Аннотация:Stereoscopic movies have experienced a surge in popularity over the past several years.
However, many of them suffer from low technical quality. Manual quality control is prevalent
in the industry, which often leads to unsatisfactory results. This work investigates the ways
to automate the quality control of stereoscopic movies and considers the task of developing
the stereoscopic video quality assessment algorithms. A review and classification of existing
quality assessment methods is presented, a separate class is devoted to methods capable
of assessing stereoscopic movies production quality. Three novel converted video quality
assessment algorithms are proposed in this work: edge-sharpness mismatch metric, cardboard
effect metric and crosstalk potential noticeability metric. A large-scale movie quality analysis
was performed based on the developed methods. The analysis revealed more than 200 scenes
containing noticeable and potentially problematic artifacts. A comparison of more than 100
stereoscopic movies in terms of crosstalk noticeability was conducted.