Content area
Due to the widespread use of point cloud, the demand for compression and transmission is more and more prominent. However, this cause various losses to the point cloud. It is necessary for application to evaluate the quality of point cloud. Therefore, we propose a new point cloud quality assessment (PCQA) metric named statistical information similarity (SISIM). First, we preprocess point cloud (PC) by scaling based on density and then project PC into texture maps and geometry maps. In addition, the SISIM based on Natural Scene Statistics (NSS) is proposed as texture features under the premise of proving that the texture maps meet NSS. Furthermore, we propose to extract geometry features based on local binary patterns (LBP) on account of the phenomenon that LBP maps of geometry images vary with different distortions. Finally, we predict the quality of PCs by fusing texture features with geometry features. Experiments show that our proposed method outperforms the state-of-the-art PCQA metrics on three publicly available datasets.