Reproduce “JPEG AI Image Compression Visual Artifacts: Detection Methods and Dataset” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Daria Tsereh, Mark Mirgaleev, Ivan Molodetskikh, Roman Kazantsev, Dmitriy Vatolin
Researchers from MSU developed five specialized methods to automatically detect specific visual artifacts (texture, color, text, boundary distortions) introduced by JPEG AI image compression, which existing metrics poorly capture. They also created and subjectively validated a dataset of 46,440 such artifacts, providing crucial tools and data for improving neural codecs.
Key insights
- Neural network-based image compression codecs generate distinct types of visual artifacts that are poorly captured by traditional objective image quality metrics.
- Specialized detection methods, designed to differentiate neural artifacts from general compression distortions, are necessary for accurately assessing the perceptual quality of learning-based codecs.
- Rigorous subjective validation is essential for ensuring that automatically detected artifacts are perceptually relevant and align with human visual experience, making datasets valuable for practical application.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.