3D point cloud compression system based on multi-scale structured dictionary learning

Abstract

The invention provides a 3D point cloud compression system based on multi-scale structured dictionary learning. A point cloud data division module outputs a voxel set after point cloud division and voxel block sets of different scales; a geometric information encoding module outputs an encoded geometric information bit stream; a geometrical information decoding module outputs decoded geometrical information; an attribute signal coding module outputs a coefficient matrix of sparse coding and a learned multi-scale structured dictionary; wherein the attribute signal encoding module outputs a learned multi-scale structured dictionary, the attribute signal compression module outputs a compressed attribute signal bit stream, the attribute signal decoding module outputs a decoded attribute signal, and the 3D point cloud reconstruction module completes reconstruction. The system is suitable for point cloud signal lossless geometry and lossy attribute compression, the reconstruction quality ofhigh-frequency detail information is gradually improved in the direction from coarse to fine along the signal scale by utilizing the natural hierarchical division structure of the point cloud signal,and a remarkable performance gain can be obtained.

Description

technical field [0001] The invention relates to a solution in the technical field of 3D point cloud data compression, in particular to a 3D point cloud compression system based on multi-scale structured dictionary learning. Background technique [0002] In recent years, with the rapid development of 3D data acquisition equipment and display systems, and with the help of powerful GPU computing capabilities, real-world scenes and objects can be digitized in real time into high-detail and high-precision 3D point data. 3D data and technology have been widely used in many emerging fields, including virtual / augmented reality, mixed reality, telepresence, panoramic communication, automatic driving and robot navigation. There are various 3D modeling technologies. For example, RGB-D frames and multi-view video technology containing depth information can model 3D scenes and objects, but they do not support real-time rendering; polygonal meshes can use the relationship between vertices...

AI Technical Summary

Problems solved by technology

There are various 3D modeling technologies. For example, RGB-D frames and multi-view video technology containing depth information can model 3D scenes and objects, but they do not support real-time rendering; polygonal meshes can use the relationship between vertices and points Connectivity reconstructs the 3D surface of an object, but cannot model 3D data that does not satisfy the manifold topology

However, the 3D-2D projection of V-PCC will introduce inevitable projection distortion, while the attribute space transformation of G-PCC only depends on geometric information, without considering the statistical characteristics of the attribute signal itself, which reduces the compression quality of the attribute signal

Images