KonJND-1k
收藏资源简介:
KonJND-1k数据集由北京大学视觉技术国家工程研究中心创建,主要用于评估压缩图像的用户满意度和机器满意度。该数据集包含1,008张原始图像和76,000张压缩图像,涵盖了不同压缩参数下的图像质量变化。数据集的创建过程涉及通过主观测试收集人类对图像质量的感知数据,并通过多种图像质量评估模型生成机器感知标签。该数据集主要应用于图像压缩优化和图像质量评估领域,旨在提高压缩图像的用户体验和机器分析性能。
KonJND-1k dataset was developed by the National Engineering Research Center for Visual Technology at Peking University, primarily for evaluating user satisfaction and machine satisfaction of compressed images. This dataset comprises 1,008 original images and 76,000 compressed images, covering image quality variations under different compression parameters. The construction of this dataset entails collecting human perceptual ratings of image quality through subjective tests, as well as generating machine-perceived labels using multiple image quality assessment models. This dataset is mainly applied in the fields of image compression optimization and image quality assessment, with the goal of improving the user experience of compressed images and the performance of machine-based image analysis.




