As part of the TRANSMIXR work on metrics and evaluation, the Distributed and Interactive Systems (DIS) group from CWI presented the work “PointPCA+: Extending PointPCA objective quality assessment metric”, and “QAVA-DPC: Eye-Tracking Based Quality Assessment and Visual Attention Dataset for Dynamic Point Cloud in 6 DoF” at the 2023 IEEE International Conference on Image Processing (ICIP) and the 2023 IEEE International Symposium on Mixed and Augmented Reality (ISMAR), respectively.
With the rise of immersive XR applications, there is a growing interest in high-fidelity representations of real-world objects and scenes, particularly point clouds. The first paper, “PointPCA+: Extending PointPCA objective quality assessment metric” explores the use of PCA-based descriptors for full-reference quality assessment of point clouds, as an extension of the PointPCA metric. The algorithm competed in three of the five tracks of the Grand Challenge “Point Cloud Visual Quality Assessment”, reaching the podium in all the tracks. A summary of all the efforts of DIS on point cloud objective metrics from can be found here: https://www.dis.cwi.nl/point-cloud-objective-quality-metrics/.
Three-dimensional contents such as point cloud can never be visualised in their entirety at once, due to the natural occlusions that occur in such contents. Thus, understanding how users perceive such contents, and what features of the content are more salient in terms of visual attention, is fundamental to optimise compression and delivery without compromising the user experience. However, due to limitations in rendering and eye-tracking technologies, datasets that explore visual attention for dynamic point cloud contents are currently not available. The second paper, “QAVA-DPC: Eye-Tracking Based Quality Assessment and Visual Attention Dataset for Dynamic Point Cloud in 6 DoF”, fills this gap by proposing a visual attention dataset comprised of eye gaze data acquired in VR in the context of a subjective quality assessment study. The dataset, which is publicly available, offers raw gaze data alongside visual attention maps for both pristine and distorted point cloud contents, and can be used for gaze data processing, visual saliency prediction, and visual saliency-assisted objective quality assessment. The data can be found here: https://github.com/cwi-dis/ISMAR_PointCloud_EyeTracking
Scientific articles:
- Zhou, X., Viola, I., Alexiou, E., Jansen, J., and Cesar, P., 2023, October. QAVA-DPC: Eye-Tracking Based Quality Assessment and Visual Attention Dataset for Dynamic Point Cloud in 6 DoF. In Proceedings of the IEEE International Symposium on Mixed and Augmented Reality (ISMAR). IEEE
- Zhou, X., Alexiou, E., Viola, I., and Cesar, P., 2023, October. PointPCA+: Extending PointPCA objective quality assessment metric. In 2023 IEEE International Conference on Image Processing (ICIP). IEEE