Eye-Tracking Data Extension for Omnidirectional Image Quality Assessment Database (OMNIQAD)

This paper extends the available OMNIQAD database with a general analysis of eye-tracking data obtained for various omnidirectional images from this database. Eye-tracking data were collected during subjective experiments using a head-mounted display with eye-tracking capability. A total of 29 subjects (19 males and 10 females) with age ranging from 18 to 25 years were assessed. Attention is devoted to the description of the data collection process using the Unity-designed application and the possibilities of data processing with respect to finding notable regions of interest in presented omnidirectional images, which are briefly discussed. The paper also points out possible usage of the obtained data, such as refining or retraining the currently available models for perceptual-based video coding with respect to compression artifacts, defining a model for a new objective quality metric for omnidirectional images or training and testing attention models for omnidirectional images.