transmission and consumption in the current media system. In this paper, we propose a best-effort projection based compression method for point cloud attributes. To take advantage of the well-developed 2D compression algorithm, regularized 3D point cloud is projected onto specified planes as different views while position information and related attributes are preserved. Joint depth- and color-dependent block-wise prediction has been utilized to further reduce the inter-view redundancy between projected 2D images. The experimental results have shown that the point cloud is successfully reconstructed via corresponding decoding process. Our method has presented strong competitiveness in both lossless and lossy compression of attributes of 3D point cloud.