Revisit Point Cloud Quality Assessment: Current Advances and a Multiscale-Inspired Approach

Area of Science:

• Computer Vision
• 3D Data Processing
• Multimedia Quality Assessment

Background:

• Full-reference point cloud quality assessment (PCQA) is crucial for 3D services.
• Traditional methods struggle with point cloud distortions in geometry and attributes.
• Existing PCQA approaches face limitations due to inadequate key point and neighborhood selection.

Purpose of the Study:

• To propose PQI, a novel and efficient metric for indexing point cloud quality.
• To address limitations of current PCQA methods by improving key point and neighborhood considerations.
• To develop a robust quality assessment metric suitable for diverse point cloud applications.

Main Methods:

• PQI utilizes scale-wise key points for uniform distortion perception across the point cloud.
• A multiscale framework is employed to obtain key points and implicitly embed neighborhood information.
• Feature similarity is computed at each scale using geometry and attribute differences, then weighted for the final score.

Main Results:

• PQI demonstrates superior performance across multiple standard PCQA datasets.
• The metric consistently achieves high accuracy in assessing point cloud quality.
• Experimental results validate PQI's effectiveness in various distortion scenarios.

Conclusions:

• PQI offers a simple yet effective solution for point cloud quality assessment.
• The proposed metric exhibits low complexity and flexibility, making it suitable for practical applications.
• PQI advances the field of 3D point cloud quality indexing and analysis.