A. Deniz Aladagli
Omnidirectional Video Encoding and Delivery Using Advanced Processing Techniques
A. Deniz Aladagli is a PhD Fellow and Marie Curie Early Stage Researcher in the Institute for Digital Technologies. His research is focused on the provision of quality of service in virtual reality (VR) video streaming services, using advanced video analysis and processing techniques.
Deniz received his B.Sc. degree in Computer Science and Engineering from Sabanci University. Later, he continued his studies to earn his M.Sc. degree in Game Technologies from Middle East Technical University where he worked on deferred rendering methods. Before starting his doctoral research, he was working as a software engineer in Simsoft, developing serious games and simulations.
PhD research description
Deniz's doctoral research looks into the state of the art in technologies related to the streaming of omnidirectional or virtual reality (VR) videos. He is in the process of developing a novel framework for the efficient delivery of high quality VR video services.
A VR video is a spherical video where all possible viewing directions are available to a viewer, which can be interactively selected in real time using a head mounted display (HMD). Due to their large size, a very high bandwidth that is not commonly available in the average household is required to stream high quality VR videos. Previous research establishes that streaming only the visible portion of the VR video is sufficient for a good viewer experience, since only a portion of the whole spherical video is visible during a VR video viewing. However, the user's viewing direction (head orientation) is not known beforehand. As there are delays that occur from network conditions and other processes related to video services, partial streaming methods suffer from problems such as missing parts or low quality in the viewer's visual field. There are also other factors which effect the complete delivery pipeline such as storage requirements, encoding and decoding limitations and processing artifacts.
The proposed framework takes a holistic approach. Multiple aspects are considered such as quality adaptation, video partitioning, efficient storage, perception based video processing, and view direction prediction. Most recently, video saliency analysis is being investigated in order to achieve more accurate personal view direction prediction during VR video viewings.
Awards, grants or scholarships received
Marie Curie Cloudscreens grant, Loughborough University London, 2015-2018.
Interests and activities
Deniz has an avid interest in games, which leaked into both his professional life and studies. He also enjoys wide variety of sports such as skating, skiing and sky diving.