Online-learning based video reconstruction for adaptive bitrate video streaming

The bitrate of streaming videos can vary over time due to channel conditions, and videos are subject to heavier compression and downsampling when the channel conditions are poorer, which may affect the quality of the video. We proposed methods based on online learning to reconstruct these severely degraded videos when the channel is bad using the temporal self-similarity of videos when the channel is good.

Joshua Peter Ebenezer
Joshua Peter Ebenezer
Graduate Researcher and Assistant Director of LIVE

My research interests include image and video quality assessment, natural video statistics, and deep learning.