Contents I. Introduction
In recent years, omnidirectional videos – also referred to as 360° or panoramic videos – have provided an opportunity to, e.g., explore videos of places in a more immersive way than using traditional 2D videos. The camera technology to record these immersive contents has significantly evolved and improved, thus enabling the creation of high-quality 360° or panoramic videos. One of the first more widely adopted omnidirectional video cameras, the GoPro Omni, was launched in 2016 [1] and can be equipped with six GoPro cameras. The Insta360 Pro 2 was released as a professional VR camera in 2018, providing 8K stereoscopic video resolution [2]. In 2021, the Kandao Obsidian Pro 360° camera was released, providing up to 12K stereoscopic video resolution [3]. Stereoscopic, high-resolution 360° videos are a promising technology that enables the user to perceive a more immersive experience of real-life places and scenes via omnidirectional videos. Until now, very few studies are available that investigate the effect of stereoscopic vs. non-stereoscopic 360° video considering quality, or even additional Quality of Experience (QoE) indicators or constructs such as user behaviour, immersion, and visual comfort. Hence, the goal of this study is to analyse to which extent the exploration behaviour, level of immersion, video quality, and visual comfort of people consuming 8K 360° video are different for stereoscopic versus non-stereoscopic contents. With this aim, in Section II, selected previous studies on the evaluation of QoE of a stereoscopic video viewing session in general, and stereoscopic 360° videos in particular are identified. Following this, in Section III the creation process of the dataset using the Kandao Obsidian Pro 360° camera is described, as no 360° video QoE or source dataset is publicly available that also provides the stereoscopic 360° video content. Within this study, we present the dataset “AVT360-8K” of five 360° video scenes, with each video having a duration of around 30s. The stitched 360° source video material is made publicly available and was encoded with the visually lossless Apple ProRes 422 HQ codec, directly providing both the non-stereoscopic and stereoscopic (single view, half over-under) versions, together with the data obtained during the subjective test The subsequent HEVC (libx265) encoding with a CRF of 20 was necessary to enable a proper play out of the stereoscopic 360. video sequences and because Mistika VR does not offer CRF-based encoding settings for HEVC. The encoded videos are also included in the database. In Section IV, we describe the test method we used to investigate the influence of non-stereoscopic vs. stereoscopic video on exploration behaviour, immersion, video quality and visual comfort. To this aim, a subjective test with 30 subjects was conducted, using a within-subject design where all subjects watched all the videos. During playout, head-rotation data was recorded for exploration behaviour analysis. After each video, the perceived video quality, level of immersion, and visual comfort were evaluated having subjects provide ratings on four respective individual scales. In Section V, results from the subjective test are presented and analyzed, and in Section VI, the study is summarized and concluded.
https://github.com/Telecommunication-Telemedia-Assessment/AVT-360-8K
