HSOID: Subjective Quality Evaluation on HDR Stereoscopic Omnidirectional Images.

High dynamic range Stereoscopic Omnidirectional Images (HSOI) can provide users with an immersive experience closer to the real world. Compared with the existing omnidirectional images, HSOIs usually have more complex perceptual factors due to the particularities of imaging, coding and visualization, making HSOIs quality assessment more challenging. Up to now, there is no research on subjective quality assessment of HSOIs. To solve this problem, this paper establishes the first subjective HSOI quality Database (denoted as HSOID), in which a set of the distorted HSOIs with encoding distortion and visualization distortion is generated, and these distorted HSOIs are subjectively scored based on the criteria for subjective evaluation. In the HSOID, the representative HSOIs of different natural scenes (indoor, outdoor, day, night, etc.) are created as the original HSOIs, then, they are encoded with the JPEG XT standard into symmetric or asymmetric distorted HOSIs, respectively. For the visualization on the head mounted display with standard dynamic range, the encoded HSOIs undergo five tone mapping operations, respectively, which may result in different degrees of tone mapping distortion. Thus, a set the distorted HSOIs is generated, and they are subjective-scored so that the HSOID with subjective evaluation is established. Finally, by analyzing the subjective quality of different distorted images, the visual perceptual characteristics of HSOI are explored. At the same time, the HSOID is also tested with some representative objective quality metrics of visual contents. The results show that the design of the HSOID is reasonable. The database can be used to benchmark objective quality evaluation methods of the distorted HSOIs.