LIVE1(LIVE Public-Domain Subjective Image Quality Database)

质量评估研究强烈依赖于主观实验来提供校准数据以及测试机制。毕竟，所有 QA 研究的目标都是做出与人类观察者的主观意见一致的质量预测。为了校准 QA 算法并测试其性能，需要一组图像和视频的数据集，其质量已由人类受试者进行排名。 QA 算法可以在该数据集的一部分上进行训练，并在其余部分上进行测试。
在 LIVE（与德克萨斯大学奥斯汀分校的心理学系合作），进行了一项广泛的实验，以从人类受试者那里获得一些因不同失真类型而失真的图像的分数。这些图像是为了支持通用形状匹配和识别的研究项目而获得的。

Quality assessment (QA) research heavily relies on subjective experiments to provide calibration data and test benchmarks. After all, the overarching goal of all QA research is to generate quality predictions that align with the subjective opinions of human observers. To calibrate QA algorithms and evaluate their performance, a dataset of images and videos whose quality has been ranked by human subjects is required. QA algorithms can be trained on a subset of this dataset and tested on the remaining portion.

At LIVE (in collaboration with the Department of Psychology at The University of Texas at Austin), a large-scale experiment was conducted to obtain quality scores from human subjects for images distorted by various types of distortions. These images were acquired to support a research program focused on general shape matching and recognition.

搜集汇总

背景与挑战

背景概述

LIVE1是一个公开的图像质量评估数据集，由德克萨斯大学奥斯汀分校于2004年发布，包含人类受试者对多种失真类型图像的主观评分，用于校准和测试质量评估算法。该数据集支持通用形状匹配和识别研究，提供主观实验数据以促进图像质量预测模型的开发。