From Global to Granular: Revealing IQA Model Performance via Correlation Surface

2026-01-29

AI Summary

This work addresses the limitations of conventional image quality assessment (IQA) evaluation metrics—such as PLCC and SRCC—which rely on global scalar summaries that fail to capture local ranking consistency across different mean opinion score (MOS) ranges and MOS differences, and are sensitive to the quality distribution of the test set. To overcome this, the authors propose Granularity-Modulated Correlation (GMC), a novel framework that constructs a three-dimensional correlation surface parameterized by MOS and absolute MOS difference. GMC introduces a granularity modulator and a distribution regulator, integrating Gaussian-weighted correlation, conditional correlation computation, and non-uniform distribution regularization to enable, for the first time, a structured, fine-grained visualization of IQA performance as a joint correlation surface. Experiments demonstrate that GMC reveals model behavior discrepancies invisible to traditional metrics, offering a more informative and reliable paradigm for IQA model evaluation, comparison, and deployment.

Technology Category

• Machine Learning: Evaluation and Analysis
• Computer Vision: Multi-modal Vision
• Search and Optimization: Evaluation and Analysis

Application Category

• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• User Modeling, Personalization and Recommendation: Metrics for user behavior and evaluating success
• Economics, Online Markets and Human Computation: LLM based quality controls for crowd work

Abstract

Evaluation of Image Quality Assessment (IQA) models has long been dominated by global correlation metrics, such as Pearson Linear Correlation Coefficient (PLCC) and Spearman Rank-Order Correlation Coefficient (SRCC). While widely adopted, these metrics reduce performance to a single scalar, failing to capture how ranking consistency varies across the local quality spectrum. For example, two IQA models may achieve identical SRCC values, yet one ranks high-quality images (related to high Mean Opinion Score, MOS) more reliably, while the other better discriminates image pairs with small quality/MOS differences (related to $|\Delta$MOS$|$). Such complementary behaviors are invisible under global metrics. Moreover, SRCC and PLCC are sensitive to test-sample quality distributions, yielding unstable comparisons across test sets. To address these limitations, we propose \textbf{Granularity-Modulated Correlation (GMC)}, which provides a structured, fine-grained analysis of IQA performance. GMC includes: (1) a \textbf{Granularity Modulator} that applies Gaussian-weighted correlations conditioned on absolute MOS values and pairwise MOS differences ($|\Delta$MOS$|$) to examine local performance variations, and (2) a \textbf{Distribution Regulator} that regularizes correlations to mitigate biases from non-uniform quality distributions. The resulting \textbf{correlation surface} maps correlation values as a joint function of MOS and $|\Delta$MOS$|$, providing a 3D representation of IQA performance. Experiments on standard benchmarks show that GMC reveals performance characteristics invisible to scalar metrics, offering a more informative and reliable paradigm for analyzing, comparing, and deploying IQA models. Codes are available at https://github.com/Dniaaa/GMC.

Problem

• Image Quality Assessment
• Correlation Metrics
• Performance Evaluation
• Quality Distribution
• Ranking Consistency

Innovation

• Granularity-Modulated Correlation
• correlation surface
• Image Quality Assessment
• local performance analysis
• distribution regularization