JPEG-AIC-4 Example Image Dataset
This dataset contains test images and reconstructed subjective quality scores in JND units. It is provided as an example image quality dataset for the Final Call for Proposals on Objective Quality Assessment (AIC-4).
📦 Download the Image Dataset
Download JPEG-AIC-4 Example Images (ZIP)
🔗 Download Subjective Data (GitHub Repository)
GitHub: JPEG-AIC-4 Datasets
📄 Related Documentation:
- ISO/IEC JTC 1/SC29/WG1 N101157, REQ "Final Call for Proposals on Objective Image Quality Assessment (AIC-4)",
107th Meeting, Brussels, Belgium, April 2025.
- ISO/IEC JTC 1/SC29/WG1 N101211, REQ "AIC-4 Proposal Submission Template",
107th Meeting, Brussels, Belgium, April 2025.
- ISO/IEC JTC 1/SC29/WG1 N101056, ICQ "Common Test Conditions on Objective Quality Assessment",
107th Meeting, Brussels, Belgium, April 2025.
- ISO/IEC JTC 1/SC29/WG1 N101058, REQ "Use Cases and Requirements for Image Quality Assessment v6.0",
106th Meeting, Online, January 2025.
This image dataset contains two folders: one contains high-resolution test images and another with their cropped versions used in the subjective study.
Folder Descriptions:
- full_resolution_images: Contains full-resolution test images and their corresponding sources, compressed using six codecs: AVIF (1), JPEG (2), JPEG 2000 (3), JPEG XL (4), VVC (5), and JPEG AI (6).
- PTC_images: Contains cropped versions of the test images used in the triplet comparison tests for subjective quality evaluation.
The JPEG AIC-3 testing methodology was used for both the subjective study and the JND scale reconstruction. For a detailed explanation of the testing methodology and the reconstruction process, please refer to the following publications:
- Jenadeleh et al., Subjective Visual Quality Assessment for High-Fidelity Learning-Based Image Compression, 2025. arXiv:2504.06301
- Testolina et al., Fine-Grained Subjective Visual Quality Assessment for High-Fidelity Compressed Images, 2024, accepted at the Data Compression Conference (DCC). arXiv:2410.09501
You can find more information on JPEG AIC on JPEG.org.