Enhanced Semantic Communications and VVC-Based Hybrid Video Coding System
Prabhath Samarathunga, Yasith Ganearachchi, Thanuj Fernando, Jayasingam Adhuran, Thushan Sivalingam, Nandana Rajatheva, Anil Fernando · IEEE Access · 2026
Video, the dominant media format on the Internet, is pushing conventional video coding techniques to their limits with a wide range of bandwidth-intensive and delay-sensitive applications. This phenomenon is creating renewed interest in learned video coding, and building on the principles of semantic communications, it has the potential to enhance and augment the effectiveness of the video coding process. We propose a semantics-based video codec implemented using autoencoders and the integration of bidirectional hierarchical frame prediction into semantic encoding and decoding. This approach aims to better exploit temporal redundancies in video without requiring extensive training or presharing of trained neural networks. It exceeds the rate-distortion performance of the state-of-the-art in both conventional and learned video codecs for the test scenarios and demonstrates better tolerance to bitstream errors compared to the latest conventional video codecs. The proposed system demonstrates the potential to evolve into a new video coding paradigm that consistently outperforms state-of-the-art conventional and learned video codecs, particularly for emerging applications where semantic communication systems are becoming integral.