Low-Complexity Patch-based No-Reference Point Cloud Quality Metric exploiting Weighted Structure and Texture Features

2025-03-19

AI Summary

Point cloud compression, transmission, and rendering often introduce perceptually significant distortions, posing challenges for no-reference subjective quality assessment—particularly in balancing generalizability, real-time inference, and reference-free operation. To address this, we propose PST-PCQA, a lightweight no-reference point cloud quality assessment method. PST-PCQA introduces a novel block-wise adaptive weighting fusion mechanism that jointly extracts weighted structural and textural features, integrating local and global contextual information to predict Mean Opinion Scores (MOS). It leverages deep block-based feature extraction, multi-scale pooling, and correlation-aware weighted fusion, followed by an ultra-lightweight fully connected regression head. Evaluated on three major benchmark datasets, PST-PCQA achieves state-of-the-art performance (PLCC > 0.92) with fewer than 50K parameters, enabling real-time deployment on edge devices.

Technology Category

• Computer Vision: 3D Computer Vision
• Machine Learning: Calibration & Uncertainty Quantification
• Knowledge Representation and Reasoning: Qualitative Reasoning

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Web performance, measurement, and characterization
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Graph Algorithms and Modeling for the Web: Algorithms and analysis for incomplete, noisy, or partially observed Web-related graphs

Abstract

During the compression, transmission, and rendering of point clouds, various artifacts are introduced, affecting the quality perceived by the end user. However, evaluating the impact of these distortions on the overall quality is a challenging task. This study introduces PST-PCQA, a no-reference point cloud quality metric based on a low-complexity, learning-based framework. It evaluates point cloud quality by analyzing individual patches, integrating local and global features to predict the Mean Opinion Score. In summary, the process involves extracting features from patches, combining them, and using correlation weights to predict the overall quality. This approach allows us to assess point cloud quality without relying on a reference point cloud, making it particularly useful in scenarios where reference data is unavailable. Experimental tests on three state-of-the-art datasets show good prediction capabilities of PST-PCQA, through the analysis of different feature pooling strategies and its ability to generalize across different datasets. The ablation study confirms the benefits of evaluating quality on a patch-by-patch basis. Additionally, PST-PCQA's light-weight structure, with a small number of parameters to learn, makes it well-suited for real-time applications and devices with limited computational capacity. For reproducibility purposes, we made code, model, and pretrained weights available at https://github.com/michaelneri/PST-PCQA.

Problem

• Evaluates point cloud quality without reference data.
• Uses patch-based analysis to predict Mean Opinion Score.
• Designed for real-time applications with low computational complexity.

Innovation

• Low-complexity patch-based quality metric
• Integrates local and global features
• Light-weight structure for real-time applications