Dynamic data reshaping for 3D mesh animation compression

Guoliang Luo, Xin Zhao, Qiang Chen, Zhiliang Zhu, Chuhua Xian

Abstract: Effective compression of 3D mesh animation data has been increasingly used in a variety of multimedia systems including virtual reality, gaming, remote transmission, display and storage. In this work, we propose a spectral clustering-based dynamic reshaping model that is performed on spatio-temporal segments to enhance the compression of 3D mesh sequences. After the lossy compression of spatio-temporal segments through Principal Component Analysis (PCA), we first compute a spectral clustering of all the PCA elements. Then, we introduce three novel reshaping schemes (namely, Row-wise matrix scheme, Arch-wise matrix scheme, and Curl-wise matrix scheme) of the PCA elements within each cluster. Through extensive experiments and comparisons, we show our model can substantially improve the compression performances on various 3D mesh sequences.