English

Learned transform compression with optimized entropy encoding

Machine Learning 2021-05-05 v2

Abstract

We consider the problem of learned transform compression where we learn both, the transform as well as the probability distribution over the discrete codes. We utilize a soft relaxation of the quantization operation to allow for back-propagation of gradients and employ vector (rather than scalar) quantization of the latent codes. Furthermore, we apply similar relaxation in the code probability assignments enabling direct optimization of the code entropy. To the best of our knowledge, this approach is completely novel. We conduct a set of proof-of concept experiments confirming the potency of our approaches.

Keywords

Cite

@article{arxiv.2104.03305,
  title  = {Learned transform compression with optimized entropy encoding},
  author = {Magda Gregorová and Marc Desaules and Alexandros Kalousis},
  journal= {arXiv preprint arXiv:2104.03305},
  year   = {2021}
}

Comments

Published as a workshop paper at ICLR 2021 neural compression workshop