A motion-compensated inter-frame attribute coding scheme for dynamic dense point clouds
2023 IEEE 25th International Workshop on Multimedia Signal Processing (MMSP)(2023)
摘要
This paper addresses the problem of compressing the colour attributes of dynamic dense point clouds. Inter-frame prediction and motion compensation are key to removing temporal redundancies and thereby reaching major compression gains. For that purpose, an attribute compression scheme combining intra-frame and motion-compensated inter-frame predictions is presented. It is built upon the Test Model developed by MPEG within the Geometry-based Point Cloud Compression (G-PCC) activity to encode dense dynamic point clouds. More precisely, a motion-compensated inter-frame prediction is introduced into the RAHT encoding scheme, leveraging the local motion field already present in the geometry encoder. The best prediction mode according to rate-distortion optimization is decided at each node of the octree, encoded by means of arithmetic coding using a binary prediction tree and signalled to the decoder. Experimental results are provided using the MPEG test sequences and common test conditions. They demonstrate very significant compression gains, with average BD-Rates of -15.2%, -18.3% and -18.1% on Y, Cb and Cr colour components respectively, when compared with the current scheme with intra-frame only attribute compression.
更多查看译文
关键词
Dynamic point cloud,dense point cloud,point cloud compression,attribute coding,RAHT transform,inter-frame prediction,motion compensation
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络
Chat Paper
正在生成论文摘要