Natural video quality prediction based on temporal and spatial features
Event date:  December 9, 2019 4:00 PM  to 5:00 PM


Due to the wide range of different natural temporal and spatial distortions appearing in user generated video content, blind prediction of natural video quality is a challenging research problem. In this talk, we discuss different approaches for tackling the problem. We also propose a two-level method for modeling video quality: low complexity features, including temporal features derived via statistical analysis of motion vectors, are computed for every second frame, whereas the high complexity spatial features are computed for a representative subset of frames only. We have tested the technique on recently published databases by using both hand-crafted features and deep convolutional neural networks for spatial feature extraction, with promising results.



Jari Korhonen received the M.Sc. (Eng.) degree in information engineering from University of Oulu, Finland, in 2001 and the Ph.D. degree in  telecommunications from Tampere University of Technology, Finland, in 2006. Currently, he is with the School of Computer Science and Software Engineering, Shenzhen University, China, where he has worked as Research Assistant Professor since 2017. From 2001 to 2006, he was Research Engineer with Nokia Research Center, Tampere, Finland. In 2007, he was with École Polytechnique Fédérale de Lausanne, Switzerland, and from 2008 to 2010 with Norwegian University of Science and Technology, Trondheim, Norway. From 2010 to 2017 he was with Technical University of Denmark. His research interests include both telecommunications and signal processing aspects in multimedia communications. Recently, his research focus has been primarily in image and video analysis and enhancement, including development of models for blind prediction of subjective video quality.




University of Konstanz, Powerwall Room, C202

The talk will be transmitted to Stuttgart, VISUS, Powerwall Room