Compression of Plenoptic Point Clouds

3D Point Cloud Super-Resolution via Graph Total Variation on Surface Normals

“…Towards this goal, a k-neighborhood graph G is initially constructed for all 3D points V in a point cloud. We next divide 3D points V in a point cloud into two disjoint sets-red nodes V1 and blue nodes V2 where V1 ∪ V2 = V and V1 ∩ V2 = ∅via bipartite graph approximation [19]; i.e., finding a bipartite graph that best approximates the original non-bipartite graph in terms of Kullback-Leibler (KL) divergence 4 . An example for the bipartite graph approximation is shown in Fig.…”

Colour Compression of Plenoptic Point Clouds Using Raht-Klt with Prior Colour Clustering and Specular/Diffuse Component Separation

“…This recent popularity of both, 3D point clouds and light fields, can be attributed to the same enduring goal: to achieve digital representations of the world around us, which are ever more realistic and more versatile, but still remaining practical to use, store, and distribute. Realising the potential of both 3D point clouds and light fields towards achieving this goal, recent publications [7,8,9,10] are now considering a new direction: a point cloud representation of a light field. The plenoptic point cloud [8,9] is a natural extension of a 3D point cloud to a Surface Light Field (SLF) [11].…”

“…Realising the potential of both 3D point clouds and light fields towards achieving this goal, recent publications [7,8,9,10] are now considering a new direction: a point cloud representation of a light field. The plenoptic point cloud [8,9] is a natural extension of a 3D point cloud to a Surface Light Field (SLF) [11]. Such a representation is very promising for applications that require both, the versatility and convenience of a point cloud, and the richness of visual information provided by a light field.…”

“…As shown in [8,9,10], these representations are amenable to efficient compression, partly because existing image and point cloud coding techniques can be extended to them fairly easily. In [8,9], four extensions of the Region-Adaptive Hierarchical Transform (RAHT) method [12,13] (which is part of the emerging MPEG-PCC standard for attribute coding [1]) are introduced to compress the plenoptic colour vectors. In [10], light field images are mapped to a point cloud to obtain a SLF, then the points' "view maps" are encoded using a B-Spline wavelet basis and compressed spatially using existing point cloud codecs.…”

“…Our focus in the current paper is the compression of the multiview colour vectors in plenoptic point clouds [8,9], assuming a lossless geometry. We wish to improve upon the rate-distortion (R-D) performance shown for the best-performing RAHT extension in [8,9], RAHT-KLT.…”