Stereoscopic quality datasets under various test conditions

摘要

Quality assessment on emerging stereoscopic 3D media is still in an early stage when compared to 2D image/video quality assessment. New characteristics (e.g. specific artifact perception, visual strain, perceived depth) and various requisite factors in stereoscopic systems (e.g. system-introduced crosstalk, screen size, viewing position, scene content, camera baseline) should be taken into account. However, most subjective quality assessments on stereoscopic 3D have been focused on coding artifacts and their influence on the perceived viewing experience. Thus, demand on sharing datasets on perceived visual quality under various requisite factors of stereoscopic systems is increasing. In this paper, we publish three visual quality datasets, including crosstalk stereoscopic (relates crosstalk perception to scene content, crosstalk level and camera baseline), crosstalk auto-stereoscopic (relates crosstalk perception to scene content, crosstalk level and viewing position) and QoE stereoscopic (relates QoE to scene content, camera baseline, screen size and viewing position). Moreover, the laboratory environments, visual stimuli, testing methodologies for generating these datasets and results analysis are also detailed in the paper.