Researchers propose new method to compress 3D light field point clouds

Researchers from Brazil have developed a new compression technique for plenoptic point clouds (PPCs) that significantly improves storage efficiency for light field data. Published on arXiv under the title *'Geometry-Based Compression of Plenoptic Point Clouds'*, the work introduces a method that applies a Karhunen-Loève transform to color attributes before multiple attribute coders with intra prediction capabilities. This approach can be seamlessly integrated into MPEG's existing geometry-based point cloud compression (G-PCC) standard.

Testing across various spatial resolutions showed the new compression scheme delivers competitive results compared to existing methods like RAHT-based or video-based PPC solutions. The authors claim their coder represents the new state of the art in PPC compression efficiency. Plenoptic point clouds represent light from multiple viewing directions, providing higher realism than traditional point clouds by associating each point with multiple colors rather than a single value.