LFDD Extended: Subjective Quality Assessment under Light Field Display Viewing

The presented dataset, LFDD Extended: Subjective Quality Assessment under Light Field Display Viewing, provides new subjective quality scores for a set of light field images sourced mainly from the original LFDD dataset and from LFQD. The primary distinction is that this subjective evaluation was conducted using a Light Field Display, rather than a traditional 2D display. This provides an assessment of quality under full autostereoscopic viewing conditions. The quality assessment employed a Multi-Stage Double-Stimulus Impairment Scale (MS-DSIS) methodology, adapted from ITU-R BT.500 recommendations. For each test sequence, subjects were presented with an unimpaired reference and its distorted counterpart. The evaluation unfolded in three sequential stages to ensure granular data collection: Detection of Difference: Participants indicated if any difference was perceived between the two stimuli. Assessment of Annoyance: If a difference was detected, participants determined if the observed difference was in any way annoying. Severity of Annoyance: If the distortion was deemed annoying, participants indicated the severity of that annoyance. The presented dataset includes raw scores, as well as Mean Opinion Scores (MOS). The repository also includes code used in the evaluation (both for data acquisition and data analysis).