.png){kind=icon}
.png){kind=icon}
Return
End-to-end learned video compression: A comprehensive review
.png){kind=icon}
PRE
AI
DOI：10.1016/j.neucom.2026.133839.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
The explosive growth of video data presents an urgent need for more efficient compression techniques. Although conventional video coding standards have undergone decades of iterative optimization, they are gradually approaching their theoretical limits. End-to-end learning video compression, as an emerging paradigm, utilizes deep neural networks to jointly optimize various compression modules, demonstrating enormous potential for development. This review aims to provide a comprehensive overview of the latest developments in this field, systematically review the evolution of inter-frame coding architecture, and clearly outline the development trajectory of video compression technology from early residual coding to new video compression paradigms. In some scenarios, end-to-end learning of video compression performance has surpassed the versatile video coding test model, while also highlighting ongoing challenges, including high computational complexity, difficulty in handling complex motion, and a lack of standardization. The review offers an in-depth analysis of the performance and complexity of typical methods, summarizes the current challenges and future development directions, and aims to provide valuable insights for future research in this domain.
Keywords:
end-to-end learning
video compression
deep neural networks
inter-frame coding
computational complexity
Journal
IF:
6.5
Papers:
2.5W
Citations:
6.5W
Organization
View more organizations