Deep learning for video compression

Video compression has been traditionally addressed using signal processing techniques. However, recent advances in machine learning can be exploited to design data-driven building blocks for next-generation video codecs. This activity has the objective of employing basic as well as advanced types of neural networks, e.g. generative adversarial networks and their bidirectional extensions, in order to improve the performance of existing codecs, or define new and innovative coding paradigms that can significantly boost the rate-quality performance with respect to the state of the art.

ERC Sector:

  • PE7_7 Signal processing


  • Video codecs
  • Data compression

Research groups