Clustering and DCT Based Color Point Cloud Compression

In this paper, a new point cloud compression method is proposed. The 3D color point cloud is firstly mean-shift clustered into many homogeneous blocks based on the similar spatial (XYZ) information of each point. Based on the RANdom SAmple Consensus (RANSAC) algorithm, those points being clustered in the same block are fitted by a 3D plane and all these points belonging to the same block are projected to this corresponding plane. For every plane an optimal rectangle bounding box is identified and is divided into n × n grids, the color (RGB) information associated with each grid point is replaced by the average of RGB values of all the projected points falling in this grid. Finally, a 2D DCT (Discrete Cosine Transform) transform is performed on these n × n grids points. The compressing ratio can reach 32 with negligible spatial and color distortion.