Comprehensive coding metric

Unicorn is the very first, versatile, learning-based solution to support the compression of geometry and attribute information of any given point cloud (e.g., static or dynamic, dense or sparse, object or scene) in both lossless and lossy modes.

Better performance

Unicorn provides significant performance gains to existing standards. As for geometry coding, it outperforms GPCC and V-PCC with about 90% and 75% bitrate reduction at the same geometry distortion for solid object samples and >35% BD-Rate gains for dynamically acquired LiDAR sequences. Compared with existing learning-based solutions, Unicorn also improves the compression efficiency noticeably.

Low computation complexity

Unicorn is a low-complexity approach. Its prototyping implementation exhibits comparable runtime measures to G-PCC’s geometry codec.