THE FIRST COMPREHENSIVE DATASET WITH MULTIPLE DISTORTION TYPES FOR VISUAL JUST-NOTICEABLE DIFFERENCES

本数据集名为‘THE FIRST COMPREHENSIVE DATASET WITH MULTIPLE DISTORTION TYPES FOR VISUAL JUST-NOTICEABLE DIFFERENCES’，由哈尔滨工程大学、南洋理工大学和复旦大学合作创建。数据集包含106张源图像和1,642张JND映射图，涵盖25种不同的图像失真类型。创建过程中，研究者首先从现有的图像质量评估（IQA）数据集中选择失真图像作为JND候选，然后通过众包主观评估进行精细JND映射选择。此数据集旨在解决现有JND模型仅限于压缩失真类型的问题，通过包含多种失真类型，扩展JND模型的应用范围，如优化图像传输过程和提高社交媒体图像质量。

This dataset, named "THE FIRST COMPREHENSIVE DATASET WITH MULTIPLE DISTORTION TYPES FOR VISUAL JUST-NOTICEABLE DIFFERENCES", was co-developed by Harbin Engineering University, Nanyang Technological University, and Fudan University. It contains 106 source images and 1,642 JND maps, covering 25 distinct image distortion types. During its development, researchers first selected distorted images from existing image quality assessment (IQA) datasets as JND candidates, then conducted fine-grained JND map selection via crowdsourced subjective evaluation. This dataset aims to address the limitation of existing JND models, which are restricted to compression-related distortion types. By incorporating diverse distortion types, it expands the application scope of JND models, such as optimizing image transmission processes and enhancing the quality of social media images.