VideoSet

A new methodology to measure coded image/video quality using the just-noticeable-difference (JND) idea was proposed in [1]. Several small JND-based image/video quality datasets were released by the Media Communications Lab at the University of Southern California in [2, 3]. In this work, we present an effort to build a large-scale JND-based coded video quality dataset. The dataset consists of 220 5-second sequences in four resolutions (i.e., 1920x1080, 1280x720, 960x540, and 640x 360). For each of the 880 video clips, we encode it using the H.264 codec with QP=1, ..., 51 and measure the first three JND points with 30+ subjects. The dataset is called the ``VideoSet", which is an acronym for ``Video Subject Evaluation Test (SET)". This work describes the subjective test procedure, detection and removal of outlying measured data, and the properties of collected JND data. Finally, the significance and implications of the VideoSet to future video coding research and standardization efforts are pointed out. All source/coded video clips as well as measured JND data included in the VideoSet are available to the public here.[1] Joe Yuchieh Lin, et al., Experimental design and analysis of JND test on coded image/video. Proc. SPIE, Applications of Digital Image Processing XXXVIII, September 22, 2015.[2] L. Jin, et al., Statistical study on perceived JPEG image quality via MCL- JCI dataset construction and analysis, in: IST/SPIE Electronic Imaging, International Society for Optics and Photonics, 2016.[3] H. Wang, et al., MCL-JCV: A JND-based H.264/AVC video quality assessment dataset, in: 2016 IEEE International Conference on Image Processing (ICIP), 2016, pp. 1509–1513.Update:As requested by some users, we prepared a copy of videoSet hosted on BaiduYun. Please try both servers, i.e. AWS in this page and BaiduYun and find the one that works better for you. Date: 02/14/2017.A typo fixed. The second header in 960x540_1st.csv, changed from anchor to jnd. Date: 03/01/2017Citation was updated in README_V2. The dataset paper was accepted by JVCIR and PDF attached. Date: 05/09/2017.

文献[1]提出了一种利用恰可察觉差（just-noticeable-difference, JND）理念评估编码图像/视频质量的新型方法论。美国南加州大学媒体通信实验室于文献[2, 3]中发布了数个基于JND的小型图像/视频质量数据集。本研究致力于构建一个大规模的基于JND的编码视频质量数据集。
该数据集包含4种分辨率（即1920×1080、1280×720、960×540及640×360）下的220段5秒视频序列。针对这880个视频片段，我们采用H.264编解码器，以QP=1至51的参数进行编码，并邀请30余名受试者完成前三个JND点的测量工作。本数据集被命名为"VideoSet"，其全称为"Video Subject Evaluation Test (SET)"。
本研究详细阐述了主观测试流程、异常测量数据的检测与剔除方法，以及所采集JND数据的特性。最后，本文指明了"VideoSet"对未来视频编码研究与标准化工作的重要意义与应用价值。"VideoSet"中包含的所有原始/编码视频片段及实测JND数据均可在此公开获取。
[1] Joe Yuchieh Lin 等，《编码图像/视频JND测试的实验设计与分析》，发表于SPIE会议：《数字图像处理应用第38届研讨会》，2015年9月22日。
[2] L. Jin 等，《基于MCL-JCI数据集构建与分析的JPEG图像感知质量统计研究》，收录于：国际光学与光子学学会，IST/SPIE电子成像会议，2016年。
[3] H. Wang 等，《MCL-JCV：一个基于JND的H.264/AVC视频质量评估数据集》，发表于：2016年IEEE国际图像处理会议（ICIP），2016年，第1509–1513页。
更新说明：应部分用户要求，我们在百度网盘上准备了"VideoSet"的备份副本。请尝试本页面提供的AWS服务器与百度网盘两个渠道，选择最适配您的下载方式。日期：2017年2月14日。
修正一处拼写错误：960x540_1st.csv中的第二个表头，将anchor更改为jnd。日期：2017年3月1日。
更新了README_V2中的引用信息。本数据集相关论文已被《Journal of Visual Communication and Image Representation (JVCIR)》录用，并附带有PDF文件。日期：2017年5月9日。

搜集汇总

背景与挑战

背景概述

VideoSet是一个大规模JND（Just Noticeable Difference）视频质量评估数据集，包含220个5秒视频序列，覆盖四种分辨率，并使用H.264编码器以不同QP值编码。数据集通过主观测试测量JND点，适用于视频编码研究和标准化工作。