Image Intrinsic Scale Assessment: Bridging the Gap Between Quality and Resolution

2025-02-10

AI Summary

This paper addresses the lack of systematic quantification of Image Intrinsic Scale (IIS)—the maximum scalable resolution at which human-perceived quality is optimal—by formally introducing the Image Intrinsic Scale Assessment (IISA) task. To support IISA, we construct IISA-DB, the first benchmark dataset comprising 785 expert-annotated image-scale pairs. We further propose WIISA, a novel weakly supervised label generation strategy based on scaling trajectories, which produces high-quality IIS pseudo-labels without requiring pixel-level annotations. Additionally, we adapt and train multiple IQA models for IISA. Experiments demonstrate that WIISA significantly improves prediction accuracy of mainstream IQA models on IISA, yielding an average gain of 12.6%. We publicly release the dataset, source code, and pre-trained models to establish a new benchmark and toolkit for scale-aware image quality modeling.

Technology Category

• Computer Vision: Image and Video Retrieval
• Machine Learning: Scalability of ML Systems
• Search and Optimization: Learning to Search

Application Category

• Search and Retrieval-Augmented AI: Efficiency and scalability of Web search engines
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets
• Systems and Infrastructure for Web, Mobile and WoT: Web performance, measurement, and characterization

Abstract

Image Quality Assessment (IQA) measures and predicts perceived image quality by human observers. Although recent studies have highlighted the critical influence that variations in the scale of an image have on its perceived quality, this relationship has not been systematically quantified. To bridge this gap, we introduce the Image Intrinsic Scale (IIS), defined as the largest scale where an image exhibits its highest perceived quality. We also present the Image Intrinsic Scale Assessment (IISA) task, which involves subjectively measuring and predicting the IIS based on human judgments. We develop a subjective annotation methodology and create the IISA-DB dataset, comprising 785 image-IIS pairs annotated by experts in a rigorously controlled crowdsourcing study. Furthermore, we propose WIISA (Weak-labeling for Image Intrinsic Scale Assessment), a strategy that leverages how the IIS of an image varies with downscaling to generate weak labels. Experiments show that applying WIISA during the training of several IQA methods adapted for IISA consistently improves the performance compared to using only ground-truth labels. We will release the code, dataset, and pre-trained models upon acceptance.

Problem

• Quantifies relationship between image scale and perceived quality.
• Introduces Image Intrinsic Scale (IIS) for highest quality perception.
• Develops WIISA strategy to enhance IQA method performance.

Innovation

• Introduced Image Intrinsic Scale (IIS)
• Developed WIISA weak-labeling strategy
• Created IISA-DB dataset for training