Abstract:
Presently, the data traffic is increasing for video conferencing, online education, gaming and watching videos on Netflix, Amazon Prime, YouTube and other OTT platforms. And, the service users are always demanding high definition and high-quality video facilities day by day. However, in order to transmit video data across the Internet's constrained bandwidth effectively, video compression is a necessary task. In last few decades, various video compression algorithms, such as non-learning and learning were standardized. But still some improvements are needed for effective video related services. We propose a deep learning based Deep Recurrent Auto Encoders (DRAE) approach which contain various modules for implementing an efficient video compression technique. The experimental outcome shows our model achieves state-of-the-art learned video compression performance in terms of both PSNR and MS-SSIM.
Page(s):
5949-5959
DOI:
DOI not available
Published:
Journal: Journal of Theoretical and Applied Information Technology, Volume: 100, Issue: 20, Year: 2022
Keywords:
Compression
,
Recurrent Auto Encoders
,
Video
,
Deep Neural Networks