讲座主题：
Evaluation of Image Quality Assessment Models
主讲人姓名及介绍：
马柯德现任香港城市大学计算机科学系助理教授。2012 年毕业于中国科学技术大学，获工学学士学位，2017 年毕业于加拿大滑铁卢大学，获电子与计算机工程专业博士学位。主要从事感知图像处理与计算视觉相关研究。
报告摘要：
Image quality assessment (IQA), a long-standing task in the field of image and multimedia processing, has evolved rapidly in the past two decades, and has also gained increasing attention from both academic and industry for its broad applications. Conventional IQA model comparison generally follows a three-step approach. First, pre-select a number of images to form the test set. Second, collect the mean opinion score (MOS) for each image in the test set to represent its true perceptual quality. Third, rank the competing models according to their goodness of fit on the test set. The one with the best result is declared the winner. In this talk, we will discuss the limitations of this conventional method in terms of the representativeness of test samples and the risk of overfitting. We will then introduce a series of alternative IQA model comparison methods, and put them in a boarder context of computer vision.
图像质量评估（IQA）是图像和多媒体处理领域的一项长期任务，在过去20年中发展迅速，也因其广泛的应用而越来越受到学术界和工业界的关注。传统的IQA模型比较一般采用三步法。首先，预先选择一些图像，组成测试集。第二，收集测试集中每张图像的平均意见得分（MOS），以代表其真实的感知质量。第三，根据测试集的拟合度对竞争的模型进行排名。具有最佳结果的模型被宣布为获胜者。在本讲座中，我们将讨论这种传统方法在测试样本的代表性和过度拟合的风险方面的局限性。然后，我们将介绍一系列替代的IQA模型比较方法，并把它们放在计算机视觉的背景中。
