Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

MS-ISSM: Objective Quality Assessment of Point Clouds Using Multi-scale Implicit Structural Similarity

Zhang Chen, Shuai Wan, Yuezhe Zhang, Siyu Ren, Fuzheng Yang, Junhui Hou

TL;DR

This work addresses objective quality assessment for irregular point clouds by eliminating explicit point-to-point correspondences. It introduces MS-ISSM, which uses Radial Basis Function-based implicit representations to compare local feature coefficients across multi-scale chroma, luma, and curvature features, producing a perceptual distortion score via learned mappings. A ResGrouped-MLP with Log-Modulus preprocessing, residual blocks, and channel/global attention translates multi-scale coefficient differences into MOS, achieving superior results on multiple public PCQA benchmarks. The approach demonstrates strong generalization, computational efficiency, and alignment with human visual perception, offering a practical tool for scalable PCQA in real-world applications.

Abstract

The unstructured and irregular nature of point clouds poses a significant challenge for objective quality assessment (PCQA), particularly in establishing accurate perceptual feature correspondence. To tackle this, we propose the Multi-scale Implicit Structural Similarity Measurement (MS-ISSM). Unlike traditional point-to-point matching, MS-ISSM utilizes Radial Basis Functions (RBF) to represent local features continuously, transforming distortion measurement into a comparison of implicit function coefficients. This approach effectively circumvents matching errors inherent in irregular data. Additionally, we propose a ResGrouped-MLP quality assessment network, which robustly maps multi-scale feature differences to perceptual scores. The network architecture departs from traditional flat MLPs by adopting a grouped encoding strategy integrated with Residual Blocks and Channel-wise Attention mechanisms. This hierarchical design allows the model to preserve the distinct physical semantics of luma, chroma, and geometry while adaptively focusing on the most salient distortion features across High, Medium, and Low scales. Experimental results on multiple benchmarks demonstrate that MS-ISSM outperforms state-of-the-art metrics in both reliability and generalization. The source code is available at: https://github.com/ZhangChen2022/MS-ISSM.

MS-ISSM: Objective Quality Assessment of Point Clouds Using Multi-scale Implicit Structural Similarity

TL;DR

This work addresses objective quality assessment for irregular point clouds by eliminating explicit point-to-point correspondences. It introduces MS-ISSM, which uses Radial Basis Function-based implicit representations to compare local feature coefficients across multi-scale chroma, luma, and curvature features, producing a perceptual distortion score via learned mappings. A ResGrouped-MLP with Log-Modulus preprocessing, residual blocks, and channel/global attention translates multi-scale coefficient differences into MOS, achieving superior results on multiple public PCQA benchmarks. The approach demonstrates strong generalization, computational efficiency, and alignment with human visual perception, offering a practical tool for scalable PCQA in real-world applications.

Abstract

The unstructured and irregular nature of point clouds poses a significant challenge for objective quality assessment (PCQA), particularly in establishing accurate perceptual feature correspondence. To tackle this, we propose the Multi-scale Implicit Structural Similarity Measurement (MS-ISSM). Unlike traditional point-to-point matching, MS-ISSM utilizes Radial Basis Functions (RBF) to represent local features continuously, transforming distortion measurement into a comparison of implicit function coefficients. This approach effectively circumvents matching errors inherent in irregular data. Additionally, we propose a ResGrouped-MLP quality assessment network, which robustly maps multi-scale feature differences to perceptual scores. The network architecture departs from traditional flat MLPs by adopting a grouped encoding strategy integrated with Residual Blocks and Channel-wise Attention mechanisms. This hierarchical design allows the model to preserve the distinct physical semantics of luma, chroma, and geometry while adaptively focusing on the most salient distortion features across High, Medium, and Low scales. Experimental results on multiple benchmarks demonstrate that MS-ISSM outperforms state-of-the-art metrics in both reliability and generalization. The source code is available at: https://github.com/ZhangChen2022/MS-ISSM.
Paper Structure (16 sections, 18 equations, 7 figures, 6 tables)

This paper contains 16 sections, 18 equations, 7 figures, 6 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Problem Formulation
  4. Points Correspondence and Perceptual Distortion Measurement
  5. Feature Implicit Representation and Multiscale Perceptual Distortion Calculation
  6. Proposed MS-ISSM
  7. Overview
  8. Multi-scale Features Extraction
  9. Feature Implicit Representation
  10. ResGrouped-MLP Regression
  11. Experimental Evaluations
  12. Datasets and PCQA Metrics under Comparison
  13. Evaluation Criteria
  14. Performance Comparison
  15. Ablation Studies
  16. ...and 1 more sections

Figures (7)

  • Figure 1: The difference between the MS-ISSM and the traditional point-to-point method.
  • Figure 2: The schematic diagram depicts implementing the MS-ISSM solution. (1) Multi-scale features are extracted from the normalized distorted and original point clouds. The chroma, luma, and curvature features of each point cloud are calculated under high-, medium-, and low-quality conditions. (2) The RBF implicit representation is used to calculate the coefficient values for each feature, and multi-scale feature coefficient differences are calculated. (3) The ResGrouped-MLP is designed to map the multi-scale features coefficient differences to perceptual quality scores.
  • Figure 3: Schematic illustration of the Multi-scale implicit feature extraction framework for point cloud.
  • Figure 4: The Proposed ResGrouped-MLP Network.
  • Figure 5: Data Pipeline and Validation Strategy
  • ...and 2 more figures