作者: Yan Wang , Guangtao Zhai
DOI: 10.1007/S11220-017-0155-X
关键词: Subjective rating 、 Eye movement 、 Multiple image 、 Subjective quality 、 Data mining 、 Computer vision 、 Eye tracking 、 Subjective video quality 、 Artificial intelligence 、 Image quality 、 Psychology
摘要: Subjective quality assessment serves as the foundation for almost all visual related researches. Size of image databases has expanded from dozens to thousands in last decades. Since each subjective rating therein be averaged over quite a few participants, ever-increasing overall size those calls an evolution existing test methods. Traditional single/double stimulus based approaches are being replaced by multiple tests, where several distorted versions original one displayed and rated at once. And this naturally brings upon question how arrange images on screen during test. In paper, we answer performing viewing with eye tracker different types arrangements. Our research indicates that isometric arrangement imposes less duress participants more uniform distribution fixations movements therefore is expected generate reliable ratings.