Panoramic and immersive images are becoming quite popular, because of the recent introduction of a variety of powerful virtual reality (VR) devices. A topic of interest in this context is the quality assessment research of immersive VR images which is strongly
related to the dataset used in experiments. Thus, we develop a model that expresses the joint impact of spatial resolution s and JPEG compression quality factor qf on immersive image quality, and try to make our database of immersive images with few stitching errors and MOS data publicly available as a resource for the picture quality and virtual reality communities.
The goal of our study is to understand the joint impact of compression and spatial resolution on immersive image quality. All the provided images are of 4096×2160 spatial resolution and compressed without loss. The database embodies thousands of snapshots in order to greatly alleviate stitching errors and other distortions. The snapshots composing each immersive image are separated by uniform angular increments of 2-degrees in both the meridian and parallel dimensions. We envision that these ground-truth samples will prove to be useful for future subjective and objective quality assessment research studies.
The predicted and subjective MOS data of 15 images (Snowfield, Forest, Apartment, Desert, Telestudio, Darkroom, Parlour, Brightroom, Brickhouse, Chamber, Seaside, Fair, Race, Stadium, Pavilion) for cross validation.
Img f-j can be found in the upper part of this page. The remaining 10 images (img a-e and img k-o) can be downloaded from the following links.