Research And Implementation Of Point Cloud Compression Based On Octree Encoding

Point Cloud Compression has recently received a lot of attention in the field of Computer Vision and Robotics,as well as in the field of Virtual Reality(VR)and Augmented Reality(AR).It's emerging as one of the quickly improving technology,but it is still facing some hurdles which need to be overcome so it can be used efficiently in practical scenarios.There have been a lot of techniques involved for the compression of point clouds.The size of point cloud data is still the main issue for it to be well implemented in these technologies.In this thesis,based on existing different techniques for Point Cloud Compression,I mainly give a comparison between two Point Cloud Compression techniques,i.e.,Double Buffer based Octree and the Single Octree-based Point Cloud Coding,and conduct detailed analyses on them and their techniques.In addition,I propose a modified Octree version for Point Cloud Compression called Single Octree Compression for Static Point Clouds,which has been inspired by Double Buffer Octree Compression.The algorithm of the proposed Single Octree Compression for Static Point Clouds uses similar compression model with that of Double Buffer Octree Compression.It works in such a way that it follows full lifecycle of Compression and Decompression of the Point Clouds over a network.It initializes from spatially decomposing a point cloud using octree data structures,then operates bit masking,and then serializes this structure using a binary stream to represent the octree structure for encoding and decoding points information.Furthermore,it uses point position-coding to encode additional information of points,e.g.color,NORMALS,etc.The experimental results of my version of octree show compression improvements on static point clouds compared to Double Buffer based Octree.Although there is little compromise on visual performance of the proposed approach but the compression ratio is higher than double buffer.