A Survey on Objective Quality Assessment of Omnidirectional Images

Xiangjie Sui, Shiqi Wang, Yuming Fang

Abstract: Omnidirectional (360°) image quality assessment (OIQA) has become crucial with the rising popularity of virtual reality (VR). Despite significant efforts to develop objective OIQA methods, there is a lack of comprehensive reviews of these methods, which hinders in-depth understanding and analysis. This paper presents an extensive survey of objective OIQA methods, classifying them into three principal types: 2D-plane-based, sphere-based, and viewport-based methods. We then analyze the limitations of current methods, highlighting challenges such as insufficient consideration of viewing conditions and diverse viewing behaviors. Moreover, this paper suggests potential future research directions, such as multi-modality approaches and personalized assessments. By systematically reviewing existing methodologies and proposing potential advancements, this paper aims to guide future research towards more accurate and efficient OIQA solutions.