JPEG AIC2026: A large-scale dataset for fine-grained assessment of image coding

2026-07-24

AI Summary

This work addresses the lack of large-scale benchmarks supporting fine-grained analysis in image compression quality assessment, particularly for learning-based methods. To this end, we introduce the AIC2026 dataset, comprising 70 source images selected through semantic clustering and manual curation, from which 9,618 distorted images were generated using eight traditional and four learning-based codecs across 17 configurations. The dataset spans a fine-grained quality gradient from 0.2 to 4.0 JNDs (Just Noticeable Differences), with perceptually uniform sampling guided by ColorVideoVDP. AIC2026 is the first benchmark to systematically include novel artifacts introduced by learning-based compression. Comprehensive evaluation using 24 traditional and 12 learning-based image quality assessment (IQA) methods reveals substantial discrepancies among existing metrics in capturing fine-grained perceptual differences, establishing AIC2026 as a critical benchmark for future research in compression and quality evaluation.

Abstract

Recent advances in conventional and learning-based image coding have increased the demand for benchmark datasets that support fine-grained assessment of compressed image quality, particularly for learning-based image compression methods. This paper introduces Assessment of Image Coding 2026 (AIC2026), a large-scale dataset for high-fidelity image compression containing 70 source images selected from 2,787 candidates using semantic clustering, inter-metric disagreement among objective image quality assessment (IQA) methods, and manual inspection and refinement. The dataset covers a wide range of compression artifacts produced by eight conventional and four learning-based codecs across 17 coding configurations. Each source image is encoded using seven codecs. For each source-codec pair, decoded images are provided at 20 perceptually spaced distortion levels, corresponding approximately to 0.2-4.0 just-noticeable difference (JND) units using the ColorVideoVDP (CVVDP) metric for distortion estimation, yielding 9,618 distorted images. This fine-grained sampling enables analysis of rate-distortion behavior and objective metric evaluation for subtle quality differences across a wide range of compression artifacts. We report an extensive objective analysis using 24 conventional and 12 learning-based IQA methods. The results show substantial disagreement among current IQA methods for fine-grained quality differences, particularly for artifacts introduced by learning-based codecs. The complete dataset is publicly available at https://doi.org/10.18419/DARUS-6156.

Problem

• image compression
• fine-grained quality assessment
• learning-based codecs
• compression artifacts
• image quality assessment (IQA)

Innovation

• fine-grained assessment
• learning-based image compression
• just-noticeable difference (JND)
• image quality assessment (IQA)
• compression artifacts