A Projection-based no-reference point cloud quality assessment metric based on perceptual features

• Published in: Displays Vol. 90; p. 103098
• Main Authors: Goswami, Apoorva, Lambat, Naman, Mazumdar, Pramit
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.12.2025
• Subjects: Feature extraction; Point cloud; Projection; Quality metric; Score predictor; Statistical parameters; Statistical parameters; Projection; Feature extraction; Point cloud; Score predictor; Quality metric
• ISSN: 0141-9382
• DOI: 10.1016/j.displa.2025.103098

Rapid progress in 3D sensing technologies has led to the generation of vast amounts of intricate point cloud data. Consequently, there has been a surge in efforts to develop methods for evaluating the quality of these extensive point cloud datasets. Among these methods, no-reference quality assessment has emerged as both challenging and promising, as it does not rely on having a perfect reference for comparison. However, existing 2D no-reference quality metrics often fall short when applied directly to point clouds. They tend to overlook features that are crucial for human visual perception, which is the ultimate driver of how scenes are perceived. This discrepancy highlights the need for a novel quality estimation model that can effectively evaluate point cloud quality without the need for reference points, while also capturing features that align closely with human vision. In this work, a Projection-based no-reference Point Cloud Quality Assessment metric (PB-PCQA) is proposed. The metric introduces a novel projection approach which converts every point cloud into multiple projected 2D patches that are capable of replicating the exploration behavior of a subject. The projected images are used to extract perceptual and distortion-based features. The distribution of pixel intensities in the feature channels is computed from various statistical features. Subsequently, a support vector regressor is employed to estimate the visual quality score for each point cloud. The proposed quality metric is compared with existing state-of-the-art metrics using two publicly available benchmark datasets. The results obtained are satisfactory and indicate an improvement in performance over existing metrics in the literature. In addition, we also performed an ablation study to analyze the importance of each individual feature used in the proposed metric. Here, we also analyze the impact of different distortions on the performance of the proposed metric. •A novel point cloud projection approach which considers all possible viewing angles.•PB-PCQA model includes perceptual and distortion features from projected patches.•PB-PCQA extracts statistical parameters for estimating feature distribution.•Proposed PB-PCQA is compared with FR,RR,NR metrics and results indicate improvement.•Rigorous ablation study to analyse effect of distortion types on visual quality.