Multidimensional multiscale parser compression of 3D meshes

Abstract

In this paper, we proposed two 3D mesh compression methods based on the “multidimensional multiscale parser”. The main idea of the first method is to transform the 3D object into a 2D image using the geometry image [3]. The second method is to project the wavelet transform of the object into a 2D image. The coding is processed using the MMP on these 2D images. At low bitrates, (from 0.3 to 1 bit/vertex) we have a better result in the order of 0.5 dB than the simple wavelet transform method [1]. Moreover, our method consists in processing the data progressively during acquisition while reducing considerably the memory. The problem of scan-based processing arises when compressing very large volumes of data using a minimum of memory resources. Knowing that the 3D meshes with a high degree of precision have sizes exceeding several million points, the difficulty of processing quickly arises related to this kind of data. With our scan-based method, we were able to reach levels memory even smaller than in the wavelet transform method (WT) of [1] with a better compression quality.