JPEG AI Image Compression Visual Artifacts: Detection Methods and Dataset
This addresses the reliability issue of neural image compression for applications where visual quality is critical, though it is incremental as it focuses on detection rather than solving the artifact problem itself.
The paper tackles the problem of visual artifacts in learning-based image compression by proposing methods to detect, localize, and quantify three types of artifacts (texture/boundary degradation, color change, and text corruption) that occur specifically in neural compression but not in traditional codecs. They applied these methods to create a dataset of 46,440 validated artifacts from 350,000 images, which can be used for testing and improving neural image codecs.
Learning-based image compression methods have improved in recent years and started to outperform traditional codecs. However, neural-network approaches can unexpectedly introduce visual artifacts in some images. We therefore propose methods to separately detect three types of artifacts (texture and boundary degradation, color change, and text corruption), to localize the affected regions, and to quantify the artifact strength. We consider only those regions that exhibit distortion due solely to the neural compression but that a traditional codec recovers successfully at a comparable bitrate. We employed our methods to collect artifacts for the JPEG AI verification model with respect to HM-18.0, the H.265 reference software. We processed about 350,000 unique images from the Open Images dataset using different compression-quality parameters; the result is a dataset of 46,440 artifacts validated through crowd-sourced subjective assessment. Our proposed dataset and methods are valuable for testing neural-network-based image codecs, identifying bugs in these codecs, and enhancing their performance. We make source code of the methods and the dataset publicly available.