Blind Quality Assessment of Point Clouds Based on 3D Co-Occurrence Statistics.

While there has been considerable progress in quality assessment for various types of media, evaluating the quality of point clouds remains a major challenge due to the complexity of the associated applications and the nature of the content. To address this issue, this paper proposes a novel point cloud quality assessment metric based on 3D co-occurrence statistics. The proposed approach involves a voxelization strategy, where the concept of a co-occurrence matrix is extended to 3D to compute the occurrence of a pair of voxels in the 26 possible directions. Selected Haralick features are then computed and concatenated based on the selected color space. A regression step is used to map the features to the ground truth, which is represented by the subjective scores associated with the point cloud models. Experimental results show the effectiveness of using 3D cooccurrence statistics for point cloud quality assessment (CO-PCQA). The proposed metric outperforms most of the recent full-reference and no-reference quality metrics reported in the literature.