Multi-View Aggregation Transformer for No-Reference Point Cloud Quality Assessment

•We propose a novel multi-view aggregation transformer (MVAT) framework, which aggregates the content and positional context feature among multiple projected images to obtain effective and accurate quality-aware feature. Experimental results show that our MVAT achieves better performance than state-of-the-art NR metrics and comparable performance with classic FR metrics.•To effectively reflect the contribution of different projected images, we propose a content fusion module (CFM), which uses cross-guided learning strategy to reasonably fuse the projected image features by adaptively weighting.•To effectively explore the contextual relationship among the projected images, we propose a bidirectional context fusion module (BCFM), which comprehensively utilizes the feature sequences of forward and backward images to extract the positional context feature among the projected images.