Object-based Compression of 3D Animation Geometry

Abstract

In multi-object based three-dimensional (MO-3D) animations, the geometry and connectivity data of the constituent objects change across the frames. These data require large disk space for storage and high bandwidth for real-time transmission over the network compared to single-object based 3D (SO-3D) animations. The existing geometry compression methods for 3D animations consider each frame as a single 3D object for compression. This global consideration of the multiple 3D objects as a single object does not exploit fully the local spatio-temporal redundancies exhibited by the individual 3D objects for the animation duration. This work proposes an efficient geometry compression method for MO-3D animations by segmenting the animation geometry data across the frames into objects followed by applying the principal component analysis (PCA) based compression on the geometry data of the segmented individual 3D objects. To compare the performance of the proposed method, a few MO-3D animations are created using the existing public domain SO-3D animations. The simulation results show that the proposed object-based PCA (OPCA) method gives better performance in terms of objective and subjective quality metrics compared to the other PCA based geometry compression methods applied globally on the MO-3D animations.